Thanks for the review! I've got a few suggestionss/questions though.
I've been out of the loop for a while and looking at Sandforce and other newer drives. So some Sandforce drives have the I/O limitations that were intended for the SF-1200 and some have SF-1500 like performance?
I'm surprised the Corsair F40 does so well. I thought the lower capacity drives performed worse than the 120GB versions, but it holds up really well. Or is this just a special case with the 40GB one and the 60GB is worse than the 40GB one? The 60GB Sandforce are also much better value than the 40GB ones, 50% more capacity at >20% more price. I find it strange you didn't include them and mentioned the 64GB C300 to be the value drive at their price point.
I'm pretty sure the Indilinx 60GB, the unlocked SF-1200 60GB and the X25-M 80GB are the most popular drives out there, which makes them great reference points, but they're not in the review. The former two are not on Bench either unfortunately. Do you have some around and could test them?
You tested the Crucial drives on the ICH10R, right?
Also, I would appreciate some blog posts or small articles about developments with newer FWs. I remember the FW development improving the Indilinx drives significantly, and am always wondering how accurate your older reviews still are given there are newer FWs out now. It would also be nice if you could list the tested FW version in Bench.
It would also be great if you could look at SSD performance in Macbooks. I want to put one in my Macbook Pro (Late 2008), but all the talk of freezing has me hesitating, and I haven't seen an in-depth look at this issue. Is it related to what kind of SSD you use, and does it make a difference wheter you have a late 2008 or mid 2009? It would also be interesting to see how tha lack of TRIM actually affects different drives under OSX.
I posted and asked on many forums and did not found an answer to my MBA ( Macbook Air ) question.
Why did the MBA do so well in its test, while its performance data were below the King of SSD controller, Sandforce?
No one could answer. There were number of review pointing out that their MBA actually feels snappier then their Macbook with Sandforce or Intel SSD. Although this is impossible when first heard, numerous other review site seems to confirm similar findings. Of cause there is no way to test it out since the MBA does not have an regular SATA slot.
Now this article actually print out the truth. The same Toshiba SSD controller used in MBA's SSD, is top of the chart in BOTH Synthetic Benchmarks and Real World usage ( Anand Bench ) Benchmarks. What we have been talking as the Holy Grail of SSD Performance Delta, the 4K Random Read / Write, didn't matter when Toshiba was literally the bottom of the chart in those test.
There is a reason why Apple choose a Inferior part ( To us at the time ) instead of Sandforce. The argument for choosing it because of always On GC doesn't make sense, since Sandforce has the same capabilities within the Firmware itself.
One reason would be Toshiba is a NAND manufacture itself, and buying NAND and Controller directly from Toshiba would be cheaper. The other reason being Toshiba ( properly involve Sandisk as well since their JV ) had a controller chip which is very fast.
There has to be an missing pieces in our performance test, something that these companies knows and we dont.
I'd like to see more real world tests - and I don't consider the AnandTech Storage Bench to be "real world" - it's still a bench, like PCMark.
But yes, you are right: synthetic tests tell us little about the performance you actually get from an SSD. There are more unknown variables than we think. You may see big differences in benches like SysMark or PCMark - and even bigger differences in synthetic tests like AS SSD or even IOMeter. But these scores tell us little about REAL world performance - and with REAL I mean things like: - "How long does it take to start Photoshop while running Virus Scan?" - "How long does it take to start iTunes while unzipping a not-so-much-compressed zip-file?"
That's the things I care about. And interestingly, you often get COMPLETELY different results, than what you would think when looking at synthetic tests or "half-synthetic"-tests like PCMark or AnandTech Storage Bench.
I used to run a lot of those types of tests, however I quickly found that if you choose your iometer and other benchmarks appropriately they don't any new data. And often times they are so limited in their scope (e.g. launch an application with virus scan in the background) that you don't see any appreciable differences between drives. Most high end SSDs are fast enough to do most of these types of tasks just as quickly as one another. It's when you string a bunch of operations together and look for cumulative differences in response time or performance that you can really begin to see which one is faster. These types of scenarios are virtually impossible to perform with consistency by hand, that's where our test suite comes in.
AnandTech Storage Bench, PCMark and even SYSMark do what is necessary - they measure performance of a more complex usage case. PCMark Vantage is a great showcase of truly light workload I/O performance, while SYSMark is more CPU bound and shows you how small the differences can be. Our own benchmark offers a more modern set of usage models (we actually do run photoshop while virus scan is active and actually edit images in photoshop, all while doing other things as well).
All of these tests are application based, they are simply scripted or isolate the I/O component specifically. They give us a look into bursts of activity that's, again, near impossible to reproduce by hand with a stopwatch.
Benchmarking a specific task usually just repeats some information we've already presented, fails to present the bigger picture or shows no repeatable difference between drives. I can absolutely add those types of benchmarks back in, however I originally pulled them out because I believed they didn't add anything appreciable to the reviews.
Of course this is your site, if you guys would like me to present some of that data I definitely can :)
The problem is that the synthetic tests you do are hard to interpret for just about anyone. "What drive is the best for this usage profile?" is still really hard to answer after reading your reviews (not that anyone else does a better job).
And even if there is little difference between todays drives in the level loading time tests you used to do, we don't know that even if you do. Right now the average AT reader reads this test and doesn't know that the more expensive drives won't load his games noticeably faster or perform better when doing video editing.
Maybe you should give recommendations for certain usage profiles, like video editing, photo editing, gaming etc. Even if you're just saying that there's not gonna be a noticeable difference.
It might help if you included statements like "you don't see any appreciable differences between drives. Most high end SSDs are fast enough to do most of these types of tasks just as quickly as one another." a bit more often in the articles. While we might be interested in the technical data, it would usually be foolish to buy SSDs by things other than size, price, and reliability.
But that's the thing. You don't see any difference in an application a "normal" homeuser would use. We see huge differences in those synthetic tests, but in reality, you don't have any faster loading time.
Of course you can test it like this and say: "You don't see much difference between these three SSDs in "real world" application tests. Get the cheapest SSD (or most robust, whatever)."
Or another position (I think the one you are currently in) is: "You don't see much difference between these three SSDs in "real world" application tests so let's stress them some more and base our verdict on those stresstests."
The thing is: a) You don't know how the SSDs would REALLY react if you stress them in reality like this. They are still synthetic tests and unless you can prove that there are scenarios where differences appear (without any influence of some kind of bench program) they don't tell us that much.
b) I think we have to begin to widen our horizon a little bit. Why exactly is it, that you don't see any beneftit using a, let's say, 50k IOPS drive and a 15k IOPS drive? Shouldn't you see some significant faster load times?
Im telling you this because of future SSDs. We get 30k IOPS, soon 60k IOPS, and in one year maybe over 100k IOPS. The score in your benches gets bigger and bigger...and bigger... And what exactly is it the user gets? NOTHING because everything else in his computer is limiting his SSD (which is already happening right now!)!
I agree that you have to test hardware in scenarios, where nothing else is limiting your subject. That's why you use a 4GHz i7 when testing GPUs. That's why you test CPUs game performance using a very low resolution. But I think it's really important that you also test scenarios a user experiences in reality. And that means in this case: "real world" benches. And yes, there will be nearly no difference there. But isn't this the thing I want to know? If I spend 600$ on a fking RevoDrive and nothing loads faster, I WANNA KNOW ABOUT IT!
Real World testing of SSD's should be done in a worst case scenaio on the lowest common denominator
They should be plug and play on XP machines without any tweaks on the slowest computer you use to amplify the differences between drives
I use a copy/paste test on ATOM CPU's to guage the Real World differences between Platter drives and SSD's
Using 200MB of data (900+ files in 80 or so folders), I simply time a copy/paste of that data on the ATOM computer
Using a faster computer WILL reduce the "Relative" speed gap between drives to the point where it becomes hard to tell which of two drives is actually the fastest
Using Windows 7 with its funky caching scheme will make ALL the drives appear to copy and paste at the same speed on the ATOM core and therefore cannot be used for this test
A 40GB Vertex 2 can copy and paste this data in 55 seconds (3.6MB/sec) A 5400RPM Western Digital Laptop drive does it in 54 seconds A 7200RPM Western Digital Desktop Drive takes 17 seconds
ALL testing was done under XP-SP2 without ANY tweaks! All tests were repeated for accuracy
Sandforce SSD's are HORRIBLE at handling data that is NOT compressible or that is already on the drive in compressed form
Any drive that requires Windows 7 or multiple tweaks just to give you "Synthetic" numbers that have no bearing in the Real World are worthless
Show us how they compare in a worst case scenario on the least common denominator for results we can use please
I'm tired of hearing how great Sandforce drives are when they can't even beat a 5400RPM laptop drive in a Real World test such as the one I've just described
I see several good reasons to use XP and None to using Windows 7
The number 1 OS is still XP and has the highest user base so why does OCZ think the public will spend an extra $200 for Windows 7 just to use their overhyped SSD's?
Why doesn't OCZ just build SSD's for the majority of people on XP instead of making their customers jusmp though all these hoops just to get synthetic speeds from their drives that have little to do with real world results?
Representing the true nature of "random" access on the desktop is very difficult to do. Most desktops don't exhibit truly random access, instead what you get is a small file write followed by a table update somewhere else in the LBA space (not sequential, but not random). String a lot of these operations together and you get small writes peppered all over specific areas of the drive. The way we simulate this is running a 100% random sweep of 4KB writes but constraining it to a space of about 8GB of LBAs. This is overkill as well for most users, however what the benchmark does do is give an indication of worst case small file, non-sequential write performance. I agree with you that we need more synthetic tests that are representative of exactly what desktop random write behavior happens to be, however I haven't been able to come across the right combination to deliver just that. Admittedly I've been off chasing phones and other SSD issues as of late (you'll read more about this in the coming weeks) so I haven't been actively looking for a better 4KB test.
Now OS X feeling snappier vs. SandForce I can completely agree with. I don't believe this is 100% attributable to the data you see here, Apple also has the ability to go in and tweak its firmwares specifically for its software. I believe the ultra quick response time you see from boot and resume comes from firmware optimizations specific to OS X. Although I am going to toss this Kingston drive in my MBP to see what things look like at that point.
would you provide a different SSD test result with macbook pro?
like me, many macbook unibody or new mac book pro users use mainly for browsers and office, iphoto, itunes. we like to have ssd to make our experience better and faster. i search many website and blogs, but there are no clear answer for this. even, apple keeping quite about "TRIM" support!
after, reading your article, i am still not sure which ssd is good for my macbook unibody. i got the an idea of garbage collection which was very helpful. but didn't know, how long ssd last if we use for general purpose?
i really appreciate if you provide descriptive guideline of ssd for OS X. please, also tell us, is it worth to waiting for intel 3rd gen.? i desperately need ssd for mac book unibody! i dont mind to pay premium as long as performance stay as it is! also, i can store movies and other data in external hard drive!
Sandforce controllers give you the "illusion" of speed by writing less data to flash than contollers without hardware compression
If I wanted to test the speed of a copy and paste involving 200MB of data in the flash cells of a sandforce based controller, how can I tell exactly how much data is in the flash cells?
I mean, would Windows tell me how much compressed data is represented in the flash cells (200MB), or would Windows tell me how much compressed data is in the cells (maybe only 150MB) ?
The only way I can see fairly comparing an SSD with hardware compression and one without is to be sure you are actually writing the same amount of data to the flash cells (in this case - 200MB)
If sandforce based SSD's will only tell you how much data is represented and not what is actually on the drive, then I think the best way would be to use data that cannot be compressed
The tests I described in another post here involved copying and pasting 200MB of data which took 55 seconds on an ATOM computer with a Vertex 2 200MB / 55 sec = 3.6MB/sec
But if the 200MB was only a representation and the actual amount of data was for example 165MB in flash, then the actual throughput of my Vertex was even worse than I thought (In this case - 165MB / 55sec = 3.0MB/sec)
I need to know exactly how much data is indeed in flash or I need to start using non-compressible data for my tests
There has to be an missing pieces in our performance test, something that these companies knows and we dont. ------------------------------------------
Like smoke & Mirrors?
Sandforce Controllers Compress the data to give you the impression of speed
check the speed without compression and then compare drives
"check the speed without compression and then compare drives "
That makes no sense!
It's like saying disable branch predictions on one processor because another doesn't have it. Or disable the turbo on one car when comparing to a competing car that is naturally aspirated.
