Back to Article

  • Mumrik - Friday, November 16, 2012 - link

    10GiB a day would be a very slow day for me, so these numbers are a little scary looking. Not that I expect to have a smallish SSD for 10 years, but still... Reply
  • jmke - Friday, November 16, 2012 - link

    Using a PC drive 5 years is not a very long time with the IT industry stagnating hardware wise. I would avoid these drives for anything you plan to have work for more than 5 years :) Reply
  • irev210 - Friday, November 16, 2012 - link

    jmke, the numbers are based off SMART, not actual, which is conservative.

    Second, 20GiB of writes a day for 5 years is very aggressive for a typical client workload.

    Anyone that is doing that sort of writes is probably doing specialized work that requires something a bit higher-end than your typical office client.
  • Alexvrb - Friday, November 16, 2012 - link

    Exactly. If you are doing something like 30GB+ of WRITES per day, you should have a good MLC drive like the 840 Pro, or even an SLC scratch drive + larger MLC drive.

    The non-Pro, TLC 840 is strictly a consumer drive, and it will easily last 5-10 years in a typical consumer system.
  • robmuld - Friday, November 16, 2012 - link

    I don't understand how you can assume SMART values are worse than actual ones? How many times has a product had a problem when it's "supposed" to be working.

    This is something Anandtech should test for themselves, rather than take the manufacturer at their word. Do a continuous write test and see how many weeks it takes to fail. I'm pretty sure you'll encounter some surprises.
  • Kristian Vättö - Friday, November 16, 2012 - link

    A premature death is always a possibility but usually that's not due to the NAND endurance.

    Totally killing the drive is obviously the best way to test endurance but the problem is that it will take weeks, possibly even months to complete and I couldn't test any other SSDs during that period.
  • robmuld - Friday, November 16, 2012 - link

    Just setup a cheap box with SATA3 and let it run. Why does it have to occupy the main test rig? After doing the main benchmarks stick the drive in the torture chamber. It would add a lot of value to reviews.

    Remember when the stuttering problem was first discovered? It took some extra effort but it was worth it, and drive manufacturers ended up changing their products as a ersult
  • Sivar - Friday, November 16, 2012 - link

    Read the link xtremesystems link and then decide if Anandtech is blindly taking manufacturers on their word. Reply
  • MrSpadge - Friday, November 16, 2012 - link

    I don't think you understand which SMART values are being talked about here. Sure, if SMART doesn't find a problem with an HDD it doesn't have to mean anything. But if it finds a problem, you can be sure there is a problem.

    Here, however, we're talking about the amount of data written to the drive logged and reported via SMART. And the manufacturer is calculating the wear level indicator from these values. However, they have to be conservative here - a dying drive still showing 50% SSD life left would get them into troube. Hence real world write endurance will be better than the wear indicator reported by SMART implies (as that link at XS shows).
  • UpSpin - Friday, November 16, 2012 - link

    Your calculations look nice but are they really how SSDs will get used in real life?
    If you have a 128GB SSD with your OS and programs installed on it, you probably have used about 60GB. Then load some other data on it or even games and let's say you're left with a generous 20GB free space on the SSD.
    Most of the occupied space in the NAND won't change because system/prgram files don't get changed at all.
    So you probably only change 10GiB per day as you said, but the 60GB system files won't get changed, ever.
    So if data gets changed it will occur on the latter 60GB.
    Because 'only' 20GB are free, wear leveling will have to use the remaining 20GB most of the time.
    So finally the first 60GB won't wear at all, whereas the latter 60GB will get written to all the time, and the last 20GB will wear the fastest.

    And because SSDs are expensive people tend to not waste space on a SSD, so I think 20GB free is luxury.

