Folding@Home

[the_lone_wolf]

Originally Posted by northwind

Y...building inefficient distributed "farms" and running them at high cost doesn't make much sense at all.

i haven't heard a single person claim they're building a dedicated folding rig to make the maximum contribution to medical science, it's a technical exercise, a challenge to overcome, nobody's claiming it's the most efficient way of contributing

Have you seen some of the scores people have? Id say they are the farmers... Or they are Uber g33ks with 10PC's already

How good is the PS3 at this folding I wonder...

LMAO they think that the reason the PS3 is good at folding @ home is because of the Cell processor... Common people, its the GPU not the CPU. ffs.

http://fah-web.stanford.edu/cgi-bin/...?qtype=osstats

Wow at the GPU's.

Think I might give this a crack

[the_lone_wolf]

Originally Posted by ThEGr33k

LMAO they think that the reason the PS3 is good at folding @ home is because of the Cell processor... Common people, its the GPU not the CPU. ffs.

GPUs are much better at parallel computing, the kind of thing F@H requires, i run F@H using the processing power of my 8800Ultra graphics card, it uses about 2% of my CPU

look up CUDA and other methods of GPU computing

Originally Posted by the_lone_wolf

GPUs are much better at parallel computing, the kind of thing F@H requires, i run F@H using the processing power of my 8800Ultra graphics card, it uses about 2% of my CPU

look up CUDA and other methods of GPU computing

Aye ive read a bit about it but never got round to doing it... I got a 9800GTX which should churn out some decent numbers I think

[the_lone_wolf]

Originally Posted by ThEGr33k

Aye ive read a bit about it but never got round to doing it... I got a 9800GTX which should churn out some decent numbers I think

i remember seeing somewhere that F@H is about 10-40x faster on a decent GPU compared to a decent CPU

look at the CUDA website, depending on the application some run ~50% faster on a GPU, some nearly 500x (50,000%) faster - it's a big thing, my current CPU (Q9550) trabscodes a full length movie to wmv in about 30 minutes, my previous CPU (A64 3500) took 90 minutes or so, on my GPU can do it say 10x as fast that reduces the time to a few monutes, to transcode a full length movie!!!

the software that takes advantage of GPU power is still very much in it's infancy, but i'd predict we'll move away from the "CPU and GPU" computer and more towards a PC with two processors, one linear, one parallel, and the hardware/software decides where to send each task depending on whether it would complete faster on either type

Aye that sounds about right. It is mental how powerful the modern GPU is.

Wondering when im running this Folding at home on the GPU what effect does it have if I want to play a game at the same time? On the game that is...

[northwind]

It's not bad... I used to run seti@home back in the day, and even on the low settings it was intrusive, it took too long to react to other processing loads. It had the same sort of impact as a bad antivirus. But I had folding on this machine for a while and while it did slow things down sometimes, it was less intrusive on the high settings than the older clients were on the low settings. Very good from that point of view.

I should put it back on actually, I'm not actually anti-Folding, just anti-farming I'm having a go at a specific weirdo folding subculture here, not everyone!

Maybe we can make a team... LOL

Im just running the GPU one... 110GFlops is enough the other 9 my CPU might give is pointless imo. ha ha

[northwind]

Originally Posted by the_lone_wolf

i haven't heard a single person claim they're building a dedicated folding rig to make the maximum contribution to medical science, it's a technical exercise, a challenge to overcome, nobody's claiming it's the most efficient way of contributing

Where's the technical challenge? It's just box-building, then it sits in the corner and does it. From reading the interviews in CPC mag, a lot of these guys build several identical machines to do it.