[Kevin OBrien]

In today's double SSD review we look at the Corsair Force Series 3 in both 120GB and 240GB capacities and how they compare to the Force GT series. The main difference between these SandForce-equipped SSDs is the NAND used inside.

Corsair Force Series 3 Review

[johnw42]

The SS data is interesting. The F3 held up fairly well.

Thanks for adding the Performance Pro. It looks very similar to the Plextor M2P on the SS test.

Can you give the parameters of the SS test here (and in future articles)? I am guessing the queue depth (QD) must be at least 4, judging by the burst write speeds you give, since almost all SSDs seem to test at around 60 - 80 MB/s on your standard 4KiB random write test (which I assume has QD=1).

Also, very relevant to the Corsair Force 3 is what type of data IOMeter is writing. Did you set it up for full random or pseudo-random? Full random would be best.

This post has been edited by johnw42: 29 December 2011 - 01:44 PM

[Kevin OBrien]

Our steady state benchmarks all use fully random data over the full LBA size of the drive, while the standard benchmarks use a limited section of the drive to work within the burst/sustained constraints of garbage collection methods. The QD is also high enough to saturate the 4K limits of any single drive, although under what you might need to measure burst on a PCIe solution (but not steady since that is still slow enough on even the fastest cards).

[johnw42]

Kevin OBrien, on 29 December 2011 - 02:20 PM, said:

The QD is also high enough to saturate the 4K limits of any single drive, although under what you might need to measure burst on a PCIe solution (but not steady since that is still slow enough on even the fastest cards).

But what is the actual queue depth used in IOMeter? Are you saying it is different for each SSD? If so, that is not good methodology. I suggest using the same value for all SSDs. If you want to saturate the throughput vs. QD relation for all SSDs, QD=16 would be sure to do it. Probably QD=8 would be sufficient, but QD=16 definitely would be.

[Kevin OBrien]

Oh on single SSDs we settled on 32 since that is what most of the enterprise community is using on spec sheets and the like. Technically speaking 16 or 64 would also work, but none of the latency numbers would align . We are still working on an optimal value for PCIe solutions though since some have issues if you bring it over the point of over-saturation while others do not.

[johnw42]

Kevin OBrien, on 29 December 2011 - 02:29 PM, said:

Oh on single SSDs we settled on 32 since that is what most of the enterprise community is using on spec sheets and the like. Technically speaking 16 or 64 would also work, but none of the latency numbers would align . We are still working on an optimal value for PCIe solutions though since some have issues if you bring it over the point of over-saturation while others do not.

So QD=32 on the SS tests for all SSDs. Good to know, thanks.

[johnw42]

Why does this review not show up in the list when I click on "STORAGE REVIEWS" from the main SR website?

[Kevin OBrien]

When drives are still in the feature block of the website, it doesnt add into the reviews page yet. Basically once it drops out of that part of the news feed it gets into the review section.

[Julius Wijaya]

Hi Kevin,

pardon my foolish question, but i really need to know the thickness dimension of this drive, as i plan to get one for my laptop, and it need no more than 7mm thick.

thank you,

[Kevin OBrien]

This drive is a 9.5mm thick model, so well over 7mm.