    Or how do SSDs handle this? Reshuffle the content? But this causes additional writes and wear, too. So I think TLC SSDs lifespan is much much shorter than 12 years, maybe just the 2 or 3 years most people keep their computers nowadays.
  • twotwotwo - Friday, November 16, 2012 - link

    Yes, they reshuffle the content. The crazy thing about SSDs is that they look like hard drives do your OS, but they're doing totally different stuff underneath. Look up 'wear leveling'; here are a couple links:
  • UpSpin - Friday, November 16, 2012 - link

    Thanks, didn't knew about the 'static wear leveling' which, just as you said, also reshuffles the static content, thus makes my concerns invalid and TLC NAND a usable less expensive alternative to MLC. Reply
  • Impulses - Friday, November 16, 2012 - link

    This is why Anand talks about write amplification... Because of all the wear Keeling"leveling algorithms, and those do indeed cause more wear, it's a balancing act. Reply
  • Visual - Tuesday, November 20, 2012 - link

    No, write amplification has little to do with wear leveling. The most major cause is the fact that flash is organized in blocks that must always be written to entirely. You may want to write just a single byte, but the drive would internally have to read the whole block (say 4KB) and then erase and white the whole block with that one byte changed. Reply
  • bwat47 - Saturday, August 03, 2013 - link

    The SSD's firmware actually does "Reshuffle" the content, so you don't need to worry about the SSD trying to just wear-level the same 20gb, they are much more intelligent than that. Reply
  • David.Sucesso - Monday, August 19, 2013 - link

    UpSpin what you are talking is so true that really scares me...

    i have registered just to confirm that i only have an ssd drive 120gb ( and the space that is normally free is as 35GB to write every day. the other are files that doesnt get moved on

    now this is serious for the consumer
  • gevorg - Friday, November 16, 2012 - link

    looks like I have to find a sweet deal for Samsung 830 while its still available Reply
  • alextall - Friday, November 16, 2012 - link

    For anyone concerned about a short lifespan, I have put an SSD in all three of my machines now. My daily use can generate 10-50GiB of data. While that means I may have considerably shortened life on these drives (which is still 7-10 years according to the MLC data above), the increased productivity more than makes up for the difference. I don't even plan on keeping these computers half that long.

    Besides, with the current prices for SSDs are, and where they are (very quickly) moving, it really isn't a difficult value proposition to switch.
  • name99 - Friday, November 16, 2012 - link

    Don't waste your breath.

    The psychology here is not that of "let's have an honest discussion about usage patterns, cost, and change". The psychology is "I want to be a contrarian to show the world that I'm a deep thinker who doesn't follow the crowd".
    It's the same psychology that drives the non-ideological pundits (the one's who say stupid things attacking both parties) and, as you may have noticed, those people have not altered their shtick one bit, even after fifteen years of the internet shaming them with their atrocious record of insight.

    It's the same thing here. These people, if they were old enough, were very concerned that we were replacing floppy disks too soon. They felt that maybe we shouldn't be so quick to ditch VGA. They felt that sealed batteries were definitely going to cause trouble.
    If they knew anything about the technology, they'd be bleating that "sure, Haswell's lock elision looks good, but how can we be SURE that it won't occasionally miss a memory collision?"
    Whatever it is, they are sure that gut instinct trumps engineering skill and actual knowledge of the relevant statistics every time.

    The only way to deal with them is to avoid them (and on sites like Ars Technica) filter them out of the comment stream.
  • bebimbap - Friday, November 16, 2012 - link

    If you think about it, even if it were to last 5 years, it's quite long enough.
    in 2007 SSDs were $7.5/GB for the cheapest drives or about $1000 for 128gb worth of drive.
    in 2010 128gb drives were about $400?
    Today 5 years later 128GB 830s drives are as cheap as $70 on sale. that's $0.55/GB
    so 5 years from now you'll probably be going to buy a 2TB drive will cost you <$100
    As drive capacities increase so will their longevity due to wear leveling.
    When 2TB SSDs come out even 250-500 P/E cycles might be more than enough
  • jonjonjonj - Friday, November 16, 2012 - link

    im in the market for a 256GB ssd and wanted to get a 830. when the 840 was first release the price of 830's dropped. you could get a 256GB 830 for as low as $150. the problem is the 830's price went back up and now the 840 is now cheaper then the 830. not to mention the 830 is sold out at newegg. samsung shows a MSRP of $259.99 for the 840 on their website but you can get it $80-100 cheaper. do they know something we dont? the 840 pro is too expensive compared to the other drives. it just concerns me that the 840 is cheaper then the 830 and im not sure what to do. Reply
  • MrSpadge - Friday, November 16, 2012 - link

