Random Read/Write Speed

The four corners of SSD performance are as follows: random read, random write, sequential read and sequential write speed. Random accesses are generally small in size, while sequential accesses tend to be larger and thus we have the four Iometer tests we use in all of our reviews.

Our first test writes 4KB in a completely random pattern over an 8GB space of the drive to simulate the sort of random access that you'd see on an OS drive (even this is more stressful than a normal desktop user would see). I perform three concurrent IOs and run the test for 3 minutes. The results reported are in average MB/s over the entire time. We use both standard pseudo randomly generated data for each write as well as fully random data to show you both the maximum and minimum performance offered by SandForce based drives in these tests. The average performance of SF drives will likely be somewhere in between the two values for each drive you see in the graphs. For an understanding of why this matters, read our original SandForce article.

Iometer - 4KB Random Write, 8GB LBA Space, QD=3

Many of you have asked for random write performance at higher queue depths. What I have below is our 4KB random write test performed at a queue depth of 32 instead of 3. While the vast majority of desktop usage models experience queue depths of 0 - 5, higher depths are possible in heavy I/O (and multi-user) workloads:

Iometer - 4KB Random Write, 8GB LBA Space, QD=32

Iometer - 4KB Random Read, QD=3

Faster than a Vertex 3 Pro & The Test Sequential Read/Write Speed
POST A COMMENT

85 Comments

View All Comments

  • Mr Perfect - Thursday, February 24, 2011 - link

    I hate to bring this up here, but this should be addressed and you seem to have a direct line to the OCZ CEO(all be it in sticky notes). Can AT look into the fiasco where OCZ shipped 8 channel 25nm Vertex 2 drives bearing the exact same name and model number as the origonal 16 channel 34nm drives, and then when customers noticed the difference decided to charge people an upgrade fee to get a 16 channel drive? True, OCX later relented after a negative backlash and gave people free exchanges for a 16 channel drive, but I can't think of a faster way to burn through customer good will then a bait and switch stunt like that. What happened?

    http://www.storagereview.com/ssds_shifting_25nm_na...
    Reply
  • FunBunny2 - Thursday, February 24, 2011 - link

    Yeah. AT *seems* to be easier on OCZ than storagereview. I can't speak for their quality overall as I've only read up on what they had to say about this OCZ Vertex nonsense. I'm not going to abandon AT. Yet, anyway. Reply
  • semo - Thursday, February 24, 2011 - link

    It wasn't always the case. When Anand exposed jmicron's awfulness to the world, OCZ's CEO wasn't happy at all (the surprising thing is Anand actually mentioning this back then).

    The two things that surprise me the most are:
    1. Why didn't Anandtech as a whole report on 25nm drives. AFAIK OCZ were the first with commercial 25nm drives. Now if a commercial x86 CPU was about to be released in the next 6 months, AT would be all over that story...

    2. A lot of time has passed, and there is still no coverage on OCZ's dishonest practices. If a decent site (like AT) covered these drives early enough, a lot less people would have fallen in OCZ's trap.

    Anand did mention in his previous article that he would look in to the issue.

    I've seen the discussion and based on what I've seen it sounds like very poor decision making on OCZ's behalf. Unfortunately my 25nm drive didn't arrive before I left for MWC. I hope to have it by the time I get back next week and I'll run through the gamut of tests, updating as necessary. I also plan on speaking with OCZ about this. Let me get back to the office and I'll begin working on it

    As far as old Vertex 2 numbers go, I didn't actually use a Vertex 2 here (I don't believe any older numbers snuck in here). The Corsair Force F120 is the SF-1200 representative of choice in this test.

    Take care,
    Anand
    http://www.anandtech.com/show/4159/ocz-vertex-3-pr...
    Reply
  • semo - Thursday, February 24, 2011 - link

    "Now if a commercial x86 CPU" should have been "Now if a commercial 14nm x86 CPU". Usually new tech in the CPU, GPU and now SoC segments gets covered quite early and thoroughly. Somehow the 25nm Vertex 2 flew under the radar it seems... Reply
  • Anand Lal Shimpi - Thursday, February 24, 2011 - link

    I've been working with OCZ behind the scenes on this. I've been tied up with the reviews you've seen this week (as well as some stuff coming next week) and haven't been able to snag a few 25nm drives for benchmarking. Needless to say I will make sure that the situation is rectified. I've already been speaking with OCZ's CEO on it for the past week :)

    Take care,
    Anand
    Reply
  • lyeoh - Thursday, February 24, 2011 - link

    Hi Anand,

    Is it possible to do average and max latency measurements on the drive access (read/write) while under various loads (sequential, random, low queue depth, high queue depth)?

    I'm thinking that for desktop use once a drive gets really fast, the max latency (and how often it occurs) would affect the user experience more.

    A drive with 0.01 millisecond access but 500ms spikes every 5 secs under load might provide a worse user experience than a drive with a constant 0.1 millisecond access even though the former averages at about 80000 accesses per second while the latter only achieves 10000. Of course it does depend on what is delayed for 500ms. If it's just a bulk sequential transfer it might not matter, but if it delays the opening of a directory or small file it might matter.

    This might also be important for some server use.
    Reply
  • jimhsu - Thursday, February 24, 2011 - link

    This is needed -- however there are a lot of technical problems when trying to assess max latency. One of those is reproducibility -- if the drive has a latency of 0.01ms, but a max of 250ms, how reliable is that data point? What if something just happens to be writing to the drive while you make that measurement? (This can easily be seen by trying to do 4K reads when doing a sequential write larger than the DRAM cache of the SSD, such as the Intel G2 drives). From my personal observations, website to website reproducibility of maximum latencies, as well as minimum frame rates in reviews, is extremely poor.

    Which is unfortunate, because they can impact user experience so heavily. Aside from strict laboratory-controlled conditions and testing with specialized equipment, I find it hard to conjecture how to do this.
    Reply
  • semo - Thursday, February 24, 2011 - link

    "haven't been able to snag a few 25nm drives"
    So OCZ are sending you Vertex drives for testing so early that they don't even have housing and you're stuggling to get a hold of drives that have been sold to consumers for the past month or two?

    How come you have V3 drives months before their release date and AT does not yet even have a news article on the 25nm V2s?

    I'm sorry Anand but this will be the 1st SSD article you've written that I will not be reading. This is in protest to OCZ's handling of this CONSUMER RIGHTS issue. I'm not protesting against AT but OCZ's propaganda which unfortunately is channeled through my No 1 favourite site.

    Ryan Petersen, you know where to stick it.
    Reply
  • seapeople - Thursday, February 24, 2011 - link

    Yeah, I can't imagine why OCZ isn't knocking Anand's door down with review samples of 25nm SSD drives after that big customer snafu with their 25nm drives. I mean, they should be eager to have all the publicity on that issue as possible, right?

    And then why on earth would they rather send their newest, fastest SSD drives to a review site right now? They must know that Anand is probably busy reviewing the other SSD drives that are about to come out, so why would they bother him with their new stuff?
    Reply
  • beginner99 - Thursday, February 24, 2011 - link

    That's why I would just go with Intel. I know what you pay for. No custom firmwares and other funky stuff. Reply

Log in

Don't have an account? Sign up now