I'm not an expert on SSDs and RAID. So excuse me if my question is a bit stupid.
I always understood that it´s impossible to use TRIM with a RAID configuration because the RAID controllers can't ´pass' the TRIM commands. So the SSD performance would drop overtime when using RAID. Now the SSD controller has "always-on garbage collection". Does this mean the performance in a RAID configuration doesn't drop?
Will controllers like this leave OS X with a performance advantage over Windows 7, precisely because OS X has no support for the TRIM command? It seems that Win7 will have some system overhead in running TRIM while any OS without TRIM support will not suffer this, as garbage collection will be left to the firmware in the drive.
I second this. While the IOPS for Bench are a nice measure, I don't know how to actually translate that into real world performance differences. I.e. are the differences in IOPS between say the new Kingston drive and Sandforce drives actually noticeable, and if yes, how noticeable?
I would like to see a review that shows how much faster a Sandforce drive is than my old Vertex 1 and X25-M G2.
Huge differences in synthetic tests, but a normal desktop system is just too slow to actually benefit from this! I'm also missing these kinds of tests here. I think your earlier tests were better - now you just go through synthetic tests and show us the results, this isn't the thing I expected 1 or 2 years ago from an SSD test from anandtech.com. I'd like to see more tests which actually measure performance a user really gets when doing everyday tasks.
Wouldn't it be better to have a drive-based sort function run to test how responsive the drive is? I would guess it would stress all of the corners of SSD performance. Write a bunch, shuffle in small portions when flipping values. Maybe it wouldn't test sequential erase. I still think it would show great info.
Just a quick recommendation. I have one of these, and i can confirm that with windows 7 (ie TRIM) and a 6gbps bridge the performance is pretty delicious. Grab one, if you meet these requirements.
Well, i've seen speeds spike above 375MB/s tho ofc this could well be erroneous reporting on windows' side. I haven't actually hooked the drive up to my 3gbps ports so in all honesty, i can't compare the two - perhaps i should run a couple of benches...
It seems that you recommend drives despite the results of your own storage bench. It shows that the Kingston is the premier ssd to have if you want a drive that handles multi-tasking well.
Sandforce is nice if you do light tasks, but who the hell buys an ssd that only does well handling light tasks? No one!
"Sandforce is nice if you do light tasks, but who the hell buys an ssd that only does well handling light tasks? No one!"
Er... I do. Well obviously I would want a drive that does well handling heavy task loads as well but there are limits to how much I can pay and the cost per gig of some of the better performers is significantly higher. Maybe money is no object for you but if I'm *absolutely honest* with myself, I only *very rarely* perform the type of very heavy loads that Anand uses in his heavy load bench (it has almost ridiculously levels of multi-tasking). So the premium for something that benefits me only 2-3% of the time is unjustified.
That's why I renamed our light benchmark a "typical" benchmark, because it's not really a light usage case but rather more of what you'd commonly do on a system. The Kingston drive does very well there and in a few other tests, which is why I'd recommend it - however concerns about price and write amplification keep it from being a knock out of the park.
I have what may be an unusual access pattern -- seeks within a file -- that I haven't seen any "standard" benchmarks for, and I'm curious how drives do under it, particularly the Sandforce drives that depend upon (inherently sequential?) compression. Quite possibly, heavy database use has the same problem, but I haven't seen benchmarks on that, either.
I do meteorology and other environmental modeling, and frequently we want to "strip mine" the data in various selective ways. A typical data file might look like:
* Header stuff -- file description, etc.
* Sequence of time steps, each of which is an > array of variables, each of which is a + 2-D or 3-D grid of values
For example, you might have a year's worth of hourly meteorology (about 9000 time steps), for ten variables (of which temperature is the 2'nd), on a 250-row by 500-column grid.
So for this file, that's 0.5 MB per variable, 5 MB per time step, total size 45 GB, with one file per year.
Now you might want to know, "What's the temperature for Christmas Eve?" The logical sequence of operations to be performed is:
Now with a "conventional" disk, that's two seeks and two reads (assuming the header is not already cached by either the OS or the application), returning a result almost instantaneously. But what does that mean for a Sandforce-style drive that relies on compression, and implicitly on reading the whole thing in sequence? Does it mean I need to issue the data request and then go take a coffee break? I remember too well when this sort of data was stored in sequential ASCII files, and such a request would mean "Go take a 3-martini lunch." ;-(
I've been asking for similar for a while. What I want to know from a test is how as SSD behaves as a data drive for a real database, DB2/Oracle/PostgreSQL with 10's of gig of data doing realistic random transactions. The compression used by SandForce becomes germane, in that engine writers are incorporating compression/security in storage. Whether one should use consumer/prosumer drives for real databases is not pertinent; people do.
Yes I have been wondering about exactly this sort of thing too. I propose a seeking and logging benchmark. It should go something like this:
Create a set of 100 log files. Some only a few bytes. Some with a few MB of random data.
Create one very large file for seek testing. Just make an uncompressed zip file filled with 1/3 videos and 1/3 temporary internet files and 1/3 documents.
The actual test should be two steps:
1 - Open one log file and write a few bytes onto the end of it. Then close the file.
2 - Open the seek test file and seek to random location and read a few bytes. Close the file.
Then I guess you just count the number of loops this can run in a minute. Maybe run two threads, each working on 50 files.
Intel charging too much, surely you must be joking!
Do you know what the Dow Jones Industrial Average would be trading at if every DOW component (such as Intel) were to cut their margins down to the level of companies like Kingston? My guess would be about 3000. Something to keep in mind as we witness Bernanke's helicopter induced meltup...
This is my perspective from what I see throughout the review:
The Corsair Force drives and the Crucial RealSSD drives are very close performance-wise in most of the comparisons; however, when price is included in the scenario the Corsair Force drives (60GB/120GB) win hands down.
The only true advantage the C300 has over the Force drives from what I can tell is in scenarios where the data is mostly uncompressible - I don't think this comes close to making up for the price disadvantage of the C300 drives since most likely users will have a large mechanical drive for their storage drive rather than opting to use their C300 for that purpose.
Taking into account the fact that you can get the Corsair Force 60GB SSD for $5 more than the 40GB at NewEgg, I would say that pretty much nullifies the cost/GB advantage of the C300 at the lower capacity - at that point you're looking at $2.109/GB for the C300 and $2.031/GB for the Force 60GB assuming it has 64GB NAND capacity (that's going by dividing the Force 120GB specs in half...if we instead use the specs from the Force 40GB that would put the Force 60GB at an amazing $1.805/GB for a low capacity SSD!).
If I were to buy an SSD tomorrow it would almost definitely be the Corsair Force 60GB for its fantastic (at least by SSD standards) cost/capacity at a low price point.
Why are you assuming the Corsair drive magically has more space than it actually lists? If it's a 60GB drive, then you only get 60GB of usable space with some other amount set aside as a scratch area.
Personally, I base my purchases more on random read than anything. Based on the prices, I'd most likely pick up the Crucial C300 64GB if I was looking for a new boot drive.
Just a simple request for a G2 80gig Intel drive for future reviews. I think many of us have one of these and know the 160gig drive is significantly faster in some tests (sequential write), but almost identical in others (random read I believe).
i own an intel 40gb and intel's trim software reports it... it's part of a drive's SMART data, so any SMART reading program should be able to tell you, but i could be mistaken
"Note that not all SandForce drives are created equal here. If a manufacturer doesn’t meet SandForce’s sales requirements, their drives are capped at a maximum of 50MB/s here. This is the case with the Patriot Inferno, although OCZ’s Agility 2 voluntarily enforces the limit."
I understand there is the 'higher performance' Sandforce firmware (Vertex 2 vs Agility 2 for example) and the Agility 2-type firmware is the one that gets 'capped,' but what exactly do you mean by 'meet Sandforce's sales requirements'? Is this to say that the firmware the retail companies are licensed to use and able to ship is restricted based upon sales figres/order quantity, so that if say OCZ buys 100k SF controllers while Patriot only buys 10k, Patriot is limited in which firmware they are allowed to offer? If it's not that, what exactly do you mean? And most importantly how can we as end-users making a purchase determine this difference, is looking at IOPS all that's needed?
I didn't like the wording of the power consumption in terms of not being better at load than a mechanical hard drive. Just like your praise of Intel's hurry up and get idle (with turbo), all of these SSD's have an advantage in bursting and then getting idle.
I had mentioned it in a previous SSD article but what we really need is a moderated workload suite. Something like virus scan and total power draw during that time. While load voltages might be the same or higher for the SSD's, the total power consumption during that time will (should) be less. I think a virus scan is perfect on a normal system as it should take 10-20min and give a measureable difference between drives. Grab a laptop from one of your recent reviews (not an Atom or other super-slow mobile to prevent the SSD from waiting around), and measure the total power drawn over the scan.
This is less important in a desktop obviously but for the mobile crowd (whom sees the biggest benefit over the traditional laptop drives anyway) might just be one more criteria to help make a decision (or just show that it's the type of drive (mech/SSD) and not really differences between the SSD's themselves).
Glad to have a new SSD review though. I love reading these and am trying to justify upgrading from my 80gig Intel G2 drive that has been fantastic for the last year. The great thing about the SSD tech is that you can hand-me-down smaller capacity, slower drives and completely revolutionize an older system.
"Kingston ships the V+100 with a 3-year warranty and to Kingston's credit I haven't had any other drives die as a result of wearing out the NAND. Even if the V+100 has higher effective write amplification than the competition, your usage model will determine whether or not you bump into it."
Anand,
Firstly, thank you for being the industry standard when it comes to articles about SSD. It's a favorite topic of mine and your articles have steered a lot of purchases in my personal and professional circles (and you haven't steered us wrong!)
My request...
One of the lingering questions about SSDs is how long they'll actually last in the real-world before the NAND is fried. When I see you mentioning that you've never actually fried any NAND, that seems like an opportunity for you to give it a shot and give the public some answers.
I'd love to see a "torture test" article where you run simulated workloads against SSDs 24/7 until they can't take any more writes, and report on the results.
Yes, a longevity test. Put it on grind mode until it pops. P51 Mustangs benefited from engines tested that way.
Now, this one, should it be fully synthetic or more life like? Just place the drive in "write 0, write 1" until failure, record how many times it can be used, or create some random workload scripted in such a manner that it behaves pretty much like real usage, overusing the first few bytes of every 4k sector... if it affects any results. What am I asking is, will it wear only the used bytes, or the entire 4k rewritten sector will be worn evenly, if I am expressing myself correctly here.
On another comment, I always thought SSD drives were like overpriced undersized Raptors, since they came to be, but damn... I hope fierce competition drive the prices down. Way down. "Neck and neck to mag drives" down.
And what about defragmenting utilities? Don't they lose their sense of purpose on a SSD? Are they blocked from usage, since the best situation you have on a SSD is actually randomly sprayed data, because there is no "needle" running over it at 7200rpm in a forcibly better sequential read? Should they be renamed to "optimization tools" when concerning SSD's? Should anybody ever consider manually giving permission to a system to run garbage collection, TRIM, whatever, while blocking it until strictly necessary, in order to increase life span?
win7 "automatically disables" their built-in defrag for SSDs, though if you go in and manually tell it to defrag an SSD, it will do it without question
prefetch and superfetch are supposedly also disabled automatically when the OS is installed on an SSD, though I don't feel comfortable until i change the values myself in the registry
I have torture-tested a 50 GB Sandforce-based drive (OWC Mercury Extreme Pro RE) with the goal to destroy it. I stopped writing our semi-random data after 21 days because I grew tired. 16.5 TB/d, 360 disk fills/day, 21 days more or less 24/7 duty cycle (we stopped a few times for an hour or two to make adjustments) ~7500 disk fills total, 350 TB written The drive still performs as good as new, and SMART parameters look reasonably good - to the extent that current tools can interpret them anyway.
If I normally write 20 GB/d this drive is going to outlast me. Actually, I expect it to die from "normal" (for electronics) age-related causes, not flash cells becoming unwritable.
This is something I've been working on for the past few months. Physically wearing down a drive as quickly as possible is one way to go about it (all of the manufacturers do this) but it's basically impossible to do for real world workloads (like the AT Storage Bench). It would take months on the worst drives, and years on the best ones.
There is another way however. Remember NAND should fail predictably, we just need to fill in some of the variables of the equation...