    It's using TLC NAND and a more advanced (smaller) process technology. This is done to bring the cost down, so when ever the yield in the new process is good enough, cost should naturally be lower than the elder brother. Reply
  • jonjonjonj - Friday, November 16, 2012 - link

    thats the whole problem. i would rather buy a MLC samsung/marvell drive for the same price as the TLC 840. i figured the retailers would be trying to get rid of their 830's but the prices are going up as the 840's price got slashed. you can get an 840 for $150 right now. Reply
  • Impulses - Friday, November 16, 2012 - link

    Then get an M4 or Plextor M5... I bought a second 830a few weeks ago because I knew the 840 was being positioned as it's direct price replacement and the 840 Pro would carry a premium for a while. Reply
  • MadMan007 - Friday, November 16, 2012 - link

    The lower price of the 840 is simply MSRP versus street price. Reply
  • Beenthere - Friday, November 16, 2012 - link

    With the compatibility and SSD reliability issues, NAND endurance is of little concern. Most folks will be replacing their SSD numerous times over the course of 3-5 years due to reliability issues. When consumer grade SSDs become as reliable as enterprise class SSDs, then NAND eundurance might be of concern. Consumer grade SSDs are disposable goods right now. Reply
  • fuzzymath10 - Friday, November 16, 2012 - link

    Many of us have been running first-gen X25-M drives for 3+ years, and will probably continue to do so for a few more, since 80GB is still sufficient for a system drive. I think the capacity will be the first restriction resulting in replacement; speed and reliability will be secondary issues unless you own an SSD for the sole purpose of benchmarking. Reply
  • Iketh - Saturday, November 17, 2012 - link

    you've been purchasing the wrong SSDs haven't you Reply
  • Death666Angel - Saturday, November 17, 2012 - link

    Still running my original OCZ Agility and OCZ Vertex 2 here without issues. Reply
  • iwod - Saturday, November 17, 2012 - link

    is there a software that records how many data i actually write per day or over a week?

    With a 16GB Memory system and No Virtual Memory. While Windows 7/8 are already doing its job trying to prefect as much data as possible i rarely sees my system memory being fully ulterised. So i am guessing i am doing very little write per day.
  • Kristian Vättö - Saturday, November 17, 2012 - link

    Many SSDs report Total Bytes Written (TBW) or Host Writes SMART value, of which both can be used to determine the amount of writes. I'm sure there are applications that record IO activity as well but I'm not aware of any good ones (the one we use is under very strict NDA). Reply
  • iwod - Saturday, November 17, 2012 - link

    I was wondering when i am reading 2bit allows 4 signals while 3 bits allows 8 signals, doesn't that means 3bits should actually double the capacity of 2bit cell? Reply
  • Kristian Vättö - Saturday, November 17, 2012 - link

    With MLC, you need four voltage states because there are four possible values (""00", "01", "10" and "11"). However, that is only two bits of data.

    TLC stores three bits but they can be organized in eight different way ("000", "001", "010", "100", "011", "110", "101" and "111"), hence you need eight voltage states.

    In other words, the increase in capacity is linear but the amount of voltage states (and thus the endurance) is exponential.
  • quixver - Saturday, November 17, 2012 - link

    I work with fairly large c++ code bases, and my compiler makes gratuitous use of the disk. With a traditional hard drive - a full rebuild of one of my projects took around 4 hours. With an SSD that time has come down drastically - under an hour in fact (the code base has grown since I switched to SSD - so the actual improvements are more than 4x ).

    So - if my SSD lasts me a year - it's more than paid for itself. And replacing the data is no biggie - all my data is either on dropbox, git/hg repos or a raided NAS.
  • spixel - Sunday, November 18, 2012 - link

    10gb a day?????

