Folding Forum

This is no marketing talk - back in May 2008, the author of these lines spoke with Vijay Pande of Pande Group [one of their projects is Folding@home] at Stanford University about the need for more efficient way for researching protein life. The problem then was that the world's most efficient GPU for Folding@home was the GeForce GTX 280,which was capable of simulating 600 nanoseconds of protein folding in 24 hours.

Today, in 24 hours of non-stop running, Dual-GPU GeForce GTX 295's are capable of running a protein folding simulation at 1400 nanoseconds, or 1.4 milliseconds. That means that unfortunately, we need 714 GTX295's to simulate a single second of protein folding. If you want to simulate 24 hours, you need 61,689,600 GeForce GTX 295 graphics cards, i.e.123,379,200 GT200-class graphics processors. Before you ask - yes, that number is higher than the overall number of every manufactured GT200 chip in the world. And that was "just" for Folding@home. At that time,a single CPU core was able to process around 4 [yes, four] nanoseconds,i.e. 150 times slower than the GeForce GTX 280, yet alone newer GPUs.For 24 hours of protein simulation you need 123 million GT200 GPUs or18,506,880,000 Core 2 CPUs. Yes, that's 123 million versus 18 billion.

And yet you still need those "slow" CPU clients to do the finer details of the simulations because the chainsaw accuracy of the GPU clients doesn't cut it close to the bone, so to speak.

gucci904 wrote:This is no marketing talk -

You certainly sold me....
Get your friends and relatives signed up and get them to fold for your team. Tell them they'll make millions if they get in first place.

7im wrote:And yet you still need those "slow" CPU clients to do the finer details of the simulations because the chainsaw accuracy of the GPU clients doesn't cut it close to the bone, so to speak.

Thereby proving that my 40-50 PPD WinMe running Ver. 4 F@H on my print-server/bit-bucket still gets WUs and is providing useful addition to the project.

gucci904 wrote:Today, in 24 hours of non-stop running, Dual-GPU GeForce GTX 295's are capable of running a protein folding simulation at 1400 nanoseconds, or 1.4 milliseconds. That means that unfortunately, we need 714 GTX295's to simulate a single second of protein folding.

1400 nanoseconds is 1.4 MICROseconds. You need 714,000 GTX295's if your stats are accurate.

Yeah but how much time is really needed to get the relevant information? I had assumed that only the first few milliseconds were interesting. Would you ever need 24 hours of simulation?

I would like to see the software get better, the simpler it is to download,install and configure the more people will do it, I think the GPU and CPU should start working together at some point as was said above the GPU does course crunching and the CPU is more flexable for some of the more specific parts of the problem from what I see currantly you have a CPU on one WU and the GPU on another.
also the ability to employ as many cores as possable this is the system tray software. you could have a comand line program but a simpler UI would be helpfull maybe even monitor things like GPU/CPU temp to keep from damaging systems.
as for the display I know the simulation runs much faster with out it, I dont need a real time image just one that updates every min or so would be nice.
I see the day when all computers will have distrubited computing imbeded in them, that allows the owner to make available the as much power of there systems as they choose.

I7 860 2.8Ghz 4Gig RAM Geforce GTX 275 896MG from FalconNW

My thinking, yes its great to have theses huge gpu farms, like atlas, but the overall success is done by the smaller cpu and smp folks, who just whirl away the small stuff.

patonb wrote:My thinking, yes its great to have theses huge gpu farms, like atlas, but the overall success is done by the smaller cpu and smp folks, who just whirl away the small stuff.

Don't agree on labeling any aspect of being the part that contributes the most to 'overall succes'.

Every client has it's niche in which it excels, now maybe one of the involved scientists could give their opinion on which client they feel has the most benefit to the project but that to would be personal opinion and maybe even preference for the client they are the most involved in running simulations on.

bhampton wrote:I would like to see the software get better, the simpler it is to download,install and configure the more people will do it, I think the GPU and CPU should start working together at some point as was said above the GPU does course crunching and the CPU is more flexable for some of the more specific parts of the problem from what I see currantly you have a CPU on one WU and the GPU on another.
also the ability to employ as many cores as possable this is the system tray software. you could have a comand line program but a simpler UI would be helpfull maybe even monitor things like GPU/CPU temp to keep from damaging systems.
as for the display I know the simulation runs much faster with out it, I dont need a real time image just one that updates every min or so would be nice.
I see the day when all computers will have distributed computing imbedded in them, that allows the owner to make available the as much power of there systems as they choose.

I7 860 2.8Ghz 4Gig RAM Geforce GTX 275 896MG from FalconNW

Actually Mac OSX has had distributed computing as part of the OS for some time now called Xgrid. In the sharing control panel one can set up to "share" their computer's cpu. There are some projects out there that use Xgrid but I haven't done it in a while because I have been folding.