PDA

View Full Version : GPU Folding


RiverRicer
10-01-2007, 09:48 PM
OK, guys and gals! Since we're getting some 'uber folders' with 'them thar Sony things' on the team, I thought it might be time to try out the GPU folding client just to save face. The problem was I didn't have an ATI card, only nVidias, so I jumped on fleaBay and snagged an X1900 for ~$100. When the card arrived, I downloaded the latest Catalyst drivers and the GPU client, then installed both. Fired the thing up and, voila, the GPU runs 600 points per day using one core of my Brisbane box! The other core is folding regular work units, so the box is putting out around 750/day intead of 300. Do I get my cookies and milk now?

peti1212
10-02-2007, 02:30 AM
Wow, nice. You mind buying me one of those cards. I'll start folding even more if you do. :) You get some cookies and milk.:grin:

Here:

http://www.displayit-info.com/food/images/desserts/2233.JPG

Good Eating. :)

tyle6
10-02-2007, 03:31 PM
i had the same idea:grin::grin: i saw what those PS3's were doing and tossed the GPU client on my x1650!! not sure what its gettin a day in points but its turning out close to a unit a day easy! all well the P4 its in is running the CPU client aswell!! Being a P4 with HT the CPU client doesn't seem to be affected by the GPU client speedwise what so ever or vice versa!!
The GPU client is just amazing! anyone with a radeon should give it a whirl!!

Nocutius
10-02-2007, 08:16 PM
Those radeons are sure worth it for folding. The only thing i don't understand is why nvidia doesn't have any cards in the ring. The 8800s would surely kick some ass.

RiverRicer
10-03-2007, 01:21 AM
Peti: Thanks for the munchies! :grin:

Tyle6: Fold on, Dude! :keepfolding:

Nocutious: Below is part of the FAH faqs that alludes to some problems with nVidia cards. In any case, they don't work well for this application. :confused:
---------------------------------
GPU and OS support
Which cards are supported?
We now support serveral classes of GPU boards, including X1600, X1800, and X1900 class GPU's from ATI. At the launch, we supported X1900 cards only. X1800 cards do not provide the performance seen in X1900's and so we strongly recommend X1900 class cards (although we now officially support X1800). X1900 and X1800 cards are actually quite different -- they have different processors (R520, R530 vs. the R580 [in the X1900 series]). The R580 makes a huge difference in performance -- its 48 pixel shaders are key, as we use pixel shaders for our computations. Also note that the card should have more than 256MB (i.e. at least 512MB) of RAM, otherwise the GPU client will put a huge load on the client machine (although we do note that the 256MB X1950Pro using PCIe does work reasonably well).

What about video cards with other (non-ATI) chipsets?
The R580 (in the X1900XT, etc.) performs particularly well for molecular dynamics, due to its 48 pixel shaders. Currently, other cards (such as those from nVidia and other ATI cards) do not perform well enough for our calculations as they have fewer pixel shaders. Also, nVidia cards in general have some technical limitations beyond the number of pixel shaders which makes them perform poorly in our calculations.

Kougar
10-05-2007, 09:42 AM
Those radeons are sure worth it for folding. The only thing i don't understand is why nvidia doesn't have any cards in the ring. The 8800s would surely kick some ass.

Nvidia can't be bothered to support F@H, since Stanford refuses to use CUDA to write their application in. This is because Stanford writes their program in DX10 code so they will not be locked into a vender specific GPU. Nvidia's drivers are reportedly still to buggy and for whatever reason they have presently refused to fix them to allow Stanford to code with them. http://forum.folding-community.org/ftopic16684-0-asc-195.html

I wish I had some stats on how many people went out and bought ATI cards just to fold on, I've seen a surprising number of users doing this with used/new ATI cards. Maybe it would convince Nvidia it was worth their time.

Nocutius
10-11-2007, 12:57 PM
Anyone tried the new gpu console yet. I did and i noticed something strange. Previously the console would shut down when you started a game, now when i played cnc-tiberium wars id didn't. The console ran and did its thing and the game ran with no problems as well.
Then later, straight after finishing with cnc, i tried stranglehold, console did not shut down, i get into the game, and it runs like crap, cause it's obviously much more graphically demanding than cnc. So i quit the game and then the console shuts itself down also, not before but only after i quit the game.
I already thought i could leave the thing running 24/7 but apparently not.

Kougar
10-13-2007, 08:20 PM
I would be really surprised if you could. From what I understand the GPU doesn't have the ability like a CPU to prioritize threads... the GPU simply runs what it is told to run. So if you run multiple graphical apps that try to use it at the same time the GPU will run all of them and likely kill the performance for all of them.

Raptorfury
10-27-2007, 08:42 PM
Any word on supporting the HD2900Xt cards ? i currently have 1 and waiting for rma for the second card ( should put up someserious Numbers )

Kougar
10-28-2007, 05:51 PM
There is no word as of yet, although it is expected to arrive at some point. I tried to check right now but their forums are down for maintenance.

SonicWRX
10-29-2007, 04:55 PM
I checked last night and the current post on ATI drivers said that the HD2X00 cards are still not supported and a new core will need to be made for them.

Every one out there in ATI country is chomping at the bit for these cards to be supported. I am sure that it is coming.

Kougar
10-30-2007, 06:30 PM
Thanks for the info. :)

One little known fact is Stanford already figured out a way to use both GPUs for those Folding@home Crossfire users out there, so I imagine they will be fairly focused on work for a new core for ATI hardware. With ATI set to launch shrinked R600s it would give the new folding core a wider base of supported hardware than before.

