Inconsistent disk mark with Passmark 7.0?!

[krsboss]

Hi, I think the best way I can illustrate what I mean is by posting a couple of screenshots:





...sooooo...the passmark tests show a massive disparity between read & write rates, whilst the ATTO benchmark shows that they are both pretty similar.

I was just wondering if anyone had any feedback on this, and if there was anything I could do to get passmark to register the correct speed.

Naturally I understand that the passmark test & atto benchmark are two different tests and they would not necessarily produce the same numbers, however it is the massive difference between the read & write speeds (600 MBps) that I don't understand. I would not questions the read rate of passmark showing ~900 MBps, whilst atto showing ~1000 MBps if the write rate of ~250 MBps shown by passmark was closer to the ~950 MBps of atto!!!

Anyway, thanks for taking your time to read this post,

cheers, krs

[David (PassMark)]

The ATTO write results were between 18MB/sec and 989MB/sec.
And would no doubt vary even more if you change the IO settings and Queue depth.

Given the range was 18MB/sec to 989MB/sec, I don't see how you could conclude that 255MB/sec was a either incorrect, or a "massive difference".

To me it looks like the drive's performance is very sensitive to the test conditions.

You should be able to get similar results from PerformanceTest using the advance disk test (under the advance menu) and fiddling the parameters to match ATTO.

[krsboss]

Hi! Thanks for getting back to me.

The reason that I think that this result is inconsistent or incorrect is from comparing benchmarks from my computer & from downloads that are available; see below:



Uploaded with ImageShack.us

This shows my computer (Monolith v5) with a single Vertex 3 SSD, compared against another persons benchmark which also has a Vertex 3. The read speeds are practically identical, whilst the write speeds on his are vastly superior, more in line with what I would have expected from the atto benchmarks.

As I now have 2 x Vertex 3 in RAID0, my read / write speeds are practically what they were before..which would be expected...

...I am not trying to replicate the atto benchmark, merely curious as to why the results I'm getting are so different to firstly what I would have expected* and secondly results from the same hardware that I can download.

*on previous incarnations of my system I was getting similar read / write speeds.

Cheers, krs

[wonderwrench]

First you are using a SandForce based drive that uses data compression to speed up writing and reduce write amplification. ATTO uses compressible data and PT does not so I would guess that's where the discrepancy in write performance is coming from . How much bench testing have you done on your drive? Many bench programs write gigs of data every time they are run. SandForce based drives will throttle if hammered hard enough. This problem is claimed to be partially rectified in the Vertex 3 but from what I can see its still a problem. Best way to see if the drive is throttling and has slowed write performance would be to secure erase the drive followed by a clean windows install. Never full format the drive or use third party tools to wipe or format the drive. Never clone to the drive either. Any of the above may cause throttling and poor write performance.

I use Acronis TrueImage to backup and restore my Vertex 3 MI, normal images only. No sector by sector backups or cloning.

Bill