A huge thanks goes out to AnandTech reader Andrei and some very attentive Xtreme Systems forum members. I just got an email pointing me to this thread where one particular number from our OCZ Vertex 3 and Intel SSD 510 articles was called into question. The problem? The 4KB random read numbers for the Vertex 3 were supiciously high. The reality? They were incorrect. 

I was just alerted to the error and quickly powered up the SSD testbed to recreate the test. It looks like the original numbers were either run at a queue depth of 32 or accidentally copied from one of the runs of 4KB random write tests. Either way the number was incorrect and has been fixed in all affected articles.

The updated numbers don't change our conclusions. The Vertex 3 is still the fastest next-generation SSD we've tested thus far and it still maintains a random read performance advantage over the Intel SSD 510.

The integrity of our test data is something we all take very seriously here. Errors like these do you a disservice and hurt the reputation I've worked so hard over the past 14 years to build. I do hope this oversight hasn't negatively impacted your opinion of AnandTech - we aren't perfect, but we strive to be. I do apologize to all of you for the error and I will be restructuring how I run and record my Iometer tests to avoid this particular issue from cropping up again. 

My sincere thanks goes out to Andrei and the XS folks who helped track down the error and inform us of its existence.

POST A COMMENT

49 Comments

View All Comments

  • Anand Lal Shimpi - Thursday, March 3, 2011 - link

    I just wanted to chime in and express my sincere gratitude to all of you. It's extremely embarrassing to have something like this happen and your outpouring of support always makes it much, much easier to deal with. All of us at AT wait with bated breath whenever we post an article - we're always so nervous to see how you all will respond. Our aim is to always do right by you, so I do appreciate that you guys take the time to let us know when you feel like we're doing a good job (and when you feel like we aren't).

    One of the first things I tell new folks who join the AT family is that you have to check your ego at the door. We are serve the readership and this idea of writers putting themselves above and beyond the readers is ridiculous. We're all on the same team here and there's no room for ego on said team :-P

    Thanks again for your support and for reading the site :)

    Take care,
    Anand
    Reply
  • cynic783 - Thursday, March 3, 2011 - link

    Anand is excellent and here's why: Making a small error is inevitable. To correct the error requires acknowledging the error. That's a BIG deal in my life experience. If you have the humility to acknowledge an error then you are still open to learning, and if you are still open to learning, then you are still improving. Anand, good a site as it is, is still improving!

    Bravo Anand for acknowledging and fixing errors like this.

    I have an Intel G2 160 and X25-32G and have not decided what my next drives will be yet. Leaning towards Vertex 3 but the reliability question is, well, "it's out there".
    Reply
  • neotiger - Thursday, March 3, 2011 - link

    This is very disappointing.

    Any reason why SandForce 2000 is so slow (or why C300 is so fast)?
    Reply
  • TheExodu5 - Thursday, March 3, 2011 - link

    Thanks for fixing that up! It's nice to see true integrity is still around these days.

    I have a question, if you will. Why was QD = 3 chosen for the test? I have seen other reviews for the Vertex 3 and it actually performs quite poorly at QD = 1...almost 50% slower than the Intel drive, at around 13MB/s. Is QD = 3 simply more common in a real world scenario?
    Reply
  • dhanson8652 - Thursday, March 3, 2011 - link

    I'd say it's common. More common? I don't know but, common enough to be valid, Yes.

    QD=1 assumes nothing at all is happening on your PC when you make a request.

    QD=3 is more likely with all the background updaters (AV, quicktime, adobe, firefox, stream, etcetera) plus any background activity like Windows Search indexing, Torrents, AV scans, and such.
    Reply
  • iwod - Saturday, March 5, 2011 - link

    Interesting question. Majority ( I remember somewhere up to 95% ) of the trace inside Anand benchamarks are Queue 1 - 2.

    P.S - Which site was that? The only thing i found was one thread. Which is different to Q1.
    Reply
  • Conficio - Thursday, March 3, 2011 - link

    Anand,
    thanks for the correction.

    I admire your courage to stand up for your errors in public and with full responsibility. It is just human to make errors, the difference we can make is hwo we deal with them. You just showed us the best way to deal with any of them.

    Thanks

    Kaj
    Reply
  • B - Thursday, March 3, 2011 - link

    +1 Reply
  • Aigoo - Thursday, March 3, 2011 - link

    <3 you anand.

    Nice to see you acknowledging and quickly fixing any errors that arise. Stuff happens, so as long as it gets fixed then all is well!
    Reply
  • slickr - Thursday, March 3, 2011 - link

    First with Nvidia than with Intel and now with OCZ!

    This website just burrows itself deeper and deeper into the ground. People can not trust this website anymore, nor should they!

    This kind of "oversights" or "accidents" are becoming too frequent for them to be just that. I'm sure its actually something more like being on the payroll of these companies and making their results look better than realistically.
    Reply

Log in

Don't have an account? Sign up now