sieanr said:
The cell is less limited than a GPU, but its more limited than a CPU. Thats why its faster than a desktop CPU in certain areas - its in-between a GPU and CPU. The things it can do well are the same things a GPU can do faster. Also, Folding at home is a dumb way to estimate speed since the assign different types of workloads to different processors. If you gave the Cell the same workloads as a general purpose CPU then it would do terrible. |
If by 'more limited' you mean 'slower if not optimized, but can do exactly the same stuff', then you are correct (not that you notice this slowness much if at all, Linux PS3 is fast aplenty barring some RSX issues). Otherwise you are dead wrong. The Cell can execute the same code an ordinary PPC can.
Why else do you think PPC linux runs on the PS3 without much modifications? Heck, the first Linux versions for the PS3 where PPC64 Linux with a different bootloader and some hardware support.
(Now, fully optimal Cell code is indeed a rather different beast (what with all the SPE's you need to program in a direct fashion to get the best out of it) but unlike a GPU Cell can just run any old PPC64 code you throw at it. Usually fast enough too. Not so odd, because it does contain a standard PPC core as well as it's SPE's)
---
Secondly, where did you read that Folding makes a difference for workunits between platforms anyway? I can't find that, all I can find is how lyrical the people over at Folding@Home are with the Cell/GPU code being so much faster at solving their units than the standard code is.







