Very slow write speed on SSD S3520

[DRobi14]

I have experienced horribly slow write performance on both my S3520 240GB drives on Windows 7, and would like some guidance.

The details:

I have 4 Intel SSDs : 2 x 530 Series (SSDSC2BW120A4 and SSDSC2BW240A4), and 2 x S3520 240GB (SSDSC2BB240G7). I have the same very disparate performance with both pairs. In addition, I have 3rd S3520 in the field with the same problem.

The pair I will highlight is as follows:

530 Series 120GB

F/W version : DC32

S/N: PHD A41 660 0Z8 120 7GN

DC S3520 240GB

F/W version : N2011021

S/N: BTD V72 750 FRY 240 AGN

The problem: I have a proprietary OS running on a Pentium server box, which has its own driver to talk to these SSDs in IDE mode. The 530 performs very well, in line with benchmark data I can find on the Web. The S3520 performs well also, except In certain cases, where its average sector write time is 26x (yes, 26) that of the 530. We have managed to produce a specific test case where we write 15,634 single sectors of a 13MB file, with specific content in each sector, and the same non-consecutive order of disk addresses being written. The average write time for sector writes to the 520 is 0.06ms, and the average write time for the S3520 is 1.56ms.

At first we thought there must be some compatibility problem between our disk driver and this specific drive.

But I have constructed a Windows program which writes these same sectors to a second pair of drives, in the same order, on Windows 7 (using FILE_FLAG_NO_BUFFERING).

The result is the same: < 1 sec to write the 15,634 sectors on the 530, and > 20 sec on the S3520.

I can only conclude that the S3520 has some severe firmware problem which makes it sensitive to the content (and/or the order) of certain data being written.

Does anyone have any experience with similar problems here?

[idata]

Hello DLRobinson,

I was reviewing this post with a higher level technician and we're trying to reproduce the issue as per your indications.Just for testing purposes, would it be possible to run a similar test on a different and newer operating system intended for server applications such as Windows* Server 2012 R2 or Windows* Server 2016 while we work on this?Best regards,Eugenio F.

[DRobi14]

Eugenio,

Sorry, I don't have any active non-virtual Windows Server installations, and none at all of the 2012 or 2016 variety.

Also, I don't see how it would help you for me to run this test in that environment - the discrepancy between the 530 and the S3520 is plain to see in Windows 7.

Furthermore, the problem is not motherboard or CPU or OS dependent, since I can reproduce this behavior on 2 very different PCs with 2 very different OSes,

regards

Dave

[idata]

Hello DLRobinson,

Thanks for you reply. We understand your concern, however we just want to discard the possibility of the slowdowns being caused by the SATA driver on those operating systems.Since server versions are not option, could you run the test on a newer operating system such as Windows* 10?Best regards,Eugenio F.

[DRobi14]

Engenio,

Sorry, I can't run the test program on Windows 10 either - I do have a Windows 10 laptop but no way to install the SSDs in it.

But to your point about perhaps the slowdowns are caused by the disk driver - I can already demonstrate the problem on TWO different OSes (mine and Windows 7) with COMPLETELY different disk drivers (one written by Pebble Ridge Systems and one by Microsoft). It is unlikely in the extreme that both those disk drivers (A) work perfectly with the 530 series and (B) are responsible for the S3520 problem.

You should pursue the test case on Windows 7 with your internal firmware diagnostics, to find the cause of the problem.

regards

Dave

[idata]

Hello DLRobinson,

We didn't have any luck reproducing the test, it was showing expected write/read speeds on Windows* 7.

As mentioned before, our engineers are requesting would really like to confirm if the issue happens on other operating system to confirm if the issue is related to the drive.

Please let us know if there's anything else we can do for you.

Best regards,

Eugenio F.