SonicWRX
10-31-2007, 02:21 PM
I have been trying to keep an eye on whats going on. I know that i am waiting to upgrade any computer components until after the holidays. I am very eager to get a new ATi and start it folding. Witch it funny cause I am not folding with the two cards I have now.... (sorry) Just too much going on with things. I hope to get back at it soon.

On that note anyone here know of a good way to have consoles run on a schedule? I was thinking of using windows scheduler and kicking off the task at say 9am and set the setting "Stop task if it runs for XX hours XX minuets" I'd like it to run for 10 hours a day and no longer. I am fairly sure this will work. Any thoughts?

ZhengHe
02-19-2008, 07:28 AM
I actually haven't tried that, but in theory there is no reason why it can't work. The problem arises with actually figuring out how to do it. If the F@H console won't natively allow for scheduling then you'll have to go digging a bit to find out what values you would have to reset in order to get this running. Assuming there is a method there really is no reason it shouldn't work, but the question is what you can alter or add on in order to accomplish this. Perhaps there is some external F@H software, which you can use in conjunction with the console to manipulate it in this way.

I was looking at the F@H FAQ on which Graphics Cards it can use in conjunction withe GPU client and something very interesting stuck me. Compatability ends with none HD cards: http://ati.amd.com/products/home-office.html. Is there a problem with using these cards or are they simply in the process of trying to get them on board with the GPU client as well? Do the HD cards have fewer pixel shaders or is it another issue, which prevents them from being used effectively?

The F@H GPU FAQ also mentions that Nvidia cards perform poorly, which is why they were not made compatable with this GPU client. Someone else also mentions a proprietary architecture, which precludes the development of folding software for these cards. So are they cards completely incapable of using a GPU folding client or will they simply run extremely poorly and perhaps be a big buggy or unstable?

SonicWRX
02-19-2008, 05:20 PM
The last time i checked the Newer ATI cards are not supported due to problems with the FAH core needing to be rewritten from scratch. As the Chip architecture totally changed.

The other problem is that the FAH cores need to produce good results. If the results are flawed for any reason they can not use them. They had a working Core for the HD cards but a flaw was found and it had to go back to the drawing board. I belive that they are testing a new one, but they are now having performance problems. Stanford will not release a core to the public unless it shows advantages over the standard Core that is run on CPU's.

As far as Nvidia they tried but like they said the performance was poor. Thus they did not use them. Not all GPU/CPU configurations lend them selves well to the "science" of folding.

Kougar
02-22-2008, 02:34 AM
I don't really know about Windows scheduling, that isn't something I've ever used. I am not aware of any scheduling features in Folding@home either, sorry!

I see that the link from my earlier post in this thread is broken... because Stanford has done a much needed reorganization of their forums!

I am looking at the two stickies made here (http://foldingforum.org/viewforum.php?f=10&sid=fa6e38dc61d7b778939a3cb42612a43a):

Folding@Home does not currently run on nVidia hardware due to incompatibilities between Brook and nVidia display drivers. All parties are currently working on a solution, but the timeframe is unknown.

It is not currently being developed with CUDA because the aim is for Folding@Home to be a cross-platform solution and CUDA only works with nVidia GPUs.

1xxx series hardware is currently supported.

2xxx and 3xxx development was progressing well and an pre-beta client was demoed. Unfortunately a serious bug was found during internal QA which prevented its release in late Q4 07. Although PandeGroup appreciate the offers of further beta testing and QA from the community, confidence in the scientific validity of the results produced by the buggy pre-beta is low and it would not be possible to credit points for that work. A full QA run takes X days/weeks and several iterations of code updates and QA may be needed.

We anticipate a release in Q1 08 (or H1 08).

(Edited to remove implication that this is an official statement)

SonicWRX
02-22-2008, 08:52 PM
Here is the latest news on the GPU Client

Taken from Here (http://folding.typepad.com/)



We're pretty happy with the results of the QA from our new GPU core and we are working with ATI to do some more extensive QA. If that goes well, the next step is a closed beta, and then an open beta (i.e. downloads on our web site).

It's taken a while to get to this point due to many fixes and changes under the hood. Some of these changes have to do with new ATI development tools for using GPUs (these changes will help the reliability of the code as well as solve a lot of issues donor side, such as problems with drivers) as well as new science we've put into the GPU code, to bring it up to date with what we now do on the PS3 (in fact, some features of the GPU code are not present in the latest 1.3.1 PS3 version, but we hope they will be included in a future version).

I'll keep everyone posted here as we go along. It's hard to tell how long this current QA stage will take (and a lot of it depends on whether any major bugs surface), but we're happy to see this baby go out the door, at least with its first small steps.

trueg50
02-23-2008, 04:16 PM
So what do the folks at Stanford have against Nvidia? Or is ATI just funding their work?

They don't want to get locked into brand loyalty by using CUDA, however they will exclusively develop with ATI (using a much more difficult system I might add)? They are basically writing for one company any ways, they might as well do some work for the other company, and the other 60% of the graphics market

CUDA is much easier to write for than DX, and can be used on all 8 series cards, and all future Nvidia cards.

raidersforever01
02-23-2008, 05:15 PM
If they use DX it may make it easier to support larrabee if it comes instead of having to cut the team in half to use CUDA.

SonicWRX
02-28-2008, 02:25 AM
FYI if you currently are folding with your GPU do not update the the new beta5. its been declared a dud. it will not get work units and gives a windows error report sending message.