I'm still a month or two away from publishing but if you're buying for longevity, SandForce seems to last the longest, followed by Crucial and then Intel. There's a *sharp* fall off after Intel however. The Indilinx and JMicron stuff, depending on workload, could fail within 3 - 5 years. Again it's entirely dependent on workload, if all you're doing is browsing the web even a JMF618 drive can last a decade. If you're running a workload full of 4KB random writes across the entire drive constantly, the 618 will be dead in less than 2 years.
Wow, I would have expected Intel to last the longest. I am going to purchase an ssd and longevity is one of my main concerns. In fact, longevity is the main reason I have not yet bought a Sandforce drive. Well, I guess that is what happens when you make assumptions. Looking forward to the article!
A torture test like that is going to sell a LOT of SSDs, Anand. Because right now that's the only thing keeping businesses and a lot of "power users" from adopting them - "but won't they wear out soon?"
That was the exact question I got when trying to get my boss to buy me one. Though I was eventually able to convince him. :)
Synthetic Benchmarks won't show you how fast the various SSD controllers handle uncompressible data
Only a copy and paste of several hundred megabytes to and from the same drive under XP will show you what SSD's will do under actual load
First off, due to Windows 7's caching scheme, ALL drives (Slow or Fast) seem to finish a copy and paste in the same amount of time and cannot be used for this test
In a worst case scenario, using an ATOM computer with Windows XP and Zero SSD Tweaks, a OCZ VERTEX 2 will copy and paste data at only 3.6 Megabytes per second
A 5400RPM laptop drive was faster than the Vertex 2 in this test because OCZ drives require massive amounts of Tweaking and highly compressible data to get the numbers they are advertizing
A 7200RPM desktop drive was A LOT faster than the Vertex 2 in this type of test
Anyone working with uncompressible data "already on the drive" such as video editors should avoid Sandforce SSD's and stick with the much faster desktop platter drives
Using a slower ATOM computer for these tests will amplify the difference between slower and faster drives and give you a better idea of the "Relative" speed difference between drives
You should use this test for ALL SSD's and compare the results to common hard drives so that end users can get a feel for the "Actual" throughput of these drives on uncompressible data
Remember, Data on the Vertex drive's is already compressed and cannot be compressed again during a copy/paste to show you the actual throughput of the drive under XP
Worst case scenario testing under XP is the way to go with SSD's to see what they will really do under actual workloads without endless tweaking and without getting bogus results due to Windows 7's caching scheme
The issue with Windows XP vs. Windows 7 doesn't have anything to do with actual load, it has to do with alignment.
Controllers designed with modern OSes in mind (Windows 7, OS X 10.5/10.6) in mind (C300, SandForce) are optimized for 4K aligned transfers. By default, Windows XP isn't 4K aligned and thus performance suffers. See here:
If you want the best out of box XP performance for incompressible data, Intel's X25-M G2 is likely the best candidate. The G1/G2 controllers are alignment agnostic and will always deliver the same performance regardless of partition alignment. Intel's controller was designed to go after large corporate sales and, at the time it was designed, many of those companies were still on XP.
With so many XP machines out there for the foreseeable future, I would think more SSD manufacturers would target the XP market with alignment agnostic controllers instead of making the consumers jump through all these hoops to get reasonable XP performance from their SSD's
Last question..
Would OS agnostic garbage collection like that on the new Kingston SSD work with Sandforce controllers if the manufacturers chose to include it in firmware or is it irrelevant with Duraclass ?
I still think SSD's should be plug and play on ALL operating Systems
Personally, I'd rather just use the drives instead of spending all this time tweaking them
This seems like a worrying trend, though time will tell how reliable SSDs are long-term. What's the situation with 2Xnm? And where does SLC fit into all that regarding reliability, performance, pricing, market usage trends?
The 2xnm lifetimes will be shorter. The flash lifetime is a function of feature size, so as processes shrink lifetime will as well. At the same time, it's only likely to be a concern if you're doing very IO intensive activities. My 128GB Indilinx drive is at 96% after a year in my main desktop. At this rate I'll be nowhere near the maximum write limit when the flash dies of old age in another 4-9 years even if I keep it as my main drive the whole time which is doubtful.
If scaling is linear with feature area, as the 50% drop in life-cycles from 50 to 34 nm implies, 22nm 2 level flash will last about 2500 write cycles. This still will probably be long enough not to matter. 14nm flash and beyond might not be unless they switch to SLC for consumer SSDs, and only use MLC for memory cards and thumb drives. With 10x the allowed number of write cycles SLC flash should remain good for several additional process shrinks.
OTOH that might be a moot point because the power levels needed to write flash doesn't drop with each process, and unless something changes flash is expected to hit a will in the teens because the wires on the chip won't be able to carry the load without melting.
Anand, couldn't you just continue running your random write test 24/7 until it falls dead, or is this not feasible?
Maybe it's not feasible with the Kingston drive because the random write performance is too slow? Which could bring up speculation that it's intentionally handicapped so this type of test is avoided?
Really interesting article as usual. I had no idea that there was such a gulf between various Sandforce controllers. There was one line however that perplexed me, "The V+ 100‘s sequential read speed is excellent, just a hair above the top drives from Intel and Crucial. There’s not much room for improvement here unless you go to a 6Gbps interface."
Why didn't you use a Sata 6 controller? I think that most of us buying boards today will be looking at Sata 6 since so many today's current SSDs clearly outstrip Sata 3, to say nothing of the upcoming 500MB/s and better drives coming down the pipeline. I'm sure you have a good reason, and I know there has been some weird controller issues with AMD I believe, but I'm very curious! Thank you.
Puts it at around $1.60/GB and if you skip past the synthetic benchmarks and look at AT Storage Bench, its a pretty formidable drive. Makes me a little sad I spent more than this on my 160 X25M G2 a while back.
Brilliant...just needs controller info added. I think I'll do some searching later tonight if I have time.
Is total capacity listed in GB (10^9) or GiB(2^30)? What about usable capacity? I wish people paid more attention to this, especially in area such as this were flash storage (typically in GiB) is being used for a drive (typically in GB).
Anand, Do you think a short description of which controller each drive uses could be added to the table on page 2? It seems to me that it would be much more instructive to sort the drives based on this, rather than the often rebadged brand.
There's a lot of stats for the drives, however I'm not any closer in choosing a drive.
I have 3 systems: 1. Windows 7. This system is mostly lots of opened web browsers hogging memory, email, maybe a Word document, video encoding, and reference software (i.e. encyclopedia). I assume the browser is caching to the drive, same with the virtual memory and other background apps. I don't turn the machine off, so boot time is not important. For video encoding, I'd like to keep the data on a HDD since it is a lot of storage that SDD's are not affordable at storing, and it is most likely CPU bound. 2. Media Center. Want a SDD for the main drive (OS and apps), but will use a larger/slower drive for the media storage. Need something low heat. This is what I really want a SDD for. 3. Linux. Mostly emailing and web browsing. All the apps I use are cloud apps.
Obviously the fastest drive would be the one I want if money is no object. But I want to know how much faster my experience is going to be for how much more money? My goal is a snappy system and less time waiting around. I want to know what I'd sacrifice by getting a "value" drive versus the higher quality.
Also, MTBF is important. I hate losing data! Hope the drives have SMART or some other methods of warnings.
If all you want to do is boot up quickly and load your apps really fast, then a sandforce drive is perfect for you
If you want to manipulate data already on the drive or copy & paste to the same drive, then you are much better off with a 7200RPM desktop drive as far as I can tell
Anyone wanting to spend their time doing actual work instead of constantly tweaking their SSD to get synthetic benchmarks that have little bearing on real world results should avoid sandforce based drives like the vertex 2
Please read my other posts in this thread for more info
I've ready your other posts and you are selecting a very specific situation which most people will not encounter (that is copy/paste using XP with no SSD tweaks). That's like putting in an NVIDIA video card and not installing the drivers and complaining about performance. The XP OS was designed so far before SSD's were even available to enterprise markets let alone the consumer space its no wonder they don't perform well.
The vast majority of people are no longer on XP (trust me I was one of the last to hold out due to my dispise for Vista). But you are artificially setting limitations on a product and then using it to generalize performance in all situations that are to you "real world".
I have real concerns about the Sandforce drives due to variable performance. I have an Intel 80gig G2 SSD and am still amazed at how well it does (to me) in the most important benchmarks, that is random read/write and sequential read. Even with the cheap(er) pricing of SSD's and the *hopefully* significant decrease once we get to the next node with Intel's 3rd generation these are still not capable of storing the majority of our data (unless you think 1-2 grand is chump change). That being said these are mainly going to be access drives and rarely used for moving large amounts of data around (which is why the sequential write becomes much less important for day to day use, besides the odd large install such as a video game).
I keep my 80gig with about 45GB free. Win7 install (including User data), a couple programs, the current 1-3 games I'm playing and that's it. I have a 250GB secondary 7200rpm drive for music/movies/etc. and a larger external HDD for backup and infrequently used data.
Quote: these are mainly going to be access drives and rarely used for moving large amounts of data around (which is why the sequential write becomes much less important for day to day use, besides the odd large install such as a video game). ---------------------------------------------------------------------------------------
In that case, I'd grab a Crucial Real SSD Best read speeds now on Sata 2 and reasonable for Sata 3
64GB is very reasonably priced for the read speeds you will get
Quote: The vast majority of people are no longer on XP (trust me I was one of the last to hold out due to my dispise for Vista). But you are artificially setting limitations on a product and then using it to generalize performance in all situations that are to you "real world". -------------------------------------------------
Where did you get that incorrect data? It will be a few years before Windows 7 sells more copies than XP
XP is still used on more computers than Vista and Windows 7 combined!
Any SSD manufacturer who misleads their customers into thinking that their SSD's are plug and play compatible with XP by omission of pertinent information should go into spam marketing!
Why should the consumer spend another $200 for a new OS to use a Vertex when Intel SSD's work just fine with XP without all the tweaks?
I can't stand all the SSD articles with no real world benchies at all. I don't care about transfer speed or even IOPS. I care about how fast it boots, how fast it loads my work, displays my porn, plays my games, etc.
I only see theoretical performances of the drives, while this is some indication of how it would perform in real case scenarios, its not the most accurate and its hard to translate into real world.
I would like some tests that test all of the SDD's abilities at once. For example running a virus scan while pasting 2GB or big and small data, while moving hundreds of pictures into a image viewer, all on top while downloading a torrent and surfing the web.
Objective performance measurements are good to have, however these comments are suggesting some difficulties in interpretation because the majority of us don't know how to relate to a synthetic benchmark. My recommendation is to create some user profiles (gamer, video artists, graphics artist, serious web surfer, media PC, low power, etc). Have some people utilize the machines and ask for their feedback.
Because here's my bottom line: I don't want to pay for more than I'll notice. If PEOPLE can't tell the difference between a top synthetic rated drive versus middle of the pack....then I don't want to pay extra.
For example: If a more expensive SSD drive will not speed up my Media Center PC, I don't want to pay the extra. I want to know what a real person thinks and experiences. Rather talk to my buddies than get comments from a robot performing synthetic tests.
Dear Anand, I would love a reply here, since I've raised these points in the past, and been ignored. "Concerned enough to recommend running it with 20% free space at all times (at least). The more free space you have, the better job the controller can do wear leveling." I feel the need to critique this statement. If TRIM is not active, there is no difference whatsoever if you use 10% or 100% of the LBAs avalible to the user, since the drive can't tell the difference. The only way you could make sure 20% of the NAND is free at all times is to not partition it in the first place (or not include it in a partition after a secure erase).
Also, you've included 4KB random write at QD 32, but still not 4KB random read. If you compare your results, you will see a much larger difference between read performance as QD scales than you do for write. This is due to write coalessing and attenuation, while reads don't have this benefit. I feel both random read and write @ QD 32 should be included to show what the drives are capable of within the NCQ spec. Ideally would be a graph showing IOPS by QD scaling, but i understand that it may be a bit much work. When benchmarking my own SSDs, i scale 4 dimentions in IOmeter; Read:write ratio, seq:ran ratio, block size, and QD. Due to size of the data set, i only scale them against each other at 4KB and a couple of larger sizes. A few thousand data points is enough pr drive :P
Quote: “SandForce's partners who have to pay a big chunk of their margins to SandForce as well as the NAND vendor are actually delivering the best value in SSDs. Kingston and Western Digital also deliver a great value. Not Crucial/Micron and not Intel, which is not only disappointing but inexcusable. These companies actually own the fabs where the NAND is made and in the case of Intel, they actually produce the controller itself”.
