GPU Folding problem

I just got this card back from an RMA service and want to give this GPU folding a go. However, I keep getting EUE. I installed the latest drivers from the ATI-AMD website and the card is currently running at stock speeds. The install I'm using is directly from the F@H website. (console, not graphic)

Launch directory: C:\Documents and Settings\Alex\Desktop\F@H GPU
Executable: C:\Documents and Settings\Alex\Desktop\F@H GPU\FAH5.91beta3-console.exe


[18:46:13] - Ask before connecting: No
[18:46:13] - User name: inkfx (Team 32)
[18:46:13] - User ID: D978D9313C5881A
[18:46:13] - Machine ID: 3
[18:46:13]
[18:46:13] Loaded queue successfully.
[18:46:13]
[18:46:13] + Processing work unit
[18:46:13] Core required: FahCore_10.exe
[18:46:13] Core found.
[18:46:13] Working on Unit 01 [February 13 18:46:13]
[18:46:13] + Working ...
[18:46:13]
[18:46:13] *------------------------------*
[18:46:13] Folding@Home GPU Core - Beta
[18:46:13] Version 0.10 (Mon Oct 30 12:32:17 PST 2006)
[18:46:13]
[18:46:13] Compiler : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 13.10.3077 for 80x86
[18:46:13] Build host: CYGWIN_NT-5.1 vishal-gpu 1.5.19(0.150/4/2) 2006-01-20 13:28 i686 Cygwin
[18:46:13] Preparing to commence simulation
[18:46:13] - Looking at optimizations...
[18:46:14] - Files status OK
[18:46:14] - Expanded 86882 -> 443709 (decompressed 510.7 percent)
[18:46:14]
[18:46:14] Project: 2736 (Run 8, Clone 5, Gen 5)
[18:46:14]
[18:46:16] Assembly optimizations on if available.
[18:46:16] Entering M.D.
[18:46:24] Working on 1258 GPUlambda
[18:46:31] Starting GUI Server
[18:46:39] mdrun_gpu returned 114
[18:46:39] Going to send back what have done.
[18:46:39] logfile size: 7138
[18:46:39] - Writing 7676 bytes of core data to disk...
[18:46:39] Done: 7164 -> 2710 (compressed to 37.8 percent)
[18:46:40] ... Done.
[18:46:44]
[18:46:44] Folding@home Core Shutdown: EARLY_UNIT_END

what drivers do you happen to be running?

Catalyst 7.1

Download option #2 http://ati.amd.com/support/drivers/xp/radeonx-xp.html

from the FAH website, FAQ on the GPU

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

* Catalyst driver version 6.5 [download from ATI or from a Stanford mirror ]. Version 6.10 [see main ATI Folding@Home page or download driver from ATI ] with the newest core appears to be working now (but requires more testing to be sure). Versions 6.6 and 6.7 will work, but at a major performance hit; 6.8 and 6.9 will not work at all. Due to all the complexities of support, we will support only versions 6.5, 6.10, and 6.11. Version 6.12 does not work correctly, but we are expecting that the 7.x series to solve that problem.
* DirectX: 9.0c (4.09.0000.0904) or later, [download here] , which yields d3dx9_30.dll (the critical part for FAH)

Click to expand...

Hm, so I guess 7.1 does not solve the problem. I will try switching to 6.11.

Let us know how it goes.

So far so good with the Catalyst 6.11 drivers. I had a hard time locating a good download server in America but finally got it from techpowerup. I havn't finished a frame but at least I don't EUE right away. Thanks for all the help guys, it's greatly appreciated.

you are welcome to the help
got any overclock going on it?
whats the time per frame?

No overclock on the video card. Looks to be about 30 minutes per frame.

[19:08:30] Completed 12
[19:53:36] Completed 13
[20:20:42] Completed 14
[20:50:31] Completed 15
[21:20:17] Completed 16
[21:49:50] Completed 17
[22:19:45] Completed 18
[22:50:05] Completed 19
[23:19:59] Completed 20

Atleast it is working now!

Is my time normal? I also have 2 instances running on my E6400 @ 3.2Ghz.

For maximum ppd with the cpu and gpu clients, you should run only 1 cpu client. To give you a perspective on the frame times gpus are capable of, an X1950 XTX @ 695 does 5:40 frames. 256 MB X19x0 cards do around 8:00/frame. To really maximize ppd you should run the SMP client under wmware/linux and the gpu client.

You'll also need an overclocking tool to get maximum performance out of the card. You don't actually have to overclock it, though it should easily, but the drivers will not recognize FAH as a 3D app and run at much slower clocks without a OC utility like ATITool or ATI Tray Tools. I use both and both work.

ChasR said:

For maximum ppd with the cpu and gpu clients, you should run only 1 cpu client. To give you a perspective on the frame times gpus are capable of, an X1950 XTX @ 695 does 5:40 frames. 256 MB X19x0 cards do around 8:00/frame. To really maximize ppd you should run the SMP client under wmware/linux and the gpu client.

Click to expand...

So with only 1 CPU client running my GPU will get more ppd than if I ran 2 CPU clients + GPU client? Since F@H is not recognized as a 3D app. I would have to overclock the 2D speeds, right?

You shouldn't have to OC at all to get 3D clocks, running either ATITool or ATI Tray Tool will force 3D clocks.

Right now I overclocked the card to 700/800 with Rivatuner 2.0 FINAL. Let's see how much this will improve frame time. (dual CPU instances running)

with 2 cpu instances running at same priority as gpu your gpu frame times will not be very good.
My xtx at 700/800 gets about 5.45 running with 1 cpu client.

With the current mix of windows cpu wu's, you will get more ppd running 1 windows cpu client and 1 gpu client.

For more ppd, run 1 gpu client, and run linux smp folding client under vmware.
Your gpu frame times will be slower, but overall your ppd will be higher.

inkfx said:

Right now I overclocked the card to 700/800 with Rivatuner 2.0 FINAL. Let's see how much this will improve frame time. (dual CPU instances running)

Click to expand...

dude. You just need to force 3d clocks (mentioned a couple times in here, if you still dont know, just ask again ), and yes,
1 uniprocessor client + 1 GPU client
or
1 SMP (through VMWare) + 1 GPU

for max PPD, the SMP way is probably better, it may take an hour or two to get setup, ask questions, or look at Leelegend's thread on "beginner VMWare and SMP guide" or whatever it is.

Just to help you get a sight of it, running 2 uniprocessor clients, I'm not sure the PPD on that CPU since I dont pay that much attention, but I'm guessing 500ppd average with teh "wrong WU's". With the GPU ONLY, you can hit 600-800 PPD just depending on your clocks. The card in my sig, I only pushed up the core MHz, and I was getting 6 min tpf's (cant remember the PPD right now, and I'm at work) so yeah, screw the uniproc' client if you have a SMP or GPU to run

ohh! And since you're new (it looks) to the GPU folding, you need 1 CPU core to 'feed' the GPU....I guess that didn't get mentioned here anywhere