Why is my SSD performing so poorly?

[996GT2]

Just ran the AS Benchmark on my Thinkpad T61 equipped with a SuperTalent Masterdrive SX 64 GB SSD.

The SSD itself is a Samsung unit with a Samsung controller (not JMicron)...I have read that the Masterdrive SX is the same drive as the OCZ Summit.

I ran the test with AHCI enabled in BIOS after doing the proper registry edits to enable it in Windows 7.

So, based on the good reviews for the Summit, why are my results so bad?
Link

More information on my drive:
Link

 

[aceO07]

The Thinkpad 60/61 series (X or T) is crippled. It only does SATA1 speeds. Unless you want to use the drive in another system, you may want to return it and get something cheaper.

http://forum.thinkpads.com/viewtopic.php?f=29&t=73634

 

[996GT2]

But SATA I/II shouldn't affect how my random 4k write speeds are, right? I don't care if the sequential reads/writes are below 150 MB/s, but it's the random write speeds that are worrying me...

 

[alcoholbob]

No. Samsung came out with a new controller for their 256GB MLC SSD which is featured across the Summit line. The 64 and 128GB OEM Samsung drives in circulation use an older RBX controller. It's still very competent but certainly is slower than the current gen controllers.

 

[deputc26]

The masterdrive sx series uses the jmicron controller which has known and terrible random 4k write performance. Avoid all Jmicron based drives.

 

[996GT2]

Originally posted by: deputc26
The masterdrive sx series uses the jmicron controller which has known and terrible random 4k write performance. Avoid all Jmicron based drives.

What's your source for this?

According to PC Per and many other sources it uses the same Samsung controller in the OCZ Summit series. See my second link in the original post.

 

[alcoholbob]

If it was really a jmicron drive the random write performance would be closer to 0.5mb/s. There's no way that's a jmicron controller.

 

[996GT2]

Any idea why it's performing nowhere near the standard of a X-25M then in terms of the random writes? I don't think those are affected much by Lenovo capping the SATA transfer speed to 150 MB/s, right?

 

[jkresh]

from the 2nd link you have it appears that the 64gb version of your drive uses the 1st gen samsung controller (which is not bad (much better then the Jmicron) but not as good as the 2nd gen which is in the summit and the 128/256 of your drive)

 

[996GT2]

Originally posted by: jkresh
from the 2nd link you have it appears that the 64gb version of your drive uses the 1st gen samsung controller (which is not bad (much better then the Jmicron) but not as good as the 2nd gen which is in the summit and the 128/256 of your drive)

So are the results pretty normal? And...how much improvement would I get by using a full SATA 3.0 GB/s bus?

 

[soccerballtux]

Originally posted by: 996GT2

Originally posted by: deputc26
The masterdrive sx series uses the jmicron controller which has known and terrible random 4k write performance. Avoid all Jmicron based drives.

What's your source for this?

According to PC Per and many other sources it uses the same Samsung controller in the OCZ Summit series. See my second link in the original post.

Just google for it, it's common knowledge. Even the reviewers on newegg point it out here and there.

 

[soccerballtux]

Originally posted by: 996GT2
Any idea why it's performing nowhere near the standard of a X-25M then in terms of the random writes? I don't think those are affected much by Lenovo capping the SATA transfer speed to 150 MB/s, right?

Internal raid and internal write caching.

 

[alcoholbob]

what do you need all the random writes for? 2/3 of all pagefile random writes are 4k and under.

 

[Emulex]

x25-m is just superior in REAL world random writes compared to (anything in that price range). you just keep believing those posted benchmarks..

throw the drive behind a decent bbwc raid controller with 256 or 512meg you'd probably see nicer performance if you set ratio to 0% read/100% write; (Cerc or p400)

 

[glugglug]

Originally posted by: Astrallite
what do you need all the random writes for? 2/3 of all pagefile random writes are 4k and under.

Um, no pagefile operation is ever under 4K as this is the minimum page size of x86.

pagefile writes are generally multiples of 256KB, which Win7 nicely aligns with SSD blocks. When memory is needed for something else, 256KB, 512KB, 1MB, and 2MB chunks of pages all get swapped out at once.

Pagefile reads are generally only 4KB, in whatever order your application accesses the pages.

 

[alcoholbob]

Originally posted by: glugglug

Originally posted by: Astrallite
what do you need all the random writes for? 2/3 of all pagefile random writes are 4k and under.

Um, no pagefile operation is ever under 4K as this is the minimum page size of x86.

pagefile writes are generally multiples of 256KB, which Win7 nicely aligns with SSD blocks. When memory is needed for something else, 256KB, 512KB, 1MB, and 2MB chunks of pages all get swapped out at once.

Pagefile reads are generally only 4KB, in whatever order your application accesses the pages.

My post was a contradiction and I don't know why I mentioned 4k because those are read numbers. What I wanted to say was most writes were sequential regarding pagefile.