    I have only windows 7 installed on my ssd, all games get installed to my caviar green. I have 20gb per 24 hours according to ssd life, and that's just windows doing its thing while the pc is on.
  • haukionkannel - Sunday, November 18, 2012 - link

    Well, the normal user use his machine 2-3 hours in a day... max, so if you use your machine 24 hours a day, 20 gb seems to be guite low usage!
    (1-2 hours seems like good estimation of "normal" user usage in a day ...)
  • spixel - Sunday, November 18, 2012 - link

    I don't have my pc on 24 hours a day, I have a total of 10 days worktime and 220gb, which seems on the low end looking at other pictures of ssd life.

    I always wondered though, is its lifetime based on total worktime hours, or an estimate of say 3hr per day.
  • hurleydood - Thursday, March 28, 2013 - link

    The worst part is the host writes are 20GB per day but the actual NAND writes it could be close to 3.5 TB a day. My drive has 11GB a day writes or 3.2TB of host writes over 7105 hours on (296 days 24/7). By my calculations I have a write amplification of 175x and burned 4498 average P/E cycles, 11% drive life. That would be 2TB per day of NAND writes for 11GB of random host writes to the drive. 4498 average P/E cycles multiplied by 128GB drive size should give me 562TB total NAND writes. Divide that by my 3.2TB host writes and you get 175x write amplification. Reply
  • hurleydood - Thursday, March 28, 2013 - link

    Here's the pic from CrystalDiskInfo Reply
  • Laststop311 - Monday, November 19, 2012 - link

    I knew people have been over acting the nand endurance issue. SSD's should last longer then hdd's with all their moving parts something is bound to fail on a hdd before you run out of writes on a ssd. In a week I probably use about 50GB of writes and thats a high estimate, more like 30GB a week unless a buncha good movies release that week. So my 256gb crucial c300 which is a 34nm MLC product should basically last me a lifetime. Reply
  • Laststop311 - Monday, November 19, 2012 - link

    Can anyone tell me how much longer the 34nm MLC nand should last compared to the 20nm MLC evaluated in this article? Reply
  • Kristian Vättö - Monday, November 19, 2012 - link

    IMFT's 34nm MLC NAND is rated at 5,000 P/E cycles, whereas IMFT's 20nm MLC NAND is rated at 3,000 cycles. Using the same math as in the article, a 256GiB SSD with 34nm MLC NAND would last for 116.8 years if you write 10GiB a day (WA=3x).

    Obviously, that's theoretical.
  • Ao1 - Monday, November 19, 2012 - link

    The theoretical P/E count is not just based on how many times the cell can be written too; it’s also based on the ability to retain data over time. Whilst it’s possible to write well beyond the theoretical P/E count no-one (publically) knows how long data retention will last once the theoretical P/E count has been exceeded. The full picture on endurance is not available until someone can show the relationship between the P/E count and the duration that a cell is subsequently able to retain data. Reply
  • htwingnut - Sunday, February 24, 2013 - link

    Any user that is writes a lot of data shouldn't even consider one of these drives. Just buy the 840 Pro or a Crucial M4. I see the 840 as more of a casual user's SSD. I put a 120GB 840 in each of my kids' netbooks. Over the course of a year they'll likely average less than 10GB writes a day and their systems go on hibernate once or twice a day. Most users that just browse the web, do basic MS Office stuff probably won't even encroach 10GB/day, and remember that's 10GB/day 7 days a week, 365 days a year.. Most users buy a new computer or laptop every 3-4 years anyhow, so I don't see it as a big deal. I wouldn't trust any HDD that is over 4-5 years old either. I have an Intel X25-m 80GB that's been used as a "casual" OS drive for about 3 years now, and it has less than 1TB of writes on the drive. It basically housed the OS and a few apps. Reply
  • Kaos11 - Tuesday, April 30, 2013 - link

    I would like to partition my 840 into two partitions as this is my normal setup on a Spinner.
    The question is does wear levelling treat the partitions as separate or is it smart enough to still wear level the whole SSD?
  • phil13 - Wednesday, December 11, 2013 - link

    I have one that failed, only show 0.9GB, SN: 000000000, Firmware: DXT06BO0, even samsung Magician software can see it, but say that there is no samsung SSD drive.
    It does look the firmware problem, all feel ok.
    Any solutions and help? appreciate, thanks.

Log in

Don't have an account? Sign up now