Folding@Home ATI Announced:

[Dave Baumann]

http://folding.stanford.edu/FAQ-ATI.html

Now in 2006, we are looking forward to another major advance in capabilities. This advance utilizes the new, high performance Graphics Processing Units (GPUs) from ATI to achieve performance previously only possible on supercomputers. With this new technology (as well as new Cell processor in Sony’s PlayStation 3), we will likely be able to attain performance on the 100 gigaflop scale per computer. With this new software and hardware, we will be able to push Folding@Home a major step forward. We are beta testing the ATI GPU client software internally at the moment and will likely announce an open beta in four to five weeks (end of September).

R580+ vs PS3? Rumble! :p

[pascal]

:cool:
One more reason (to me) to upgrade to a new AMD/ATI combo, and be back folding.

Will it work on the X1650?

[Gnerma]

Dave do you know anything about the performance delta between R520 and R580? How big is it? Vijay isn't talking :???: Also, I assume the GPU client needs the full memory controller that is only on the cores mentioned above?

[Dave Baumann]

I don't know, but I'd be willing to guess something in the range of 0% - 200% difference! ;)

[Gnerma]

... something in the range of 0% - 200% difference! ;)
I'll hold you to that :cool:

[Geo]

http://folding.stanford.edu/FAQ-ATI.html



R580+ vs PS3? Rumble! :p


http://www.beyond3d.com/forum/showpost.php?p=815633&postcount=14

Beatcha by a full day!

[Bohdy]

Is FoldMark going to be a standard GPU benchmark now, I wonder?

Or will we just have to stick with RyderMark...? :lol:

[chavvdarrr]

Dave do you know anything about the performance delta between R520 and R580? How big is it? Vijay isn't talking :???: Also, I assume the GPU client needs the full memory controller that is only on the cores mentioned above?
Which GPUs will be supported? We have not made any final decisions on this issue. However, our software will likely require the very latest GPUs from ATI (especially now that the newest ATI GPUs support 32 bit floating point operations). Previous work of ours used NVIDIA GPUs as well, but we have now concentrated on ATI GPU's as they allow for significant performance increases for FAH over NVIDIA's GPU's (at least at the current generation). Our GPU cluster has 25 1900XT's and 25 1900 XTX's. We find a considerable performance increase of 1900XT's even over 1800XT's, due to the architectural differences between the R580 and R520 GPU's. Our code will run on R520's, but considerably more slowly than R580. We're very much looking forward to trying out R600's.
http://folding.stanford.edu/FAQ-highperformance.html

[Guden Oden]

Why aren't Nvidia in on this? :( I have no particular love for that company, but that's the card in my PC right now, and I got no money to replace it...

[Geo]

Why aren't Nvidia in on this? :( I have no particular love for that company, but that's the card in my PC right now, and I got no money to replace it...

The implication one hears is that robust (perf-wise) dynamic branching is a big part of what makes this go.

[nelg]

The implication one hears is that robust (perf-wise) dynamic branching is a big part of what makes this go.

ATI could use this advantage for marketing.

ATI - Helping save more lives than the competition.

[Geo]

The real important question --will only X1950 users be able to use this without being driven batty by their cooler fans? Or will they add a soundtrack to mask it for X1900XT/X users? :razz:

I'm not entirely kidding in that --I wonder if they'll have a slider similar to the cpu version to tell it how hard you're willing to have it push your gpu? We've seen them talk about heating the gpu up considerably in testing. I'm not looking forward to my X1900XT running at 90c 24/7, with the fan going 50%! Neither for my ears nor the longterm health of the gpu.

[SugarCoat]

need to see it get released! Then i'll be happy. Kinda miffed that they havent released an SMP compliant version, without having to jump through a flaming hoop to get it to work that is. I usually have one of my cores hacking away, but i could easily switch it to both anytime the computers idle, or to 3/4 or 4/4 cores when i get the intel Kentsfield together. Not completely impressed with how they are letting the whole SMP thing slip by. Tell ya though my CPU 1 has been getting a nice work out. It hasnt been idle more then 5 minutes in the last few months.

[Dave Baumann]

Just install multiple clients for SMP.

[SugarCoat]

its not that simple cause the client assigns a single unique ID to the computer (at least last time i checked), theres some wierd step by step process and a changing of registry keys to get it to work though. I'm just too lazy to do it so i'm waiting for a proper version.

[BRiT]

Install them into two different directories and use the "-local" cmd line flag. That tells it to read/write the config from local files and not to use the registery. Another thing to do is to set the affinity of each so each stays on their own core. Really simple indeed. Nothing I would consider to be flaming hoops.

[Geo]

http://www.techreport.com/forums/viewtopic.php?t=43530

I dunno. I scanned that and said "too much trouble". Maybe not "flaming hoops", but more than I care to deal with, and that for a text-only client anyway.

[Gnerma]

That guide makes it look much too hard. All you have to do is:

1) Put two console clients in two different directories.
2) Run each client to configure them and assign them different machine IDs (in advanced options).

Thats it. The -local flag hasn't been needed for a number of years. Start and stop the clients however you want.

[Geo]

So who besides me is folding for B3D with an X1900? Nice bump in my output last week.

[Sinistar]

I am, but only 12 hours a day on weekdays, 24 hrs on weekends. When I'm sure it not going to blow up or something:lol: , I will probably do 24/7.

[Dave Baumann]

Multi GPU client is now available. Here's the how to:

http://fahinfo.org/gpu/multi_gpu_howto.html