Introduction, Specifications, and Packaging
Today we take a look at the Corsair Force Series 3 and Force Series GT. Both are SandForce driven IMFT 25nm flash.
A couple of days ago we looked at a pair of SSD’s from Patriot. Next up is a pair of SSD’s from Corsair. These are another two SandForce controlled units, but this time it’s Async IMFT flash vs. Sync IMFT flash:
We’ll carry the Patriot Pyro (IMFT Async) into the results for comparison, and keeping the other benchmark OCZ and Intel models in with the mix of results. The OCZ Vertex 3 and Agility 3 will again share the same SandForce controller, but OCZ has been known to add many performance tweaks to their firmware. Let’s see if Corsair was able to use ‘tweaked’ firmware or instead went with the stock one provided by SandForce.
The Corsair Force 3 and Force GT are both available in the following capacities:
The added capacity points are a bonus of how IMFT can stack their dies in ‘odd’ multiples (i.e. 3 per package, making a 24GB TSOP). Varying slightly from low to high capacities (and across the two models), specs range from 490 to 525 MB/sec writes and 550 to 555 MB/sec reads. 60GB models get 80,000 4K IOPS and the rest get a rating of 85,000 4K IOPS. Corsairs specs indicate IOMeter 2008 was used for this test, and it’s important to note that 2008’s writes were a repeating pattern that is easily and fully compressible by the SandForce controller, meaning those specs were derived using fully compressible data.
I noted their 480GB* models are listed as having the same high IOPS specs, which we figure must be a misprint since *any* SandForce 2281 SSD with a capacity greater than 256GB will see a dip in 4K IOPS performance. This is due to the way the SF controller handles the mapping of LBA’s. To double capacity from 240GB to 480GB, the SandForce controller’s finite number of allocations must be reconfigured to utilize 8KB blocks (up from the standard 4KB – intentionally matched to the NTFS 4KB cluster size). This negatively impacts IOPS performance as a 4KB random write translates to the equivalent of an 8KB random write once the added overhead is taken into account.
If this is a truly hard limit, a 1TB SandForce 2281 SSD would have to again redouble its allocation unit to 16KB and would then be theoretically kneecapped to an again-halved ~20,000 4KB random writes.