Should I wipe my X-25 160GB SSD?

[Hazarrd]

Here are my results from Crystal: http://dl.dropbox.com/u/4918782/Crystal.JPG

Correct me if I'm wrong, but it appears that my read speeds are right on target, yet the write speeds are at less than half of what they should be. The drive is at about 80 percent capacity and I've used it for just over a year now. Will wiping it fix the problem? Does anyone have a good link to step-by-step for wiping?

Thanks!

 

[razel]

I wouldn't worry about it. But if it really bothers you, run the Intel Toolbox. If you have a G2 you can execute it's garbage collection (GC) that uses TRIM data. If you have a G1 run the full diagnostic. It will fill up the empty space with test data, which will force the drive to do GC.

You can also use HDDerase, but for most getting it to just run is a chore and that's on top of having to backup/restore. I only recommend doing a secure erase only if you plan on selling the drive.

Have fun!

 

[flamenko]

There can actually be two reasons why your ssd is lagging abit, the first being the fact that you are starting to near drive capacity of course. Once u fill that drive past the 60% mark, you will see performance results to lag.

Another is the fact that you may have System Restore turned on. If you do, the way Win sets its allocation points has been known to affect if not completely disable TRIM which results in slowing of the drive.

Here is an article which speaks of this. Hope this helps.

 

[Hazarrd]

Interesting. I think I have the G1 because the Toolbox wouldn't run the optimization tool. It turns out I've had System Restore on. I deleted the files and disabled it. Would this have had any effect on a G1's performance, or only a G2?

 

[flamenko]

Run a CDM and post the score here. If your restore is now shut off and the restore points have been deleted, performance will return to its original level if this is what the problem was.

 

[Hazarrd]

WOW. After running the Full Diagnostic using the Intel Toolbox and then disabling System Restore (and removing all data) all write-speed benchmarks improved significantly (close to double the speed for most of the tests).

Original CDM results: http://dl.dropbox.com/u/4918782/Crystal.JPG
New CDM results: http://dl.dropbox.com/u/4918782/Crystal2.JPG

Now I wonder whether I'll actually notice a speed increase. I'm also curious as to whether the test that CDM uses may conflict in some way with System Restore, such that it does not accurately replicate traditional use of the SDD.

Can someone tell me whether the degradation problems only affect write speeds, or do they also affect read speeds?

 

[flamenko]

Degradation can affect both read and write. Its not really a permanent degradation as, as you can see, it can be fixed.

CDM is fine on the Intel. I recognized this problem many months back and approached Intel in their forums and outside with respect to validating it. We went back and forth and back and forth and when it got to the actual test results, their engineers chose not to comment any further. Even through our phone conversations, the standard "We are looking into it" resulted and at the end of the day they did state that System Restore should be disabled in their New User SSD Guide. Things pretty much caught on from there.

Myself, I believe this this is the reasoning that so many were hollaring about SSD degradation in the past few years as I had not experienced same on several older ssds (slc and mlc) but I haven't used system restore for 3 years as well.

 

[Tsavo]

SSD's still degrade as they fill?

 

[flamenko]

Yes the Intel does as does many others. I have tested the OWC drives and they do not. I have requested many to conduct similar tests of the OCZ's to which I have yet to have anyone respond although they said they would return with the results. Its a very simple test actually whhich concerns me as to why nobody has replied saying the OCZ does not slow when filled.

 

[Hazarrd]

Does anyone on here think that using HDDErase will result in any significant improvement of performance on my SSD, given the results I posted above?

 

[bamacre]

SSD's still degrade as they fill?

I would think this would be a major problem given that SSD's have such small capacities.

 

[flamenko]

Notice how full this OWC drive is...

 

[sub.mesa]

Is that a Sandforce-based SSD? If so, it has 28 percent spare space by default. Thus 100GB SSDs are 128GiB physical NAND. That is alot more spare space than Intel gets by default (only 6.8 percent).

edit: percentage signs are broken on this forum; or at least appear to be?

 

[flamenko]

Actually, that is the 120Gb model with the same provisioning as the Intel. This is the RE model with 28 percent overprovisioning, vice the norm of 7 percent;...oh and again full.

If it will help out here, I have been asking from day one for the same full test from the OCZ where many had originally said ya I will do it and get back to ya. I have yet to have anyone get back to me.

Also, OWC warrants against exactly performance degradation through capacity or age of the ssd within the warranty period.

 

[tweakboy]

Dont use Windex, if you want to wipe it take a large tissue and clean it that way.

 

[Tsavo]

Is that a Sandforce-based SSD? If so, it has 28 percent spare space by default. Thus 100GB SSDs are 128GiB physical NAND. That is alot more spare space than Intel gets by default (only 6.8 percent).

edit: percentage signs are broken on this forum; or at least appear to be?

Percents are broken for me, too.

 

[pcslookout]

I would think this would be a major problem given that SSD's have such small capacities.

In my opinion it is. They should sell SSDs where there is extra space on them that you can not use so this does not happen. Doubt that will ever happen to benefit the consumer but who knows. Like for example a 80 GB SSD with really 100 GB or 110 GB of space. That way if it gets full or close like 70 GB max then it won't slow down much. Same with the 160 GB with like 180 to 200 GB instead.

 

[sub.mesa]

Actually, that is the 120Gb model with the same provisioning as the Intel. This is the RE model with 28 percent overprovisioning, vice the norm of 7 percent;...oh and again full.

Full by writing large files yes, that won't hurt as much as small writes which cause fragmentation on erase blocks, especially on Intel SSDs.

Could you also try benchmarks which do not zero-write? Testing zero-write on Sandforce-based SSDs would give artificially high scores due to compression going skyroof. I'm not certain on Windows, but i believe some benchmarks can also use random patterns which are not compressible.

Also, you cannot reproduce SSD performance degradation by just writing it full. Likely the physical NAND won't be nearly full if all you do is write zeroes. In theory only one NAND page could be allocated without any other physical writes. Though i do not believe the compression algorithm is that smart.

Also, testing with a 50MB sample is much too low. The filesystem may be nearly full, but the physical NAND is not!

That said, the compression could cause for more spare space, simply because the filesystem has 100GB allocated but actually only 60GB is being stored (40GB compressed) then that also means 40GB is available as spare space. This trick of course doesn't work if you only write incompressible data, especially over a period of time with many modifications, such as Windows updates + System Restore.

Also, OWC warrants against exactly performance degradation through capacity or age of the ssd within the warranty period.

Any URL of this? That would mean if there is any degradation you can send your used product back? I can't imagine they would be so dumb, since that means anyone can send their used SSDs back to the manufacturer. Only compactflash/USB sticks don't degrade; they are just always slow on small writes.

 

[flamenko]

I understand what you are asking for but CDM and AS SSD do not give accurate results of SandForce drives. With respect to 0Fill, it is the same result as ATTO which is used by the manufacturer to determine speeds for the SandForce drives.

Also, the filling of the drive, if it helps, contained small and large files as I simply copied everything I had. doubled, tripled and moved it all as a large folder to fill. There are probably much more small files than large in the folder used,

I would be a bit confused when you say that you cannot reproduce performance degradation by simply writing it full as it becomes apparent in drives as soon as one starts filling it past a certain point.

In the end, I am still waiting for another SandForce drive, specifically the OCZ, to be posted with the same test and full, whether it be 0Fill or not.