Are you forgetting that Intel spent over 2 billion on the JV with Micron? They are the ones carrying the investment risk, not the likes of OCZ.
Also Intel spend money and time on getting their products right before market and use much higher quality components throughout, not least better QA’d NAND. (Intel do not sell SSD grade NAND. They sell NAND that has to be further processed by whoever buys it).
You have consistently ignored these facts and they make a big difference.
Whilst I’m at it I can’t understand your assertion that SF drives are so good. There are countless problems with them even now after numerous f/w updates. They degrade badly with uncompressible data, they have weak GC and they don’t trim in the same way as other drives. It’s quite easy to see performance drop by 50%, even OCZ have admitted that. Read performance, without writing also degrades.
I used to look forward to your reviews but it seems your objectivity when it comes to the X25 and C300 in favour of SF drives is seriously lacking.
"Here is the thing, if you hammer the driver with say enough writes that the drive would under normal use/see in 7 days within a few hrs, the drive will slow down for 7 days, maybe longer. It does this to protect the nand life. So your guys seeing a 50% drop may actually see 30% which is the normal drop, then a further 20% because at some stage they have hammered the drive and then not realised its going to take 5 days or longer for the speed to creep back up. Also remember this write quantity slowdown is further impacted by how you use the drive after you have hammered it."
I have built probably 10 pc's in the last 2 yrs, and wonder if adding a SSD as a boot drive would be worth the effort.
I see you guys making lots of comments, but as a newbie on SSD's, I would really like to see a thread on how to introduce a SSD to a system that has a bunch of SATA drives.
Exactly how does one go about installing Windows XP Pro? I imagine that I should disconnect my SATA drives and just hook up the SSD and then do a CD install?
Can programs like CasperXP ( mirror imaging program ), be used to do a mirror image backup from an SSD to a Sata drive ( C: boot drive SSD to D: mirror drive SATA ).
My apologies if this was posted in the wrong thread.
QUOTE: Exactly how does one go about installing Windows XP Pro? I imagine that I should disconnect my SATA drives and just hook up the SSD and then do a CD install? ---------------------------------------------------------------------------------------------------------
You are correct as far as I can tell... I just spent the past 2 days retesting the copy/paste speed of a Vertex 2 with the alignment tweak recommended by OCZ
There was no speed benefit resulting from the alignment of the partition as OCZ described
Copy and paste speed was still a miserable 3.6 megabutes per second
But just to be on the safe side, I will start using ONLY Intel SSD's with Windows XP
Anand says Intel controllers are alignment agnostic and should be the best for non-compressible data under XP
So, untill other SSD manufacturers start making alignment agnostic controllers that are capeable of doing a copy and paste faster than my 5400RPM laptop drive under XP, I think I'll just stick with Intel
you noted that due to the aggressive garbage-collection this SSD will have a large write-amplification, so the flash will wear-out more quickly. What would happen when using this SSD with an OS that supports TRIM? Will TRIM support the internal GC and lead to a smaller write-amplification and thus a longer lifetime of the SSD?
Can you test how VMs behave on SSD? I understand it's partially there in synthetic IOPS and bandwith tests - that show what can you get max. from these drives. But "normal" load from multiple VMs should be different than normal load from a single OS, so drives that do good at normal load in current test migth be not performant under normal load of multiple VMs.
This test with few VMs makes sense for larger drives (160, 256GB), as there just won't be enough place for multiple OSes on small drives. Also longevity becomes more important when you have few times larger load.
I use Linux and its TRIM command support is sparse. Finally a drive that does not require the TRIM command. I know it comes at a cost of more wear and tear, but I do not mind because pros out weighs the cons. It will give me no noise from the storage system and it be about 60 times quicker than my hard drive to access files. I do not think $280 for 128 GB is not a bad deal for a SSD that does not require the TRIM command to function great. Though, one thing that comes to mind is does the Kingston SSDNow V+100 come with ECC because all hard drives includes ECC for data integrity.
True, but the file system also have problems with SSD. Not all file system in Linux are SSD aware. EXT3 and EXT4 does not have good support to handle SSD.
Linux filesystems ext3/ext4 are designed for magnetic drives - but that is also true for ntfs and hfs+, so the problems with SSD are not unique to Linux. However, it is true that while at least Windows7 does some automatic optimizations, you have to configure everything manually under Linux.
All you have to do is format the drive with ext4, add the discard option, and you are set. TRIM is enabled automatically in the background; no need to manually run the command anymore.
Even with the latest firmwares, this isn't an issue anyway.
Let me start of by saying that i'm new to SSD tech (noob) and i'm starting to learning but there are some things which where not pointed in the reviews as far as i remember.
Heck, even in this review you said that Inferno is somewhat capped to 50 mb/s if i'm not wrong?
So in short,
Where i live i'm limited to 2 models, one is Patriot Inferno and the Other A-DATA S599 with possibly 50K IOPS firmware, as Newegg and Amazon specs confirm?
Now, does that firmware insure i won't get slowdowns (at some points) like the Inferno model in this review? or there is some more about it. if there is, i really got no idea how is that possible same controller, same build? and heck in most of the test the Vertex 2 is still fastest, even faster then the other drives which now share the 50K IOPS firmware, G.Skill Phoenix pro, Corsair Force etc??
i'm mostly interested in SSD which will have good read/write speeds (4K etc) and to be fast during programs/games installation. I got 6 pcs connected in network so that means alot to me, in short does my needs require higher IOPS firmware or i would not notice any difference in performance during installation of big size programs/games.
Also, is it possible to know what revision you get without actually buying the product? is there any info on the package/sticker or part number? the difference in performance i noticed in some reviews was 2x higher random write (4K) speed on the models with OCZ "exclusive firmware. And as far that point goes i'm totally clueless does it really matter at all, and if its not how that that translate in real world usage?
Please answer i don't want to regret my purchase :(
Instead of $259, the Kingston V100+ 96GB is available for $119.99 after rebate from Buy.com. This pricing completely changes the value rating from acceptable to outstanding value! The mail-in rebate is rotating from one online store to another. Last month it was at Newegg.com; right now it's at Buy.com. Next month - who knows? Anyway, the cost per Gigabyte when tested here at Anandtech.com was around $2.70. Now it's $1.25 per Gigabyte, and a 96GB drive is large enough for Windows plus quite a few installed programs and games. Pretty good value. Think I just talked myself into ordering one .......
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
96 Comments
Back to Article
Dorin Nicolaescu-Musteață - Thursday, November 11, 2010 - link
Anand, what about the Samsung 470 Series?It's been out since August and looks like a very nice drive. Why in the world the reviews have only started to appear this week on-line.
Nickel020 - Thursday, November 11, 2010 - link
Thanks for the review! I've got a few suggestionss/questions though.I've been out of the loop for a while and looking at Sandforce and other newer drives. So some Sandforce drives have the I/O limitations that were intended for the SF-1200 and some have SF-1500 like performance?
I'm surprised the Corsair F40 does so well. I thought the lower capacity drives performed worse than the 120GB versions, but it holds up really well. Or is this just a special case with the 40GB one and the 60GB is worse than the 40GB one? The 60GB Sandforce are also much better value than the 40GB ones, 50% more capacity at >20% more price. I find it strange you didn't include them and mentioned the 64GB C300 to be the value drive at their price point.
I'm pretty sure the Indilinx 60GB, the unlocked SF-1200 60GB and the X25-M 80GB are the most popular drives out there, which makes them great reference points, but they're not in the review. The former two are not on Bench either unfortunately. Do you have some around and could test them?
You tested the Crucial drives on the ICH10R, right?
Also, I would appreciate some blog posts or small articles about developments with newer FWs. I remember the FW development improving the Indilinx drives significantly, and am always wondering how accurate your older reviews still are given there are newer FWs out now. It would also be nice if you could list the tested FW version in Bench.
It would also be great if you could look at SSD performance in Macbooks. I want to put one in my Macbook Pro (Late 2008), but all the talk of freezing has me hesitating, and I haven't seen an in-depth look at this issue. Is it related to what kind of SSD you use, and does it make a difference wheter you have a late 2008 or mid 2009? It would also be interesting to see how tha lack of TRIM actually affects different drives under OSX.
That's all for now, thanks again!
retnuh - Thursday, November 11, 2010 - link
I've had a OWC Mercury Extreme Pro 240GB in my late 2008 MBP since may, not one issue or freezing. Best upgrade you can do.http://eshop.macsales.com/shop/internal_storage/Me...
iwod - Thursday, November 11, 2010 - link
I posted and asked on many forums and did not found an answer to my MBA ( Macbook Air ) question.Why did the MBA do so well in its test, while its performance data were below the King of SSD controller, Sandforce?
No one could answer. There were number of review pointing out that their MBA actually feels snappier then their Macbook with Sandforce or Intel SSD. Although this is impossible when first heard, numerous other review site seems to confirm similar findings. Of cause there is no way to test it out since the MBA does not have an regular SATA slot.
Now this article actually print out the truth. The same Toshiba SSD controller used in MBA's SSD, is top of the chart in BOTH Synthetic Benchmarks and Real World usage ( Anand Bench ) Benchmarks. What we have been talking as the Holy Grail of SSD Performance Delta, the 4K Random Read / Write, didn't matter when Toshiba was literally the bottom of the chart in those test.
There is a reason why Apple choose a Inferior part ( To us at the time ) instead of Sandforce. The argument for choosing it because of always On GC doesn't make sense, since Sandforce has the same capabilities within the Firmware itself.
One reason would be Toshiba is a NAND manufacture itself, and buying NAND and Controller directly from Toshiba would be cheaper. The other reason being Toshiba ( properly involve Sandisk as well since their JV ) had a controller chip which is very fast.
There has to be an missing pieces in our performance test, something that these companies knows and we dont.
Chloiber - Thursday, November 11, 2010 - link
I'd like to see more real world tests - and I don't consider the AnandTech Storage Bench to be "real world" - it's still a bench, like PCMark.But yes, you are right: synthetic tests tell us little about the performance you actually get from an SSD. There are more unknown variables than we think.
You may see big differences in benches like SysMark or PCMark - and even bigger differences in synthetic tests like AS SSD or even IOMeter. But these scores tell us little about REAL world performance - and with REAL I mean things like:
- "How long does it take to start Photoshop while running Virus Scan?"
- "How long does it take to start iTunes while unzipping a not-so-much-compressed zip-file?"
That's the things I care about. And interestingly, you often get COMPLETELY different results, than what you would think when looking at synthetic tests or "half-synthetic"-tests like PCMark or AnandTech Storage Bench.
Anand Lal Shimpi - Thursday, November 11, 2010 - link
I used to run a lot of those types of tests, however I quickly found that if you choose your iometer and other benchmarks appropriately they don't any new data. And often times they are so limited in their scope (e.g. launch an application with virus scan in the background) that you don't see any appreciable differences between drives. Most high end SSDs are fast enough to do most of these types of tasks just as quickly as one another. It's when you string a bunch of operations together and look for cumulative differences in response time or performance that you can really begin to see which one is faster. These types of scenarios are virtually impossible to perform with consistency by hand, that's where our test suite comes in.AnandTech Storage Bench, PCMark and even SYSMark do what is necessary - they measure performance of a more complex usage case. PCMark Vantage is a great showcase of truly light workload I/O performance, while SYSMark is more CPU bound and shows you how small the differences can be. Our own benchmark offers a more modern set of usage models (we actually do run photoshop while virus scan is active and actually edit images in photoshop, all while doing other things as well).
All of these tests are application based, they are simply scripted or isolate the I/O component specifically. They give us a look into bursts of activity that's, again, near impossible to reproduce by hand with a stopwatch.
Benchmarking a specific task usually just repeats some information we've already presented, fails to present the bigger picture or shows no repeatable difference between drives. I can absolutely add those types of benchmarks back in, however I originally pulled them out because I believed they didn't add anything appreciable to the reviews.
Of course this is your site, if you guys would like me to present some of that data I definitely can :)
Take care,
Anand
Nickel020 - Thursday, November 11, 2010 - link
The problem is that the synthetic tests you do are hard to interpret for just about anyone. "What drive is the best for this usage profile?" is still really hard to answer after reading your reviews (not that anyone else does a better job).And even if there is little difference between todays drives in the level loading time tests you used to do, we don't know that even if you do. Right now the average AT reader reads this test and doesn't know that the more expensive drives won't load his games noticeably faster or perform better when doing video editing.
