| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: CUDA device help |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 30
|
|
| Author |
|
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges: 
|
Hi, I recently upgraded the graphics card on one of my machines with a new NVIDIA 8400GS - it's a CUDA capable device. Sorry but your card is not supported at GPUGrid. I just thought I'd save you some grief if you went there. You can go here to find supported cards. Basicly, cards in green work, cards in black work but cannot return a wu within their deadline and cards in red won't work. Just thought I'd let you know ... Cheers  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
GPU tasks also use CPU resources. I understand this. I'm asking about the nature of the BOINC client, so let me rephrase. Let's say I have a quad core system running "standard" tasks; obviously, four of them will be working in parallel. Later on, I decide to add a GPU-based task to these already running tasks. Will the BOINC client make exclusive to a CPU thread the GPU task (ie, one of the original four is placed on hold)? Or will it simply let the other tasks continue as usual and give CPU resources to the GPU task only as needed? [Edit 3 times, last edit by Former Member at Apr 15, 2009 8:57:09 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
A quad with 1 CUDA capable card will run 5 tasks, a quad with 3 CUDA capable cards will run 7 tasks, so is the theory. From version 6.8 the client will also be able to handle ATI cards of correct spec, theory too.
----------------------------------------
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Apr 15, 2009 9:09:53 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
A quad with 1 CUDA capable card will run 5 tasks, a quad with 3 CUDA capable cards will run 7 tasks, so is the theory. From version 6.8 the client will also be able to handle ATI cards of correct spec, theory too. Interesting and thanks! Is there a FAQ you guys keep updated with this information as it becomes relevant? |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
No, but someone is looking for contributors. Roll up the sleeves, read the 5000 posts on these forums on GPU processing and summarize the culmination of your findings and conclusions.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges:
|
No, but someone is looking for contributors. Roll up the sleeves, read the 5000 posts on these forums on GPU processing and summarize the culmination of your findings and conclusions.  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
No, but someone is looking for contributors. Roll up the sleeves, read the 5000 posts on these forums on GPU processing and summarize the culmination of your findings and conclusions. I find it so strange that people come to WCG, where there are no CUDA/GPU projects, to find out about these projects. Our CAs seem to know more about a technology that they do not support than do the projects that have support. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
No, but someone is looking for contributors. Roll up the sleeves, read the 5000 posts on these forums on GPU processing and summarize the culmination of your findings and conclusions. Sounds like a project! I'm game. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
|
||
|
|
BertrandRussel
Cruncher Joined: Nov 24, 2006 Post Count: 30 Status: Offline Project Badges: 
|
Thanks for stating this more precisely. Before buying the card I checked Nvidia's own list which is very general at:
http://www.nvidia.com/object/cuda_learn_products.html Fortunately the card's sole purpose was not to CUDA-crunch but to fit into a HP SFF desktop that is not easy to restock due to the format constraint. The card is low profile and passively cooled which makes it a good find. I'll stay with the WCG assuming that time will bring more improvements in efficiency... |
|||
|
||||
|
|