A few weeks ago, OCZ got caught out after moving over to 25nm on the Vertex 2s. Although this development isn’t problematic in itself, the use of 8 GB chips lead to under-usage of the SandForce controller channels. Performances were as a result halved in sequential writes and there was a reduction in capacity of around 5 GB.
OCZ did eventually correct the situation by going back to 4 GB dies, still at 25nm, on the Vertex 2s (60, 90, 120 GB) affected by the loss of space and accepted exchanges for customers who wanted this. Unfortunately the story didn’t end there and AnandTech got hold of a Vertex 2 120 GB using Hynix 32nm memory.
Performance levels still vary because, with incompressible data, Anandtech scored the Hynix 32nm at 95.8 MB/s for sequential writes, against 136.6 MB/s on the original version with 34nm IMFT Flash and 118.5 MB/s with 4GB 25nm chips! Of course these variations are masked with highly compressible data, with scores of 215 to 221 MB/s.
After some discussion, OCZ has agreed to launch specific versions detailing the memory used:
- 25nm IMFT 80, 160 and 200 GB: OCZSSD2-2VTX80G.25, OCZSSD2-2VTX160G.25, OCZSSD2-2VTX200G.25
- 34nm IMFT 60, 120 and 240 GB: OCZSSD2-2VTX60G.34, OCZSSD2-2VTX120G.34, OCZSSD2-2VTX240G.34
This will guarantee that there aren’t any changes in spec of any given model, something OCZ should have done itself in the first place. NAND and therefore performance can still vary on the original Vertex 2s, which isn't of course a perfect situation. This would improve if OCZ published the minimum expected specs for each version, that’s to say the performance level obtained with incompressible data with the least best range of NAND used.
Variations in Flash memory within the same range are not of course anything new. Nevertheless, this didn’t previously affect the highest performance models, with only the Agility range affected at OCZ. This change can only be regretted, though the launch in parallel of versions that specify the Flash used is of course a good thing.