Maybe you should give recommendations for certain usage profiles, like video editing, photo editing, gaming etc. Even if you're just saying that there's not gonna be a noticeable difference.
wumpus - Thursday, November 11, 2010 - link
It might help if you included statements like "you don't see any appreciable differences between drives. Most high end SSDs are fast enough to do most of these types of tasks just as quickly as one another." a bit more often in the articles. While we might be interested in the technical data, it would usually be foolish to buy SSDs by things other than size, price, and reliability.Chloiber - Saturday, November 13, 2010 - link
But that's the thing. You don't see any difference in an application a "normal" homeuser would use. We see huge differences in those synthetic tests, but in reality, you don't have any faster loading time.Of course you can test it like this and say:
"You don't see much difference between these three SSDs in "real world" application tests. Get the cheapest SSD (or most robust, whatever)."
Or another position (I think the one you are currently in) is:
"You don't see much difference between these three SSDs in "real world" application tests so let's stress them some more and base our verdict on those stresstests."
The thing is:
a) You don't know how the SSDs would REALLY react if you stress them in reality like this. They are still synthetic tests and unless you can prove that there are scenarios where differences appear (without any influence of some kind of bench program) they don't tell us that much.
b) I think we have to begin to widen our horizon a little bit. Why exactly is it, that you don't see any beneftit using a, let's say, 50k IOPS drive and a 15k IOPS drive? Shouldn't you see some significant faster load times?
Im telling you this because of future SSDs. We get 30k IOPS, soon 60k IOPS, and in one year maybe over 100k IOPS. The score in your benches gets bigger and bigger...and bigger...
And what exactly is it the user gets? NOTHING because everything else in his computer is limiting his SSD (which is already happening right now!)!
I agree that you have to test hardware in scenarios, where nothing else is limiting your subject. That's why you use a 4GHz i7 when testing GPUs. That's why you test CPUs game performance using a very low resolution.
But I think it's really important that you also test scenarios a user experiences in reality. And that means in this case: "real world" benches. And yes, there will be nearly no difference there. But isn't this the thing I want to know? If I spend 600$ on a fking RevoDrive and nothing loads faster, I WANNA KNOW ABOUT IT!
I hope you see my point :)
Out of Box Experience - Thursday, November 11, 2010 - link
Real World testing of SSD's should be done in a worst case scenaio on the lowest common denominatorThey should be plug and play on XP machines without any tweaks on the slowest computer you use to amplify the differences between drives
I use a copy/paste test on ATOM CPU's to guage the Real World differences between Platter drives and SSD's
Using 200MB of data (900+ files in 80 or so folders), I simply time a copy/paste of that data on the ATOM computer
Using a faster computer WILL reduce the "Relative" speed gap between drives to the point where it becomes hard to tell which of two drives is actually the fastest
Using Windows 7 with its funky caching scheme will make ALL the drives appear to copy and paste at the same speed on the ATOM core and therefore cannot be used for this test
A 40GB Vertex 2 can copy and paste this data in 55 seconds (3.6MB/sec)
A 5400RPM Western Digital Laptop drive does it in 54 seconds
A 7200RPM Western Digital Desktop Drive takes 17 seconds
ALL testing was done under XP-SP2 without ANY tweaks!
All tests were repeated for accuracy
Sandforce SSD's are HORRIBLE at handling data that is NOT compressible or that is already on the drive in compressed form
Any drive that requires Windows 7 or multiple tweaks just to give you "Synthetic" numbers that have no bearing in the Real World are worthless
Show us how they compare in a worst case scenario on the least common denominator for results we can use please
I'm tired of hearing how great Sandforce drives are when they can't even beat a 5400RPM laptop drive in a Real World test such as the one I've just described
dagamer34 - Saturday, November 13, 2010 - link
If you're buying an SSD, I see no reason why your OS should still be Windows XP.Oxford Guy - Sunday, November 14, 2010 - link
Some may not want to pay the Microsoft tax.Out of Box Experience - Monday, November 15, 2010 - link
What you see is irrelevant to what I use!I see several good reasons to use XP and None to using Windows 7
The number 1 OS is still XP and has the highest user base so why does OCZ think the public will spend an extra $200 for Windows 7 just to use their overhyped SSD's?
Why doesn't OCZ just build SSD's for the majority of people on XP instead of making their customers jusmp though all these hoops just to get synthetic speeds from their drives that have little to do with real world results?
sprockkets - Sunday, November 21, 2010 - link
Oh, I don't know, TRIM support, built in alignment support, build in optimization after running WEI for SSDs?But if you want to stick with a 9 year old OS which lacks basic security, poor desktop video rendering, etc, go right on ahead.
Anand Lal Shimpi - Thursday, November 11, 2010 - link
Representing the true nature of "random" access on the desktop is very difficult to do. Most desktops don't exhibit truly random access, instead what you get is a small file write followed by a table update somewhere else in the LBA space (not sequential, but not random). String a lot of these operations together and you get small writes peppered all over specific areas of the drive. The way we simulate this is running a 100% random sweep of 4KB writes but constraining it to a space of about 8GB of LBAs. This is overkill as well for most users, however what the benchmark does do is give an indication of worst case small file, non-sequential write performance. I agree with you that we need more synthetic tests that are representative of exactly what desktop random write behavior happens to be, however I haven't been able to come across the right combination to deliver just that. Admittedly I've been off chasing phones and other SSD issues as of late (you'll read more about this in the coming weeks) so I haven't been actively looking for a better 4KB test.Now OS X feeling snappier vs. SandForce I can completely agree with. I don't believe this is 100% attributable to the data you see here, Apple also has the ability to go in and tweak its firmwares specifically for its software. I believe the ultra quick response time you see from boot and resume comes from firmware optimizations specific to OS X. Although I am going to toss this Kingston drive in my MBP to see what things look like at that point.
Take care,
Anand
iwodo - Friday, November 12, 2010 - link
ARH, Thx,"toss this Kingston drive in my MBP to see what things look like at that point."
I never thought of that. Keep thinking mSATA was blocking anyone from testing it.
I am looking forward to see your SSD issues and MBP testing.
Tech news has been dull for number of years, SSD finally make thing interesting again.
sunjava04 - Friday, November 12, 2010 - link
hey anand,would you provide a different SSD test result with macbook pro?
like me, many macbook unibody or new mac book pro users use mainly for browsers and office, iphoto, itunes. we like to have ssd to make our experience better and faster. i search many website and blogs, but there are no clear answer for this.
even, apple keeping quite about "TRIM" support!
after, reading your article, i am still not sure which ssd is good for my macbook unibody. i got the an idea of garbage collection which was very helpful. but didn't know, how long ssd last if we use for general purpose?
i really appreciate if you provide descriptive guideline of ssd for OS X.
please, also tell us, is it worth to waiting for intel 3rd gen.?
i desperately need ssd for mac book unibody!
i dont mind to pay premium as long as performance stay as it is! also, i can store movies and other data in external hard drive!
Sincerely,
Rishi
iwodo - Saturday, November 13, 2010 - link
As of Today, i just read another review on SSD comparison. Namely Intel SSD and Sandforce,While the Sandforce wins on all synthetic benchmarks like Seq Read Write and Random Read Write.
It was booting slower, starting up Apps slower, finish task slower then Intel SSD.
And by an noticeable amount of percentage. ( 10 - 30% )
I am beginning to think there are things Sandforce dont work well at all. But again, we have yet to find out what.
Out of Box Experience - Tuesday, November 16, 2010 - link
Sandforce controllers give you the "illusion" of speed by writing less data to flash than contollers without hardware compressionIf I wanted to test the speed of a copy and paste involving 200MB of data in the flash cells of a sandforce based controller, how can I tell exactly how much data is in the flash cells?
I mean, would Windows tell me how much compressed data is represented in the flash cells (200MB), or would Windows tell me how much compressed data is in the cells (maybe only 150MB) ?
The only way I can see fairly comparing an SSD with hardware compression and one without is to be sure you are actually writing the same amount of data to the flash cells (in this case - 200MB)
If sandforce based SSD's will only tell you how much data is represented and not what is actually on the drive, then I think the best way would be to use data that cannot be compressed
The tests I described in another post here involved copying and pasting 200MB of data which took 55 seconds on an ATOM computer with a Vertex 2
200MB / 55 sec = 3.6MB/sec
But if the 200MB was only a representation and the actual amount of data was for example 165MB in flash, then the actual throughput of my Vertex was even worse than I thought (In this case - 165MB / 55sec = 3.0MB/sec)
I need to know exactly how much data is indeed in flash or I need to start using non-compressible data for my tests
Make sense?
Out of Box Experience - Monday, November 15, 2010 - link
There has to be an missing pieces in our performance test, something that these companies knows and we dont.------------------------------------------
Like smoke & Mirrors?
Sandforce Controllers Compress the data to give you the impression of speed
check the speed without compression and then compare drives
pvdw - Monday, November 15, 2010 - link
"check the speed without compression and then compare drives "That makes no sense!
It's like saying disable branch predictions on one processor because another doesn't have it. Or disable the turbo on one car when comparing to a competing car that is naturally aspirated.
Out of Box Experience - Monday, November 15, 2010 - link
It makes all the sense because it would be a fair comparison of all the SSD's availableNot all drives have compression and not all data is compressible so why not test them all without compression?
Are you afraid that OCZ would be at the bottom of the list in a fair test like this?
Well, you may be right!
Lets test them to be sure
.harm - Thursday, November 11, 2010 - link
I'm not an expert on SSDs and RAID. So excuse me if my question is a bit stupid.I always understood that it´s impossible to use TRIM with a RAID configuration because the RAID controllers can't ´pass' the TRIM commands. So the SSD performance would drop overtime when using RAID. Now the SSD controller has "always-on garbage collection". Does this mean the performance in a RAID configuration doesn't drop?
AnnihilatorX - Thursday, November 11, 2010 - link
I would think so, it would at least be better off than other drive with less aggressive garbage collection.sparky76 - Thursday, November 11, 2010 - link
Will controllers like this leave OS X with a performance advantage over Windows 7, precisely because OS X has no support for the TRIM command?It seems that Win7 will have some system overhead in running TRIM while any OS without TRIM support will not suffer this, as garbage collection will be left to the firmware in the drive.
DoktorSleepless - Thursday, November 11, 2010 - link
I would love to see loading and minimum frame rate tests from actual games like what was done here a while back.http://www.anandtech.com/show/2614/14
It's just really hard to visualize real world performance even from Vantage.
I would bet a game like Fallout 3 would benefit a lot in the minimum frame rate department since it's always loading new data from the hard drive.
Nickel020 - Thursday, November 11, 2010 - link
I second this. While the IOPS for Bench are a nice measure, I don't know how to actually translate that into real world performance differences. I.e. are the differences in IOPS between say the new Kingston drive and Sandforce drives actually noticeable, and if yes, how noticeable?I would like to see a review that shows how much faster a Sandforce drive is than my old Vertex 1 and X25-M G2.
Chloiber - Thursday, November 11, 2010 - link
I'd also like to see that, as I said before.Just look at this test from (german) CB as an example:
http://translate.google.com/translate?js=n&pre...
Huge differences in synthetic tests, but a normal desktop system is just too slow to actually benefit from this!
I'm also missing these kinds of tests here. I think your earlier tests were better - now you just go through synthetic tests and show us the results, this isn't the thing I expected 1 or 2 years ago from an SSD test from anandtech.com. I'd like to see more tests which actually measure performance a user really gets when doing everyday tasks.
pavlindrom - Thursday, November 11, 2010 - link
Wouldn't it be better to have a drive-based sort function run to test how responsive the drive is? I would guess it would stress all of the corners of SSD performance. Write a bunch, shuffle in small portions when flipping values. Maybe it wouldn't test sequential erase. I still think it would show great info.Sufo - Thursday, November 11, 2010 - link
Just a quick recommendation. I have one of these, and i can confirm that with windows 7 (ie TRIM) and a 6gbps bridge the performance is pretty delicious. Grab one, if you meet these requirements.Taft12 - Thursday, November 11, 2010 - link
Can you comment on any penalty for 3Gbps SATA?I'm not convinced any SSD can exhibit any performance impact of the older standard except in the most contrived of benchmarks.
