Relativo alle 58xx
by mhouston » Wed Sep 23, 2009 4:02 pm
On really large proteins you will see some speedup, but the current core is essentially still running in "6XX mode". On the small proteins, you will see minimal speedup, on massive proteins, much larger than currently going out public, we have seen 50% increases. Still not as much as we would like to see and we think we will be able to get down the road.
As OpenCL matures and we can migrate the core to that and begin to use the extended features available in that API and the new hardware. Note that it will take some time to do this, but moving to OpenCL should simplify support since the same code base will be able to be used across multiple vendors and hardware. If you look at some of the compute benchmarks posted for DirectCompute, the 5800 series does quite well, but most of those are small code bases and not the complexity of Folding@Home, but it does give a sense of what these parts are capable of.
Proteine piccole: più o meno stessa velocità di ora con una 4870, proteine grosse, fino a un 50% di più... schifo... una 5870 va il doppio di una 4870 e il quadruplo di una 3870
C'è da sperare per OpenCL... vabbè mi spiace ma questo progetto fa di tutto per far scappare le ATI