Intel Unable to Replicate SSD Speed Degradation Issue

[Rob Williams]

From our front-page news:

Last Monday, we linked to an article posted by PC Perspective that took a look at the loss of performance on SSDs over time, with the target of focus being Intel's X25-M. The article went on to explain that during certain usage scenarios, the sub-block level can become highly fragmented and cause extreme slow-down over time. The worst of it is that typical defragging applications, such as Diskeeper, are unable to work on the sub-block level, and as a result, attempting to defrag the drives will make the issues only worse.

Since that article was posted, it's been getting a great deal of attention around the web, and as you might expect, Intel has been paying close attention to the issue. In a response to an editor at CNET, Intel notes that they've been unable to replicate the issue thus far, but they are of course still investigating the issue. Intel is also quoted as saying, "In our estimation, the synthetic workloads they use to stress the drive are not reflective of real world use."

I have no doubts that last point is true, but as mentioned in the originating article, the idea to investigate the issue began when the writer noticed performance degradation after a few months of regular desktop use. Intel also goes on to mention that it's typical to experience slowdown when the drive is full, and on an 80GB model, that's not hard to pull off. Whether or not that factor played a role in this, we're unsure, but I'm confident we'll be hearing from either Intel or PC Perspective with updates soon.

​

In response, Intel made a statement on Thursday. "Our labs currently have not been able to duplicate these results," Intel said. "In our estimation, the synthetic workloads they use to stress the drive are not reflective of real world use. Similarly, the benchmarks they used to evaluate performance do not represent what a PC user experiences."

<table border="0" width="100%"><tbody><tr><td>Source: Nanotech: The Circuits Blog</td> <td>
</td></tr></tbody></table>​

 

[Kougar]

"In our estimation, the synthetic workloads they use to stress the drive are not reflective of real world use."

This was specifically mentioned in PC Perspective's article as well. However, it doesn't change the fact many of the enthusiasts buying these drives are going to be running quite a few synthetic benchmarking or stability testing programs. Or that they are going to be running workloads more typical of a server, either of which PCP states can lead to this problem.

It is interesting that Intel has in fact not denied the problem, only that they have not observed it. And PC Perspective wasn't the only site noticing this kind of behavior.

PC Perspective's mentioning of the new ATA commands tailored for SSDs was the first and only time I've herad of this. Anyone have more info or even a possible date the ATA spec will be updated?

 

[Rob Williams]

Intel is just being careful where this is concerned. They can't come out and claim that there is an issue unless they understand it perfectly. It no doubt exists, but they are right in that it happens with specific workloads. You're also right in that enthusiasts will be running these workloads for the sake of performance testing. I guess the moral of the story is, if you buy one, you should run your benchmarks and then format/install your OS on it.

As for the ATA spec, no idea, but I wish I did as well. I really can't see it happening soon, sadly. I contacted SATA-IO, but I'm not even sure if they are the right people to ask, to be honest.

 

[Kougar]

Yeah, no doubt Intel is being careful. If they acknowledge the issue, they lose sales and if they deny it then they lose credibility if the problem is proven. It's a tricky situation.

I believe SATA-IO is the correct organization, I didn't know there were quite so many parties involved in the SATA specs! The new SATA 6gb/s spec is supposed to add some new commands as well from what I'm seeing, but I can't confirm it is the same commands PCP specifically mentioned. It would be great if this was the case as the new 6gb/s spec is due to go official this year.

 

[Rob Williams]

It would be great if this was the case as the new 6gb/s spec is due to go official this year.

I was thinking the exact same thing. It would be nice, but seems entirely unlikely. Who knows though. Given that SSDs are growing in popularity fast, and will be much moreso in the next few years, hopefully they'll decide to revise the upcoming spec for that purpose. It seems highly unlikely, but more unbelievable things have happened, so we'll see.