Sufo - Thursday, November 11, 2010 - link
Well, i've seen speeds spike above 375MB/s tho ofc this could well be erroneous reporting on windows' side. I haven't actually hooked the drive up to my 3gbps ports so in all honesty, i can't compare the two - perhaps i should run a couple of benches...Hacp - Thursday, November 11, 2010 - link
It seems that you recommend drives despite the results of your own storage bench. It shows that the Kingston is the premier ssd to have if you want a drive that handles multi-tasking well.Sandforce is nice if you do light tasks, but who the hell buys an ssd that only does well handling light tasks? No one!
JNo - Thursday, November 11, 2010 - link
"Sandforce is nice if you do light tasks, but who the hell buys an ssd that only does well handling light tasks? No one!"Er... I do. Well obviously I would want a drive that does well handling heavy task loads as well but there are limits to how much I can pay and the cost per gig of some of the better performers is significantly higher. Maybe money is no object for you but if I'm *absolutely honest* with myself, I only *very rarely* perform the type of very heavy loads that Anand uses in his heavy load bench (it has almost ridiculously levels of multi-tasking). So the premium for something that benefits me only 2-3% of the time is unjustified.
Anand Lal Shimpi - Thursday, November 11, 2010 - link
That's why I renamed our light benchmark a "typical" benchmark, because it's not really a light usage case but rather more of what you'd commonly do on a system. The Kingston drive does very well there and in a few other tests, which is why I'd recommend it - however concerns about price and write amplification keep it from being a knock out of the park.Take care,
Anand
OneArmedScissorB - Thursday, November 11, 2010 - link
"Sandforce is nice if you do light tasks, but who the hell buys an ssd that only does well handling light tasks? No one!"Uh...pretty much every single person who buys one for a laptop?
cjcoats - Thursday, November 11, 2010 - link
I have what may be an unusual access pattern -- seeks within a file -- that I haven'tseen any "standard" benchmarks for, and I'm curious how drives do under it, particularly
the Sandforce drives that depend upon (inherently sequential?) compression. Quite possibly, heavy database use has the same problem, but I haven't seen benchmarks on that, either.
I do meteorology and other environmental modeling, and frequently we want to "strip mine" the data in various selective ways. A typical data file might look like:
* Header stuff -- file description, etc.
* Sequence of time steps, each of which is an
> array of variables, each of which is a
+ 2-D or 3-D grid of values
For example, you might have a year's worth of hourly meteorology (about 9000 time steps),
for ten variables (of which temperature is the 2'nd), on a 250-row by 500-column grid.
So for this file, that's 0.5 MB per variable, 5 MB per time step, total size 45 GB, with
one file per year.
Now you might want to know, "What's the temperature for Christmas Eve?" The logical sequence of operations to be performed is:
1. Read the header
2. Compute timestep-record descriptions
3. Seek to { headersize + 8592*5MB + 500KB }
4. Read 0.5 MB
Now with a "conventional" disk, that's two seeks and two reads (assuming the header is not already cached by either the OS or the application), returning a result almost instantaneously.
But what does that mean for a Sandforce-style drive that relies on compression, and implicitly on reading the whole thing in sequence? Does it mean I need to issue the data request and then go take a coffee break? I remember too well when this sort of data was stored in sequential ASCII files, and such a request would mean "Go take a 3-martini lunch." ;-(
FunBunny2 - Thursday, November 11, 2010 - link
I've been asking for similar for a while. What I want to know from a test is how as SSD behaves as a data drive for a real database, DB2/Oracle/PostgreSQL with 10's of gig of data doing realistic random transactions. The compression used by SandForce becomes germane, in that engine writers are incorporating compression/security in storage. Whether one should use consumer/prosumer drives for real databases is not pertinent; people do.Shadowmaster625 - Thursday, November 11, 2010 - link
Yes I have been wondering about exactly this sort of thing too. I propose a seeking and logging benchmark. It should go something like this:Create a set of 100 log files. Some only a few bytes. Some with a few MB of random data.
Create one very large file for seek testing. Just make an uncompressed zip file filled with 1/3 videos and 1/3 temporary internet files and 1/3 documents.
The actual test should be two steps:
1 - Open one log file and write a few bytes onto the end of it. Then close the file.
2 - Open the seek test file and seek to random location and read a few bytes. Close the file.
Then I guess you just count the number of loops this can run in a minute. Maybe run two threads, each working on 50 files.
Shadowmaster625 - Thursday, November 11, 2010 - link
Intel charging too much, surely you must be joking!Do you know what the Dow Jones Industrial Average would be trading at if every DOW component (such as Intel) were to cut their margins down to the level of companies like Kingston? My guess would be about 3000. Something to keep in mind as we witness Bernanke's helicopter induced meltup...
melgross - Thursday, November 11, 2010 - link
So, with all of this info presented, which drives would be best as a Photoshop scratch drive?Out of Box Experience - Tuesday, November 16, 2010 - link
Any modern platter based drive should be fine for a photoshop scratch driveKlober - Thursday, November 11, 2010 - link
This is my perspective from what I see throughout the review:The Corsair Force drives and the Crucial RealSSD drives are very close performance-wise in most of the comparisons; however, when price is included in the scenario the Corsair Force drives (60GB/120GB) win hands down.
The only true advantage the C300 has over the Force drives from what I can tell is in scenarios where the data is mostly uncompressible - I don't think this comes close to making up for the price disadvantage of the C300 drives since most likely users will have a large mechanical drive for their storage drive rather than opting to use their C300 for that purpose.
Taking into account the fact that you can get the Corsair Force 60GB SSD for $5 more than the 40GB at NewEgg, I would say that pretty much nullifies the cost/GB advantage of the C300 at the lower capacity - at that point you're looking at $2.109/GB for the C300 and $2.031/GB for the Force 60GB assuming it has 64GB NAND capacity (that's going by dividing the Force 120GB specs in half...if we instead use the specs from the Force 40GB that would put the Force 60GB at an amazing $1.805/GB for a low capacity SSD!).
If I were to buy an SSD tomorrow it would almost definitely be the Corsair Force 60GB for its fantastic (at least by SSD standards) cost/capacity at a low price point.
Anyway, that's my 2 cents!
Aikouka - Thursday, November 11, 2010 - link
Why are you assuming the Corsair drive magically has more space than it actually lists? If it's a 60GB drive, then you only get 60GB of usable space with some other amount set aside as a scratch area.Personally, I base my purchases more on random read than anything. Based on the prices, I'd most likely pick up the Crucial C300 64GB if I was looking for a new boot drive.
7Enigma - Thursday, November 11, 2010 - link
Just a simple request for a G2 80gig Intel drive for future reviews. I think many of us have one of these and know the 160gig drive is significantly faster in some tests (sequential write), but almost identical in others (random read I believe).DanNeely - Thursday, November 11, 2010 - link
Are there any tools available that will let you see how much of the write capacity of an SSD you've exhausted?Iketh - Thursday, November 11, 2010 - link
i own an intel 40gb and intel's trim software reports it... it's part of a drive's SMART data, so any SMART reading program should be able to tell you, but i could be mistakenMadMan007 - Thursday, November 11, 2010 - link
"Note that not all SandForce drives are created equal here. If a manufacturer doesn’t meet SandForce’s sales requirements, their drives are capped at a maximum of 50MB/s here. This is the case with the Patriot Inferno, although OCZ’s Agility 2 voluntarily enforces the limit."I understand there is the 'higher performance' Sandforce firmware (Vertex 2 vs Agility 2 for example) and the Agility 2-type firmware is the one that gets 'capped,' but what exactly do you mean by 'meet Sandforce's sales requirements'? Is this to say that the firmware the retail companies are licensed to use and able to ship is restricted based upon sales figres/order quantity, so that if say OCZ buys 100k SF controllers while Patriot only buys 10k, Patriot is limited in which firmware they are allowed to offer? If it's not that, what exactly do you mean? And most importantly how can we as end-users making a purchase determine this difference, is looking at IOPS all that's needed?
7Enigma - Thursday, November 11, 2010 - link
Anand,I didn't like the wording of the power consumption in terms of not being better at load than a mechanical hard drive. Just like your praise of Intel's hurry up and get idle (with turbo), all of these SSD's have an advantage in bursting and then getting idle.
I had mentioned it in a previous SSD article but what we really need is a moderated workload suite. Something like virus scan and total power draw during that time. While load voltages might be the same or higher for the SSD's, the total power consumption during that time will (should) be less. I think a virus scan is perfect on a normal system as it should take 10-20min and give a measureable difference between drives. Grab a laptop from one of your recent reviews (not an Atom or other super-slow mobile to prevent the SSD from waiting around), and measure the total power drawn over the scan.
This is less important in a desktop obviously but for the mobile crowd (whom sees the biggest benefit over the traditional laptop drives anyway) might just be one more criteria to help make a decision (or just show that it's the type of drive (mech/SSD) and not really differences between the SSD's themselves).
Glad to have a new SSD review though. I love reading these and am trying to justify upgrading from my 80gig Intel G2 drive that has been fantastic for the last year. The great thing about the SSD tech is that you can hand-me-down smaller capacity, slower drives and completely revolutionize an older system.
Thanks again.
JohnBooty - Thursday, November 11, 2010 - link
"Kingston ships the V+100 with a 3-year warranty and to Kingston's credit I haven't had any other drives die as a result of wearing out the NAND. Even if the V+100 has higher effective write amplification than the competition, your usage model will determine whether or not you bump into it."Anand,
Firstly, thank you for being the industry standard when it comes to articles about SSD. It's a favorite topic of mine and your articles have steered a lot of purchases in my personal and professional circles (and you haven't steered us wrong!)
My request...
One of the lingering questions about SSDs is how long they'll actually last in the real-world before the NAND is fried. When I see you mentioning that you've never actually fried any NAND, that seems like an opportunity for you to give it a shot and give the public some answers.
I'd love to see a "torture test" article where you run simulated workloads against SSDs 24/7 until they can't take any more writes, and report on the results.
Gonemad - Thursday, November 11, 2010 - link
Yes, a longevity test. Put it on grind mode until it pops. P51 Mustangs benefited from engines tested that way.Now, this one, should it be fully synthetic or more life like? Just place the drive in "write 0, write 1" until failure, record how many times it can be used, or create some random workload scripted in such a manner that it behaves pretty much like real usage, overusing the first few bytes of every 4k sector... if it affects any results. What am I asking is, will it wear only the used bytes, or the entire 4k rewritten sector will be worn evenly, if I am expressing myself correctly here.
On another comment, I always thought SSD drives were like overpriced undersized Raptors, since they came to be, but damn... I hope fierce competition drive the prices down. Way down. "Neck and neck to mag drives" down.
And what about defragmenting utilities? Don't they lose their sense of purpose on a SSD? Are they blocked from usage, since the best situation you have on a SSD is actually randomly sprayed data, because there is no "needle" running over it at 7200rpm in a forcibly better sequential read? Should they be renamed to "optimization tools" when concerning SSD's? Should anybody ever consider manually giving permission to a system to run garbage collection, TRIM, whatever, while blocking it until strictly necessary, in order to increase life span?
Iketh - Thursday, November 11, 2010 - link
win7 "automatically disables" their built-in defrag for SSDs, though if you go in and manually tell it to defrag an SSD, it will do it without questionprefetch and superfetch are supposedly also disabled automatically when the OS is installed on an SSD, though I don't feel comfortable until i change the values myself in the registry
cwebersd - Thursday, November 11, 2010 - link
I have torture-tested a 50 GB Sandforce-based drive (OWC Mercury Extreme Pro RE) with the goal to destroy it. I stopped writing our semi-random data after 21 days because I grew tired.16.5 TB/d, 360 disk fills/day, 21 days more or less 24/7 duty cycle (we stopped a few times for an hour or two to make adjustments)
~7500 disk fills total, 350 TB written
The drive still performs as good as new, and SMART parameters look reasonably good - to the extent that current tools can interpret them anyway.
If I normally write 20 GB/d this drive is going to outlast me. Actually, I expect it to die from "normal" (for electronics) age-related causes, not flash cells becoming unwritable.
