NVIDIA GPU F@H Client Test Results

There's a massive difference in cpu usage when using the gpu client on xp or vista, on xp the cpu usage stays somewhere between 25 and 50% whereas on vista x64 it doesn't go above 5% maybe it's the driver's but I would imagine it makes a big difference specially if you are running the smp client too.

 

im on 175.19 with the beta8 client..

 

GeForce ForceWare 177.41 for the GeForce GTX 200 series. (use a modded INF file for others I did with my 8800 GTX's)



Products supported:

GeForce GTX 280
GeForce GTX 260
nForce 780a SLI
nForce 750a SLI

WHQL release .
Adds support for NVIDIA PhysX Technology for GPU PhysX hardware acceleration on GeForce 9800 GTX, GeForce GTX 280, and GeForce GTX 260 GPUs. The GeForce v177.39 driver package includes NVIDIA PhysX System Software v8.06.12. The Unreal Tournament 3 PhysX Mod Pack can be downloaded here.
Supports single GPU and NVIDIA SLI™ technology* on DirectX 9 and OpenGL.
Supports CUDA™ Technology.
Supports GPU overclocking and temperature monitoring by installing NVIDIA System Tools software.
Please read the release notes for more information on product support, features, and known compatibility issues.

 

Yes, sir! Getting much more PPD from my single graphics card than from those stupid SMPish WUs I am getting lately (project 3065 and 2665).

BTW: The GPU client works for me much better on Vista (x64 Home Premium SP1) than on XP (32 bit Pro SP1)- uses only around 6% of CPU on average compared to a hole core on XP and doesn't have any effect on the system speed... This actually sped up my transition to Vista considerably! Using 177.35 with modified .inf on both OSes.

EDIT:

Sorry, wasn't that far with reading when I had written my post...

Damn, seems that I am having similar issues, but only on Vista... Especially playing back video causes crashes. In case of XP everything is stable and good... Gonna try to watch something with SLI disabled myself. Do you use Vista?

 

I cheated.
I gave up on SLi 8800 GTX and got myself a 280GTX

yes its vista x64

 

First time poster here, just wanna share my results with everyone

Current Test Specs:


Besides this i've noticed some people having trouble with CPU Usage.. this was not the case for me.. FaH is using an avg of 9-10% and peaks up to 20% on 1 core in my system a dualcore 2.8GHz, and has no noteworthy memory usage.

With this system im getting 5748,00 ppd ..

 

My process to install and performance

downloaded and installed NVIDIADisplayWinVista(177_35)Int from http://www.nvidia.com/object/cuda_get.html
downloaded and installed Folding@home-GPU-systray-612b8 from http://foldingforum.org/viewtopic.php?f=43&t=3500

Copied cudart.dll from C:\CUDA\bin to C:\Program Files\Folding@home\Folding@home-gpu (without doing this the client crashes unable to find cudart.dll)

CPU Load 5-10%
Performance is ~2400 iter/s

 

I beleve the 41's dont support G80 without a modified inf. For my setup this is the current software. If I had a G9x GPU I would

 

you have a G9x GPU, you wouldnt need a modified inf. in your experience with a g9x it may provide a performance boost and or stability but have you actually run it on a G80?

 

Not at all, all im saying is my current setup requires no modified inf for drivers or modification of the client. So id rather use the drivers that nvidia has blessed to use cuda with G80. afaik the newer set of drivers work without a modified inf for g9x cards but not G80 cards. That being said I searched and read these forums before starting this process. found this to be the best software setup for my current configuration

apparently the 41's require a modified inf for anything under the 200 series of gpu but include support for PHYSX and CUDA for all G9x but not G8x. That being said im not interested in the physx just the cuda and that is supported by the 35's with no modification

 

It is possible that that the 41's allow Cuda/physx on G9x and Cuda on G9x/G80 for the physx is currently not supported on G80. That being said the 35's remove the physx from the equasion and support cuda on G80 I was never a big fan of modifying inf's. And like my 1st post very simple to get running.

 

Playing around with fahmon and fahspy and apparently with my stock 8800GTX I am getting 4300PPD out of it

 

Code:

[14:33:26] + Attempting to send results
[14:33:26] - Couldn't send HTTP request to server
[14:33:26]   (Got status 503)
[14:33:26] + Could not connect to Work Server (results)
[14:33:26]     (171.64.122.76:8080)
[14:33:26]   Could not transmit unit 05 to Collection server; keeping in queue.
[14:33:31] + Attempting to get work packet
[14:33:31] - Connecting to assignment server
[14:33:32] + No appropriate work server was available; will try again in a bit.
[14:33:32] + Couldn't get work instructions.
[14:33:32] - Attempt #2  to get work failed, and no other work to do.
             Waiting before retry.

Anyone else getting this? Or is there a way to fix this?

Would really like to keep folding!

Or with all the GPU's folding now, they can't keep up with work for us to do

powerspec