Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 8
|
![]() |
Author |
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello all,
I have been using BOINC for quite a bit of time now and just "upgraded" to the latest version (6.12.34) because I thought the previous version did not support GPU computing (will read the documentation better next time!). Now it seems that BOINC can no longer connect to WCG. I get these errors in the log:
I am using Slackware 13.37-current, libcurl 7.21.4 with SSL support (previous BOINC was not complaining) From the BOINC log:
Other questions are in regard to GPU computing. Like you can see above, my laptop has 2 GeGorce 8700M GT cards. I'd like to use them as much as I can to crunch numbers, but I am confused as to how it is accomplished. I am currently using my GPU's to help in other applications with the CUDA implementation. I have the Nvidia dev-driver installed and the CUDA SDK. Other cuda enabled applications are using my GPU's for processing and I am wondering if BOINC really uses CUDA for GPU processing? GPUGrid seems really oriented toward Nvidia cards but does not have 32bit version for linux, which make it impossible for me to use it. Last question: why are projects oriented toward specific GPU processing technologies such as OpenCL, CUDA or CAL (AMD/ATI)? Does it mean that I will not be able to participate in certain project just because I use Nvidia cards? Looking forward for answers! Thanks !! |
||
|
Coleslaw
Veteran Cruncher USA Joined: Mar 29, 2007 Post Count: 1343 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Not sure if it is a similar problem, but 6.12.34 seems to have some difficulties other versions don't. There is another thread mentioning troubles with it. https://secure.worldcommunitygrid.org/forums/wcg/viewthread_thread,31757
----------------------------------------I would suggest downgrading it to 6.12.33 or go with the version recommended by WCG which is 6.10.58. You will also find that WCG does not yet have GPU computing projects but we are constantly told that HCC has one in the works. As far as your question regarding OpenCL, CUDA, or CAL, you will find just about all GPU projects supporting CUDA, but only a hand full supporting OpenCL or CAL at the moment. So, if you have nVidia, you are more likely to find uses at this time. You can find a list here: http://boinc.berkeley.edu/wiki/GPU_computing Edit: I forgot to also answer...yes some of the older BOINC clients do support GPU. So, going with WCG's recommendation will still use both of your cards. ![]() ![]() ![]() ![]() [Edit 1 times, last edit by Coleslaw at Oct 2, 2011 8:04:50 PM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Coleslaw, thanks for the reply!
----------------------------------------I am going to downgrade to 6.10.58 with the hope the SSL problem will be gone and that both of my GeForce's will be harvested to the max. I am still not sure how the application (BOINC) can harvest the most out of the GPU's unless there were a CUDA enabled version of BOINC... Sorry for my ignorance, I just started a few days ago to play around with CUDA, and so far, to allow a specific application to use the GPU's I need to recompile the software with the CUDA enabled in the code (nvcc) and I get a CUDA capable application instead of a standard C++ application that will only use the CPU. With distributed computing like BOINC and WCG, what decides if the GPU will be used? Is it BOINC or the specific project? For example, if its the project, does it mean that HCC (for example) will send me CUDA enabled packets to crunch, instead of standard packets? Thanks!!! [Edit 1 times, last edit by Former Member at Oct 3, 2011 4:45:58 PM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello lpallard,
The specific project program has to be compiled to run on a GPU. Right now, none of the WCG projects are designed for GPUs. So you have to look for non-WCG projects to use a GPU. Distributed Computing ( http://www.distributedcomputing.info/projects.html ) lists the current distributed projects, both BOINC and non-BOINC) Lawrence |
||
|
Coleslaw
Veteran Cruncher USA Joined: Mar 29, 2007 Post Count: 1343 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
As lawrencehardin said, it is up to the individual projects to come up with GPU/CUDA work units. So WCG itself doesn't make them, but the scientists of each project. If you follow the second link I posted, it tells you which projects support GPU which can be run alongside WCG with the same BOINC software. Each project also has its own requirements, so for example...you will find GPUGrid to have much higher requirements (high end cards) then many of the others. I think Moo! has the least requirements considering they only need a 64MB CUDA capable card.
----------------------------------------Edit: Oh and welcome to the forums. I think you will find WCG one of the most helpful and informative out there. ![]() ![]() ![]() ![]() [Edit 1 times, last edit by Coleslaw at Oct 3, 2011 8:31:39 PM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Coleslaw, thanks for the welcome!
I've stumbled upon a PDF document that is discussing about the Help Conquer Cancer project and CUDA. Apparently the Ontario Cancer Institute were (are?) looking to release a CUDA enabled version of HCC so owners of CUDA cards could crunch faster... Any hints as to when HCC could be 100% CUDA enabled? https://encrypted.google.com/url?sa=t&sou...1QOcrsjMbjBNw&cad=rja |
||
|
Coleslaw
Veteran Cruncher USA Joined: Mar 29, 2007 Post Count: 1343 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
WCG has been telling us for like 3 years that they are "planning" on bringing GPU work and it has been discussed a lot within the forums that HCC would be more then likely the first project to bring us to the GPU realm. Unfortunately, you will find that WCG never gives exact dates, but rather lets you see it when it gets here with almost all new arrivals and/or changes. Your best bet would be to either not purchase until they have released at least BETA wu's or to just dedicate your GPU's elsewhere until WCG is ready. BETA wu's are usually your best indicator or when something new is happening "soon". So, watch those forums closely. Also, there is plans for a HCC2 some time after HCC finishes. So, it is theoretically possible that we will finish HCC before GPU is ever released. So, like I said..."wait and see".
----------------------------------------Edit: I have also seen the .pdf of them planning CUDA. However, it seems to be the common belief that WCG is more then likely pushing for OpenCL. That way it will cover the widest variety of cards including ATI/AMD and hopefully with newer versions of BOINC Intel. Sandy Bridge GPU and newer can theoretically work if BOINC decides to support them. Supposedly newer versions of BOINC will have better GPU support built into them. Edit2: I also have a thread for people to post what GPU's they would bring here if/when WCG has a project that supports them. https://secure.worldcommunitygrid.org/forums/wcg/viewthread_thread,31724 ![]() ![]() ![]() ![]() [Edit 2 times, last edit by Coleslaw at Oct 6, 2011 2:41:36 AM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Yeah I'll keep an eye here!
Right now I have a core 2 duo machine with 2 geforce 8700m gt video adapters (not used because wcg doesnt support cuda yet) and a quad core amd athlon 64 bit with a geforce GT 430 but my goal , if I get strong gpu involvement is to build a quad gpu ( probably geforce 580gtx) with a six core phenom black edition with 8+ gb ram... no real cuda support this is not gonna happen... Boinc should also run on cuda for optimized parallel processing... |
||
|
|
![]() |