Anand Lal Shimpi - Thursday, November 11, 2010 - link
This is something I've been working on for the past few months. Physically wearing down a drive as quickly as possible is one way to go about it (all of the manufacturers do this) but it's basically impossible to do for real world workloads (like the AT Storage Bench). It would take months on the worst drives, and years on the best ones.There is another way however. Remember NAND should fail predictably, we just need to fill in some of the variables of the equation...
I'm still a month or two away from publishing but if you're buying for longevity, SandForce seems to last the longest, followed by Crucial and then Intel. There's a *sharp* fall off after Intel however. The Indilinx and JMicron stuff, depending on workload, could fail within 3 - 5 years. Again it's entirely dependent on workload, if all you're doing is browsing the web even a JMF618 drive can last a decade. If you're running a workload full of 4KB random writes across the entire drive constantly, the 618 will be dead in less than 2 years.
Take care,
Anand
Greg512 - Thursday, November 11, 2010 - link
Wow, I would have expected Intel to last the longest. I am going to purchase an ssd and longevity is one of my main concerns. In fact, longevity is the main reason I have not yet bought a Sandforce drive. Well, I guess that is what happens when you make assumptions. Looking forward to the article!JohnBooty - Saturday, November 13, 2010 - link
Awesome news. Looking forward to that article.A torture test like that is going to sell a LOT of SSDs, Anand. Because right now that's the only thing keeping businesses and a lot of "power users" from adopting them - "but won't they wear out soon?"
That was the exact question I got when trying to get my boss to buy me one. Though I was eventually able to convince him. :)
Out of Box Experience - Thursday, November 11, 2010 - link
Here's the problemSynthetic Benchmarks won't show you how fast the various SSD controllers handle uncompressible data
Only a copy and paste of several hundred megabytes to and from the same drive under XP will show you what SSD's will do under actual load
First off, due to Windows 7's caching scheme, ALL drives (Slow or Fast) seem to finish a copy and paste in the same amount of time and cannot be used for this test
In a worst case scenario, using an ATOM computer with Windows XP and Zero SSD Tweaks, a OCZ VERTEX 2 will copy and paste data at only 3.6 Megabytes per second
A 5400RPM laptop drive was faster than the Vertex 2 in this test because OCZ drives require massive amounts of Tweaking and highly compressible data to get the numbers they are advertizing
A 7200RPM desktop drive was A LOT faster than the Vertex 2 in this type of test
Anyone working with uncompressible data "already on the drive" such as video editors should avoid Sandforce SSD's and stick with the much faster desktop platter drives
Using a slower ATOM computer for these tests will amplify the difference between slower and faster drives and give you a better idea of the "Relative" speed difference between drives
You should use this test for ALL SSD's and compare the results to common hard drives so that end users can get a feel for the "Actual" throughput of these drives on uncompressible data
Remember, Data on the Vertex drive's is already compressed and cannot be compressed again during a copy/paste to show you the actual throughput of the drive under XP
Worst case scenario testing under XP is the way to go with SSD's to see what they will really do under actual workloads without endless tweaking and without getting bogus results due to Windows 7's caching scheme
Anand Lal Shimpi - Friday, November 12, 2010 - link
The issue with Windows XP vs. Windows 7 doesn't have anything to do with actual load, it has to do with alignment.Controllers designed with modern OSes in mind (Windows 7, OS X 10.5/10.6) in mind (C300, SandForce) are optimized for 4K aligned transfers. By default, Windows XP isn't 4K aligned and thus performance suffers. See here:
http://www.anandtech.com/show/2944/10
If you want the best out of box XP performance for incompressible data, Intel's X25-M G2 is likely the best candidate. The G1/G2 controllers are alignment agnostic and will always deliver the same performance regardless of partition alignment. Intel's controller was designed to go after large corporate sales and, at the time it was designed, many of those companies were still on XP.
Take care,
Anand
Out of Box Experience - Saturday, November 13, 2010 - link
Thanks AnandThats good to know
With so many XP machines out there for the foreseeable future, I would think more SSD manufacturers would target the XP market with alignment agnostic controllers instead of making the consumers jump through all these hoops to get reasonable XP performance from their SSD's
Last question..
Would OS agnostic garbage collection like that on the new Kingston SSD work with Sandforce controllers if the manufacturers chose to include it in firmware or is it irrelevant with Duraclass ?
I still think SSD's should be plug and play on ALL operating Systems
Personally, I'd rather just use the drives instead of spending all this time tweaking them
sheh - Thursday, November 11, 2010 - link
This seems like a worrying trend, though time will tell how reliable SSDs are long-term. What's the situation with 2Xnm? And where does SLC fit into all that regarding reliability, performance, pricing, market usage trends?DanNeely - Thursday, November 11, 2010 - link
The 2xnm lifetimes will be shorter. The flash lifetime is a function of feature size, so as processes shrink lifetime will as well. At the same time, it's only likely to be a concern if you're doing very IO intensive activities. My 128GB Indilinx drive is at 96% after a year in my main desktop. At this rate I'll be nowhere near the maximum write limit when the flash dies of old age in another 4-9 years even if I keep it as my main drive the whole time which is doubtful.If scaling is linear with feature area, as the 50% drop in life-cycles from 50 to 34 nm implies, 22nm 2 level flash will last about 2500 write cycles. This still will probably be long enough not to matter. 14nm flash and beyond might not be unless they switch to SLC for consumer SSDs, and only use MLC for memory cards and thumb drives. With 10x the allowed number of write cycles SLC flash should remain good for several additional process shrinks.
OTOH that might be a moot point because the power levels needed to write flash doesn't drop with each process, and unless something changes flash is expected to hit a will in the teens because the wires on the chip won't be able to carry the load without melting.
Iketh - Thursday, November 11, 2010 - link
Anand, couldn't you just continue running your random write test 24/7 until it falls dead, or is this not feasible?Maybe it's not feasible with the Kingston drive because the random write performance is too slow? Which could bring up speculation that it's intentionally handicapped so this type of test is avoided?
Skiprudder - Thursday, November 11, 2010 - link
Hello Anand,Really interesting article as usual. I had no idea that there was such a gulf between various Sandforce controllers. There was one line however that perplexed me, "The V+ 100‘s sequential read speed is excellent, just a hair above the top drives from Intel and Crucial. There’s not much room for improvement here unless you go to a 6Gbps interface."
Why didn't you use a Sata 6 controller? I think that most of us buying boards today will be looking at Sata 6 since so many today's current SSDs clearly outstrip Sata 3, to say nothing of the upcoming 500MB/s and better drives coming down the pipeline. I'm sure you have a good reason, and I know there has been some weird controller issues with AMD I believe, but I'm very curious! Thank you.
Casper42 - Thursday, November 11, 2010 - link
Just saw this on Gizmodohttp://slickdeals.net/forums/showthread.php?t=2365...
Puts it at around $1.60/GB and if you skip past the synthetic benchmarks and look at AT Storage Bench, its a pretty formidable drive. Makes me a little sad I spent more than this on my 160 X25M G2 a while back.
BoboGO - Thursday, November 11, 2010 - link
The true cost vs. useability (usable nand) price breakdowns are as follows... I think... :)SSD Price Comparison - November 11, 2010
SSD Useability % Price $/GB of USEABLE NAND
Western Digital SiliconEdge Blue 93.13% $214.99 $1.80
Corsair Nova V128 128GB 93.13% $219.99 $1.85
Kingston SSDNow V Series 128GB 93.13% $224.99 $1.89
Corsair Force F120 120GB 87.34% $229.99 $2.06
OCZ Agility 2 120GB 87.34% $229.99 $2.06
OCZ Vertex 2 120GB 87.34% $234.99 $2.10
Crucial RealSSD C300 64GB 93.13% $134.99 $2.26
Crucial RealSSD C300 128GB 93.13% $269.99 $2.27
Kingston SSDNow V+ Series 128GB 93.13% $277.00 $2.32
Kingston SSDNow V+ 100 128GB 93.13% $278.99 $2.34
Intel X25-V 40GB 93.25% $ 94.99 $2.55
Patriot Inferno 60GB 87.34% $149.00 $2.67
Intel X25-M G2 160GB 93.13% $409.00 $2.74
Kingston SSDNow V Series 30GB 93.00% $ 82.99 $2.97
Corsair Force F40 40GB 77.71% $124.99 $3.35
SeetheSeer - Thursday, November 11, 2010 - link
Brilliant...just needs controller info added. I think I'll do some searching later tonight if I have time.Is total capacity listed in GB (10^9) or GiB(2^30)? What about usable capacity? I wish people paid more attention to this, especially in area such as this were flash storage (typically in GiB) is being used for a drive (typically in GB).
SeetheSeer - Thursday, November 11, 2010 - link
Anand,Do you think a short description of which controller each drive uses could be added to the table on page 2? It seems to me that it would be much more instructive to sort the drives based on this, rather than the often rebadged brand.
Out of Box Experience - Sunday, November 14, 2010 - link
How about this info when comparing SSD's? >1. Is it truly Plug and Play on any OS?
2. Does it have Alignment Agnostic Controllers?
3. Does it handle incompressible Data really well under XP?
Since the majority of computers still run XP, this information is the most valuable to the most people
snakyjake - Thursday, November 11, 2010 - link
There's a lot of stats for the drives, however I'm not any closer in choosing a drive.I have 3 systems:
1. Windows 7. This system is mostly lots of opened web browsers hogging memory, email, maybe a Word document, video encoding, and reference software (i.e. encyclopedia). I assume the browser is caching to the drive, same with the virtual memory and other background apps. I don't turn the machine off, so boot time is not important. For video encoding, I'd like to keep the data on a HDD since it is a lot of storage that SDD's are not affordable at storing, and it is most likely CPU bound.
2. Media Center. Want a SDD for the main drive (OS and apps), but will use a larger/slower drive for the media storage. Need something low heat. This is what I really want a SDD for.
3. Linux. Mostly emailing and web browsing. All the apps I use are cloud apps.
Obviously the fastest drive would be the one I want if money is no object. But I want to know how much faster my experience is going to be for how much more money? My goal is a snappy system and less time waiting around. I want to know what I'd sacrifice by getting a "value" drive versus the higher quality.
Also, MTBF is important. I hate losing data! Hope the drives have SMART or some other methods of warnings.
Thanks,
Jake
Out of Box Experience - Thursday, November 11, 2010 - link
If all you want to do is boot up quickly and load your apps really fast, then a sandforce drive is perfect for youIf you want to manipulate data already on the drive or copy & paste to the same drive, then you are much better off with a 7200RPM desktop drive as far as I can tell
Anyone wanting to spend their time doing actual work instead of constantly tweaking their SSD to get synthetic benchmarks that have little bearing on real world results should avoid sandforce based drives like the vertex 2
Please read my other posts in this thread for more info
7Enigma - Friday, November 12, 2010 - link
I've ready your other posts and you are selecting a very specific situation which most people will not encounter (that is copy/paste using XP with no SSD tweaks). That's like putting in an NVIDIA video card and not installing the drivers and complaining about performance. The XP OS was designed so far before SSD's were even available to enterprise markets let alone the consumer space its no wonder they don't perform well.The vast majority of people are no longer on XP (trust me I was one of the last to hold out due to my dispise for Vista). But you are artificially setting limitations on a product and then using it to generalize performance in all situations that are to you "real world".
I have real concerns about the Sandforce drives due to variable performance. I have an Intel 80gig G2 SSD and am still amazed at how well it does (to me) in the most important benchmarks, that is random read/write and sequential read. Even with the cheap(er) pricing of SSD's and the *hopefully* significant decrease once we get to the next node with Intel's 3rd generation these are still not capable of storing the majority of our data (unless you think 1-2 grand is chump change). That being said these are mainly going to be access drives and rarely used for moving large amounts of data around (which is why the sequential write becomes much less important for day to day use, besides the odd large install such as a video game).
I keep my 80gig with about 45GB free. Win7 install (including User data), a couple programs, the current 1-3 games I'm playing and that's it. I have a 250GB secondary 7200rpm drive for music/movies/etc. and a larger external HDD for backup and infrequently used data.
Out of Box Experience - Friday, November 12, 2010 - link
Quote:these are mainly going to be access drives and rarely used for moving large amounts of data around (which is why the sequential write becomes much less important for day to day use, besides the odd large install such as a video game).
