| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: GPU Support Forum
Thread: GPU Support: Post Your Problems/Questions Here |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 195
|
|
| Author |
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
Just a quick question or two: Are the GPU WU's able to be crunched on linux (Running ubuntu) and if so where can I get the drivers? The GPU beta test is currently for Windows only. 
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
  |
||
|
|
GeraldRube
Master Cruncher United States Joined: Nov 20, 2004 Post Count: 2153 Status: Offline Project Badges: 
|
Hi Gerald I've just joined in this GPU Malarkey. I'm running GPUGrid flat out However the Beta's from WCG (last time out anyway) have a really short return time As a result they kick out the GPUGrid WU's as they have a higher urgency  and take over your PC's GPU(s) until they run out.Which for me, like you, with a great affection for WCG....is ideal. (I joined the moonlighting  MOT team over there too ) Dave Dave that is great!!   |
||
|
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges:
|
Seems that a GTX680 outperformed a GTX580 at SETI by ~16%. This Leaves some hope for non OpenCL projects; +16% performance for 195W rather than 244W TDP is a 45% improvement in performance per Watt. Note however that SETI use CUDA for NVidia and WCG uses OpenCL! Around Aug 2012, NVidia is thought to be releasing a GTX685, which will be GK110 (rather than GK104, or GK106 which we might see at entry/mid range levels between now and then). The GK110 will probably be for the TESLA, Quadro, and possibly enthusiast range GeForce cards, but will offer NVidia's implementation of high end compute performance. Until then it's likely that the AMD 7970 will reign supreme when it comes to OpenCL and double precision (FP64), and will thus be the best card for here. Hypernova, Existing TESLA's will work well here. About as well as a high end GTX500. However, I would not recommend buying one as an HD7970 would perform the GPU part of the WCG's HCC for GPU task about 4 or 5 times as quickly. For perspective, it doesn't mean that these are bad cards, just that the HD7970 is excellent at OpenCL thanks for the info skgiven. I do not plan for a Tesla card which is expensive by the gaming cards standard. HCC GPU will surely not be ready as a mainstream GPU project by August I suppose, so there is plenty of time to wait and see. But if in the end the AMD cards are the best ones for WCG, then I may do a deviation from my personal standard, and make an exception and use a bunch of AMD cards. I see another issue too. If HCC GPU WUs are processed in such a short time, then the WCG servers will have to supply and process in return an enormous amount of WU's. This may hit badly bandwith. I wonder if WCG has also integrated all this. There is also the issue that to make the project attractive it has to last over some time, say one year at least, otherwise nobody will bother investing in GPU boards to crunch.  |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: 
|
I see another issue too. If HCC GPU WUs are processed in such a short time, then the WCG servers will have to supply and process in return an enormous amount of WU's. This may hit badly bandwith. I wonder if WCG has also integrated all this. There is also the issue that to make the project attractive it has to last over some time, say one year at least, otherwise nobody will bother investing in GPU boards to crunch. I assume that HCC will add more work beyond the 30-days they have for the CPUs at the moment, but even so any new work would probably go quickly. And it appears possible that no other project will be ready for GPUs for quite a while. Therefore, this project may be just an oasis (or mirage) in a desert. I wouldn't buy any new card for it until I see how/if the work holds up. |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
I see another issue too. If HCC GPU WUs are processed in such a short time, then the WCG servers will have to supply and process in return an enormous amount of WU's. This may hit badly bandwith. I wonder if WCG has also integrated all this. It was mentioned they'll going to supply multiple "normal" HCC-wu's in each HCC-GPU-wu, meaning it will be similar to how some of the sub-projects already works with having multiple jobs in them. Also, for GPU there won't be any checkpointing in the middle of each wu-part, only between wu-parts, this is also similar to how some of the sub-projects works. Since the GPU can finish a "normal" HCC-wu in only a few minutes, not having checkpointing in the middle of a wu-part shouldn't be a big problem. So then it comes to server-load it shouldn't be a problem. The interesting part would be if each wu-part will use just as much CPU-time as the current beta-wu's, or if it's possible to combine things so uses significantly less CPU-time between each GPU-part.  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The checkpointing part has already been replied to by amrstrdj, only between the images packaged in 1 HCC GPU Task.
----------------------------------------In a long year gone, there were alpha tests with 5 in a task, i.e. it's not novel territory. Supposing there can be 100K per day processed via the GPU route [current total production of HCC 350-400K daily], the servers have effectively proven in the last week to be able to handle the higher volumes. We're hitting 1.25 million now with 120K shorties of HCMD2 included. If the system would not be able to handle, there is the mentioned implemented scalability, or the fallback to throttle... the whole of the grid will not be allowed to go under because of a single science feed tempest, there's no doubt about that. --//-- [Edit 1 times, last edit by Former Member at Mar 29, 2012 9:48:47 AM] |
||
|
|
piropeople13
Cruncher Joined: Jun 9, 2007 Post Count: 13 Status: Offline Project Badges:
|
I have GPUs enabled to run always in preferences, how do I know if it is working.
i7 3930k 2 x GTX 555 The temps of one of the two gpus is elevated but bonic only shows 12 thread running which is the normal amount for just cpu How do I get the other GPU running toom or check if it is working at all? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Gpu still in beta phase. More may come this week. Hopefully anyway.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello piropeople13,
WCG does not have any GPU project running now. We have an occasional Beta test of a GPU project. We have not run a test in several weeks. Keep looking for an announcement in the Beta Test Announcements forum for the next test. Lawrence |
||
|
|
piropeople13
Cruncher Joined: Jun 9, 2007 Post Count: 13 Status: Offline Project Badges:
|
Gpu still in beta phase. More may come this week. Hopefully anyway. Hello piropeople13, WCG does not have any GPU project running now. We have an occasional Beta test of a GPU project. We have not run a test in several weeks. Keep looking for an announcement in the Beta Test Announcements forum for the next test. Lawrence Thanks for the info. I am excited for them to get some work out. F@H hasn't whitelisted the GTX 555s yet on the v7 client and it bothers me that they are just idling all the time. |
||
|
|
|