GTX 470/480 cards not folding yet; bug in FAH core

[Zero82z]

Originally Posted by 0o0

hey zero what does that flag do?

It tells the client to treat the GPU as if it were an R700 generation chip.

Originally Posted by 0o0

damn this is a dilemma for me lol, I've got a 5870 and a i7 so I've gotta keep it on 24/7 for -bigadv but I don't want to have my 5870 folding 24/7 at 5% efficiency or have it idling and guzzlin power while the i7 folds, but if I remove it and replace it with a 4650 I won't be able to game :( I can't keep switching back and forth either, just don't have the will to do that... so it's either folding or gaming for me. I wish stanford would hurry up with the ati support; if they did, they would see more WU's coming in

Why again can't you just fold on your CPU and GPU at the same time?

Originally Posted by 10e

This flag tells this particular GPU2 client to use ATI folding binaries.

Not quite.

[0o0]

dang I was hoping it'd boost the efficiency or something... I don't plan on folding on my 5870 because it's not nearly as efficient as it could be. That, and it dumps a helluvalotta heat into my room to the point where it's unbearable (as in I'm at risk of heat exchaustion sitting in my seat in the same room as my 5870 with the door/windows shut).

[Zero82z]

Originally Posted by 0o0

dang I was hoping it'd boost the efficiency or something...

Boost the efficiency of what?

[0o0]

ati cards have a lot more of these 'teraflop' thingies than nv had, so I'd assume they have more raw power, but stanford decided they want to support cuda instead of havok/opencl (or w/e ati is using right now), so ati cards fold much less than what they could be folding if the client was optimized for it. i.e. per watt, ati can produce much more ppd but isn't doing that right now, which is very unfair and why I'm not going to be folding until stanford makes it fair. I could be wrong though... these are based on my googling at EOC and others.

[Zero82z]

Originally Posted by 0o0

ati cards have a lot more of these 'teraflop' thingies than nv had, so I'd assume they have more raw power, but stanford decided they want to support cuda instead of havok/opencl (or w/e ati is using right now), so ati cards fold much less than what they could be folding if the client was optimized for it. i.e. per watt, ati can produce much more ppd but isn't doing that right now, which is very unfair and why I'm not going to be folding until stanford makes it fair. I could be wrong though... these are based on my googling at EOC and others.

Your understanding of the situation is completely wrong.

[crazyhorsejohnny]

From what I've "read", ATI deserves most of the blame.

[0o0]

Could you guys explain? I know that 1/5 ati's shaders are crippled and only 'look-good-on-paper', but I don't think that the folding client can't make any use of it at all =\ is it because ati keeps changing architecture or something?

[chrisk]

Originally Posted by 0o0

ati cards have a lot more of these 'teraflop' thingies than nv had, so I'd assume they have more raw power, but stanford decided they want to support cuda instead of havok/opencl (or w/e ati is using right now), so ati cards fold much less than what they could be folding if the client was optimized for it. i.e. per watt, ati can produce much more ppd but isn't doing that right now, which is very unfair and why I'm not going to be folding until stanford makes it fair. I could be wrong though... these are based on my googling at EOC and others.

No. The nVidia architecture has better error checking (for the types of calculations done folding) and thus wastes less resources.

[Zero82z]

I don't feel like giving the whole big explanation, but it basically has nothing to do with bias and it's mostly because the F@H client is just not optimized properly for ATI's architecture.

Originally Posted by chriskwarren

No. The nVidia architecture has better error checking (for the types of calculations done folding) and thus wastes less resources.

Not quite.

[SugarJ]

Ya know, Zero, that's rather maddening. Why not explain it once, rather than just tell everyone they're wrong. It'll save time in the end, especially if you happen to be right.