---------------------------------------------------------------------------------------
In that case, I'd grab a Crucial Real SSD
Best read speeds now on Sata 2 and reasonable for Sata 3
64GB is very reasonably priced for the read speeds you will get
Out of Box Experience - Monday, November 15, 2010 - link
Quote:The vast majority of people are no longer on XP (trust me I was one of the last to hold out due to my dispise for Vista). But you are artificially setting limitations on a product and then using it to generalize performance in all situations that are to you "real world".
-------------------------------------------------
Where did you get that incorrect data?
It will be a few years before Windows 7 sells more copies than XP
XP is still used on more computers than Vista and Windows 7 combined!
Any SSD manufacturer who misleads their customers into thinking that their SSD's are plug and play compatible with XP by omission of pertinent information should go into spam marketing!
Why should the consumer spend another $200 for a new OS to use a Vertex when Intel SSD's work just fine with XP without all the tweaks?
ceomrman - Friday, November 12, 2010 - link
I can't stand all the SSD articles with no real world benchies at all. I don't care about transfer speed or even IOPS. I care about how fast it boots, how fast it loads my work, displays my porn, plays my games, etc.slickr - Friday, November 12, 2010 - link
I only see theoretical performances of the drives, while this is some indication of how it would perform in real case scenarios, its not the most accurate and its hard to translate into real world.I would like some tests that test all of the SDD's abilities at once.
For example running a virus scan while pasting 2GB or big and small data, while moving hundreds of pictures into a image viewer, all on top while downloading a torrent and surfing the web.
B - Friday, November 12, 2010 - link
Great Table. Would be nice if there was a column identifying which controller was used in each of the drives.snakyjake - Friday, November 12, 2010 - link
Objective performance measurements are good to have, however these comments are suggesting some difficulties in interpretation because the majority of us don't know how to relate to a synthetic benchmark. My recommendation is to create some user profiles (gamer, video artists, graphics artist, serious web surfer, media PC, low power, etc). Have some people utilize the machines and ask for their feedback.Because here's my bottom line: I don't want to pay for more than I'll notice. If PEOPLE can't tell the difference between a top synthetic rated drive versus middle of the pack....then I don't want to pay extra.
For example: If a more expensive SSD drive will not speed up my Media Center PC, I don't want to pay the extra. I want to know what a real person thinks and experiences. Rather talk to my buddies than get comments from a robot performing synthetic tests.
GullLars - Friday, November 12, 2010 - link
Dear Anand, I would love a reply here, since I've raised these points in the past, and been ignored."Concerned enough to recommend running it with 20% free space at all times (at least). The more free space you have, the better job the controller can do wear leveling."
I feel the need to critique this statement. If TRIM is not active, there is no difference whatsoever if you use 10% or 100% of the LBAs avalible to the user, since the drive can't tell the difference.
The only way you could make sure 20% of the NAND is free at all times is to not partition it in the first place (or not include it in a partition after a secure erase).
Also, you've included 4KB random write at QD 32, but still not 4KB random read. If you compare your results, you will see a much larger difference between read performance as QD scales than you do for write. This is due to write coalessing and attenuation, while reads don't have this benefit.
I feel both random read and write @ QD 32 should be included to show what the drives are capable of within the NCQ spec. Ideally would be a graph showing IOPS by QD scaling, but i understand that it may be a bit much work.
When benchmarking my own SSDs, i scale 4 dimentions in IOmeter; Read:write ratio, seq:ran ratio, block size, and QD. Due to size of the data set, i only scale them against each other at 4KB and a couple of larger sizes. A few thousand data points is enough pr drive :P
Ao1 - Saturday, November 13, 2010 - link
Quote:“SandForce's partners who have to pay a big chunk of their margins to SandForce as well as the NAND vendor are actually delivering the best value in SSDs. Kingston and Western Digital also deliver a great value. Not Crucial/Micron and not Intel, which is not only disappointing but inexcusable. These companies actually own the fabs where the NAND is made and in the case of Intel, they actually produce the controller itself”.
Are you forgetting that Intel spent over 2 billion on the JV with Micron? They are the ones carrying the investment risk, not the likes of OCZ.
Also Intel spend money and time on getting their products right before market and use much higher quality components throughout, not least better QA’d NAND. (Intel do not sell SSD grade NAND. They sell NAND that has to be further processed by whoever buys it).
You have consistently ignored these facts and they make a big difference.
Whilst I’m at it I can’t understand your assertion that SF drives are so good. There are countless problems with them even now after numerous f/w updates. They degrade badly with uncompressible data, they have weak GC and they don’t trim in the same way as other drives. It’s quite easy to see performance drop by 50%, even OCZ have admitted that. Read performance, without writing also degrades.
I used to look forward to your reviews but it seems your objectivity when it comes to the X25 and C300 in favour of SF drives is seriously lacking.
Ao1 - Saturday, November 13, 2010 - link
From OCZ's forum:"Here is the thing, if you hammer the driver with say enough writes that the drive would under normal use/see in 7 days within a few hrs, the drive will slow down for 7 days, maybe longer. It does this to protect the nand life. So your guys seeing a 50% drop may actually see 30% which is the normal drop, then a further 20% because at some stage they have hammered the drive and then not realised its going to take 5 days or longer for the speed to creep back up. Also remember this write quantity slowdown is further impacted by how you use the drive after you have hammered it."
http://www.ocztechnologyforum.com/fo...l=1#post528...
eckre - Saturday, November 13, 2010 - link
So this seems like a perfect drive for a RAID because of it's aggressive always on garbage collection would make up for lack of TRIM support. ?Am I correct? Thoughts?
JohnWPB - Saturday, November 13, 2010 - link
I have built probably 10 pc's in the last 2 yrs, and wonder if adding a SSD as a boot drive would be worth the effort.I see you guys making lots of comments, but as a newbie on SSD's, I would really like to see a thread on how to introduce a SSD to a system that has a bunch of SATA drives.
Exactly how does one go about installing Windows XP Pro? I imagine that I should disconnect my SATA drives and just hook up the SSD and then do a CD install?
Can programs like CasperXP ( mirror imaging program ), be used to do a mirror image backup from an SSD to a Sata drive ( C: boot drive SSD to D: mirror drive SATA ).
My apologies if this was posted in the wrong thread.
Out of Box Experience - Saturday, November 20, 2010 - link
QUOTE:Exactly how does one go about installing Windows XP Pro? I imagine that I should disconnect my SATA drives and just hook up the SSD and then do a CD install?
---------------------------------------------------------------------------------------------------------
You are correct as far as I can tell...
I just spent the past 2 days retesting the copy/paste speed of a Vertex 2 with the alignment tweak recommended by OCZ
There was no speed benefit resulting from the alignment of the partition as OCZ described
Copy and paste speed was still a miserable 3.6 megabutes per second
But just to be on the safe side, I will start using ONLY Intel SSD's with Windows XP
Anand says Intel controllers are alignment agnostic and should be the best for non-compressible data under XP
So, untill other SSD manufacturers start making alignment agnostic controllers that are capeable of doing a copy and paste faster than my 5400RPM laptop drive under XP, I think I'll just stick with Intel
(Better Flash Quality as well)
Ezekeel - Saturday, November 13, 2010 - link
Hi Anand,you noted that due to the aggressive garbage-collection this SSD will have a large write-amplification, so the flash will wear-out more quickly. What would happen when using this SSD with an OS that supports TRIM? Will TRIM support the internal GC and lead to a smaller write-amplification and thus a longer lifetime of the SSD?
mpx - Sunday, November 14, 2010 - link
Can you test how VMs behave on SSD? I understand it's partially there in synthetic IOPS and bandwith tests - that show what can you get max. from these drives. But "normal" load from multiple VMs should be different than normal load from a single OS, so drives that do good at normal load in current test migth be not performant under normal load of multiple VMs.This test with few VMs makes sense for larger drives (160, 256GB), as there just won't be enough place for multiple OSes on small drives. Also longevity becomes more important when you have few times larger load.
tecknurd - Sunday, November 14, 2010 - link
I use Linux and its TRIM command support is sparse. Finally a drive that does not require the TRIM command. I know it comes at a cost of more wear and tear, but I do not mind because pros out weighs the cons. It will give me no noise from the storage system and it be about 60 times quicker than my hard drive to access files. I do not think $280 for 128 GB is not a bad deal for a SSD that does not require the TRIM command to function great. Though, one thing that comes to mind is does the Kingston SSDNow V+100 come with ECC because all hard drives includes ECC for data integrity.Ezekeel - Sunday, November 14, 2010 - link
Why sparse? The kernel supports TRIM since 2.6.33 (early 2010).tecknurd - Wednesday, November 17, 2010 - link
True, but the file system also have problems with SSD. Not all file system in Linux are SSD aware. EXT3 and EXT4 does not have good support to handle SSD.Ezekeel - Friday, November 19, 2010 - link
Linux filesystems ext3/ext4 are designed for magnetic drives - but that is also true for ntfs and hfs+, so the problems with SSD are not unique to Linux. However, it is true that while at least Windows7 does some automatic optimizations, you have to configure everything manually under Linux.sprockkets - Sunday, November 21, 2010 - link
All you have to do is format the drive with ext4, add the discard option, and you are set. TRIM is enabled automatically in the background; no need to manually run the command anymore.Even with the latest firmwares, this isn't an issue anyway.
Ezekeel - Thursday, December 9, 2010 - link
That is all you have to do to enable TRIM, yes.However, I was talking about optimizations for an SSD (https://wiki.archlinux.org/index.php/SSD#Tips_for_... http://cptl.org/wp/index.php/2010/03/30/tuning-sol... which you all have to do manually while Windows7 afaik does at least some optimizations automatically if you install it on an SSD, like disabling (Super)PreFetch and indexing. Also partition setup with diskpart under Windows automatically takes care of a proper partition alignment while you still have to do it manually under Linux (http://randomtechoutburst.blogspot.com/2010/03/4k-...
ClagMaster - Monday, November 15, 2010 - link
You are right Mr Shimpi that this is confusing and I got a headache.I am not certain what I am going to get if I order a Kingston Drive. Performance is so diverse and the model nomenclature so similar.
I am going to get a OCZVertex 2 or an Intel X-25 G2 instead. I know what I am getting with these brands.
psyside1 - Sunday, November 21, 2010 - link
Hi Anand and all who read this.
Let me start of by saying that i'm new to SSD tech (noob) and i'm starting to learning but there are some things which where not pointed in the reviews as far as i remember.
Heck, even in this review you said that Inferno is somewhat capped to 50 mb/s if i'm not wrong?
So in short,
Where i live i'm limited to 2 models, one is Patriot Inferno and the Other A-DATA S599 with possibly 50K IOPS firmware, as Newegg and Amazon specs confirm?
Now, does that firmware insure i won't get slowdowns (at some points) like the Inferno model in this review? or there is some more about it. if there is, i really got no idea how is that possible same controller, same build? and heck in most of the test the Vertex 2 is still fastest, even faster then the other drives which now share the 50K IOPS firmware, G.Skill Phoenix pro, Corsair Force etc??
i'm mostly interested in SSD which will have good read/write speeds (4K etc) and to be fast during programs/games installation. I got 6 pcs connected in network so that means alot to me, in short does my needs require higher IOPS firmware or i would not notice any difference in performance during installation of big size programs/games.
Also, is it possible to know what revision you get without actually buying the product? is there any info on the package/sticker or part number? the difference in performance i noticed in some reviews was 2x higher random write (4K) speed on the models with OCZ "exclusive firmware. And as far that point goes i'm totally clueless does it really matter at all, and if its not how that that translate in real world usage?
Please answer i don't want to regret my purchase :(
Thanks and sorry for slight off topic :)
psyside1 - Sunday, November 21, 2010 - link
Sorry for double post there is no edit option, actually Madman007 was asking the same i did on the more appropriate way, my English is bad.
tno - Sunday, May 1, 2011 - link
"Remember that NAND is written to at the page level (4KB), but erased at the block level (512 pages)." I think you meant '512 KB.'Gaucherre - Friday, May 6, 2011 - link
Instead of $259, the Kingston V100+ 96GB is available for $119.99 after rebate from Buy.com. This pricing completely changes the value rating from acceptable to outstanding value! The mail-in rebate is rotating from one online store to another. Last month it was at Newegg.com; right now it's at Buy.com. Next month - who knows? Anyway, the cost per Gigabyte when tested here at Anandtech.com was around $2.70. Now it's $1.25 per Gigabyte, and a 96GB drive is large enough for Windows plus quite a few installed programs and games. Pretty good value. Think I just talked myself into ordering one .......