Message boards :
Graphics cards (GPUs) :
Right software version??
Message board moderation
| Author | Message |
|---|---|
|
Send message Joined: 10 Mar 09 Posts: 1 Credit: 0 RAC: 0 Level ![]() Scientific publications
|
"Message from server: Cant use cuda app for full-atom molecular dynamics: your cuda device has the wrong software version" I have been trying to get this to work for some time now.. but thats all i get, so where do you think I go wrong? I have 9950 phenom, 4GB ram and 2 x 8800gts 640mb (G80 core) (sli off currently, tryed turning on/off and rebooting with no effect[to gpugrid that is, Sli is working fine]) on vista x64 sp1, now running forceware 182.08 (also tried 180.43(beta),180.48(whql), 181.20(whql),181.22(whql) and 182.06(whql) with no success) and boinc 6.4.7 x86-64. protected mode is NOT selected and Boinc manager does find cuda devices ("CUDA devices found"). Any help please? |
|
Send message Joined: 8 Sep 08 Posts: 63 Credit: 1,696,957,181 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Unfortunately, it is your hardware that causes the problem. G80 chips do not give the required CUDA 1.1 or better capability to run GPUGRID. From the homepage you can normally access a list with compatible videocards, where the 8800GTS/340-620 are listed as not compatible. But at the moment of writing this page seems to be unavailable. Maybe try again later. Kind regards Alain |
Bender10Send message Joined: 3 Dec 07 Posts: 167 Credit: 8,368,897 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]()
|
|
|
Send message Joined: 30 Oct 08 Posts: 47 Credit: 669,991,028 RAC: 169 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Now you can crunch also "The lattice" BOINC project http://boinc.umiacs.umd.edu/ |
|
Send message Joined: 24 Dec 08 Posts: 738 Credit: 200,909,904 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
You can still do GPU crunching with your 8800GTS (G80) on Seti, Folding@home, distributed.net and maybe Milkyway... How would that work? Its not cuda capable according to the comments above and nvidia's web site. Don't folding, distributed, milkyway and lattice use cuda? BOINC blog |
Stefan LedwinaSend message Joined: 16 Jul 07 Posts: 464 Credit: 298,573,998 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
It is CUDA capable, but only has computing capability 1.0 hence it won't run on GPUGRID but should work with SETI and Folding@home. Milkyway doesn't have a CUDA app as of now... pixelicious.at - my little photoblog |
|
Send message Joined: 21 Oct 08 Posts: 144 Credit: 2,973,555 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]()
|
The G80 chip is CUDA capable. It has compute capability 1.0; GPUGRID requires compute capability 1.1 (I believe because it needs the atomic functions coding). Compute capability 1.0 is not a problem for other projects as I understand it. Edit: too fast for me, Time Bandit :) |
EdboardSend message Joined: 24 Sep 08 Posts: 72 Credit: 12,410,275 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
I had a 8800GTX (G80) and I can confirm that it worked with Folding and with SETI. |
Paul D. BuckSend message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Now you can crunch also "The lattice" BOINC project A couple of points ... * The number of tasks issued for CUDA is limited. * This is an Alpha class test * These tasks can "Hang" meaning that you now have a task occupying your GPU doing nothing real fast. * The tasks are Garli meaning that they are the long running tasks of the project ....(Which makes sense, but could mean that the GPU will be running for several days on the one task, even longer than here at GPU Grid. The early tasks were set to run for a very short time before ending, the live tasks have not been run so no real run time on the GPU has been established, CPU run times even on fast systems is upwards of 200 hours with no trickles, and at least a 10% failure rate (my experience)). * You cannot opt in, or out of these Garli tasks. * There is no opt in for only GPU tasks |
©2025 Universitat Pompeu Fabra