| 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: 26
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I know is not likely that any project would do something for pre-OpenCL 1.0 specification but just in case I offer my 2 HD4670 GPUs (CF) if you ever need them. I have already enabled my self for GPU participation on both, projects & preferences, also set BOINC client to use GPU always. My GPU is fully OpenCL 1.0b (beta) compliant.
----------------------------------------If you feel the same post your old GPGPU capable GPUs. If you want to know you CPU/GPU capabilities download this free and great system analyzer: AIDA64 Cheers. [Edit 3 times, last edit by Former Member at Jul 17, 2012 1:37:57 PM] |
||
|
|
Coleslaw
Veteran Cruncher USA Joined: Mar 29, 2007 Post Count: 1343 Status: Offline Project Badges:
|
It is highly unlikely for anyone to even bother with trying to get the outdated cards to work. Especially here since there seems to be a strong reply of lack of funding as is. I would recommend you utilize these cards at other projects that still have apps that work on them such as PrimeGrid, Collatz, and Moo!. Those projects are not OpenCL and therefore still work. That is what I do with my 4350 and some of the even lower end onboard GPU's from AMD.
----------------------------------------Edit: You may also want to direct this discussion more towards this thread... https://secure.worldcommunitygrid.org/forums/wcg/viewthread_thread,31724 ![]() ![]() ![]() ![]() [Edit 1 times, last edit by Coleslaw at Jul 17, 2012 8:41:58 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Yeah I know about those projects and I don't care about them, I want medical projects to help with, I think that comes first before any other research. And yes, I already posted that is unlikely because of that, I made this thread so people could post their "outdated" cards here, so the community have an idea of the potential and maybe someone might do something about it.
|
||
|
|
kateiacy
Veteran Cruncher USA Joined: Jan 23, 2010 Post Count: 1027 Status: Offline Project Badges:
|
You might want to see whether you can run POEM (medical project) on your GPU. I am running it on an Nvidia ION, which I believe is only OpenCL 1.0 compatible. Like you, I'm happy to be able to run a medical project on whatever hardware I have to offer.
----------------------------------------![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thanks mate, but unfortunately POEM uses 1 full core for every GPU, so I prefer WCG to use my CPU cores. I have also searched for GPU projects and found other protein research oriented but all had a similar CPU core usage or needed 64-bit floating point GPUs. Well looks like I'm officially GPU "outdated", and I put it in quoting marks because each of my GPUs can give 960 GFLOPS so I think they're still useful and powerful enough to do good, imagine thousands of such cards mostly idle, well, it is what it is.
----------------------------------------[Edit 1 times, last edit by Former Member at Jul 18, 2012 3:14:34 AM] |
||
|
|
Coleslaw
Veteran Cruncher USA Joined: Mar 29, 2007 Post Count: 1343 Status: Offline Project Badges:
|
Well the other forum I linked is for all GPU's you have to bring to the table. Maybe scroll through and you will see some of the lists. I have posted many outdated GPU's on there. Also, any project using OpenCL is pretty much gonna use a core to a certain degree to feed it. It doesn't work like CUDA or CAL so much and therefore seems to be less efficient. However, that is a long theoretical debate that differs from each individual project. I have asked the techs here more then once why certain cards that "should" meet the prerequisites weren't included in the BETA's and pretty much the only people replying are the non-techs spouting the obvious. However, nothing official from the techs. I am aware of possible memory, bandwidth, etc.. issues. I have also requested in the suggestion forums for a non-compatability list which we actually got a reply that the techs took note of the request. So, for now we wait until they get something working outside of BETA and then hope they find the need to expand further. But, I wont hold my breath.
----------------------------------------![]() ![]() ![]() ![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
To Admins: This thread has nothing to do with the Betas and so should get moved to the "GPU Support Forum".
----------------------------------------[Edit 1 times, last edit by Former Member at Jul 18, 2012 5:46:45 AM] |
||
|
|
mikey
Veteran Cruncher Joined: May 10, 2009 Post Count: 826 Status: Offline Project Badges:
|
Thanks mate, but unfortunately POEM uses 1 full core for every GPU, so I prefer WCG to use my CPU cores. I have also searched for GPU projects and found other protein research oriented but all had a similar CPU core usage or needed 64-bit floating point GPUs. Well looks like I'm officially GPU "outdated", and I put it in quoting marks because each of my GPUs can give 960 GFLOPS so I think they're still useful and powerful enough to do good, imagine thousands of such cards mostly idle, well, it is what it is. NOTHING DEFINITIVE, but there is no reason to think WCG would use any less than just short of a full cpu when it releases it gpu software. More and more AMD projects are using the cpu more and more. Nvidia cards have been doing this for a long time and the AMD cards are just catching up. More cpu usage can mean the gpu is working harder as it needs more info being sent to and from it. Besides a gpu can do 10 TIMES the work of a cpu, so dedicating a cpu to it IS cost effective. It DOES however slow down cpu only projects! ![]() ![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I'll try to get into a GPU project but I think is a big mistake what protein GPU projects are doing: not making use of 32-bit floats GPUs and if OpenCL will use a lot of or a complete CPU Core per GPU then I think is not done properly at least on the software if the hardware allows it, because we have serial buses (makes them independent from each other and more dedicated to individual hardware devices) and DMA (Direct Memory Access) from devices and this should be the case of GPUs so GPUs can actually act as graphics co-processors as efficient as they can be. GPUs can do SOME things faster than CPUs, not everything so the efficient thing to do is that GPUs use 1% at most CPU time (I'd like it to be .1%) and accelerate protein research even more. If the hardware is the efficiency bottleneck then it should be redesigned, I think the IT industry have seen how much good CPU/GPU protein research for Humanity hence the design towards it, but we need to make it efficient as well since is the greatest thing to make an altruistic effort like this, so the better we make it the most altruistic it is and Humanity benefits more; just my thoughts.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The other side of the thinking of not tapping into the potential gains from also harnessing the GPU-computing output of a ton of old GPUs:
----------------------------------------1] The techSupport cost and difficulty of maintaining old GPUs most likely hosted by old hardware and old OSes (with each component being not designed to do GPU-computing) -- would not be negligible by any reasonable stretch of the imagination to say the least, and potentially nightmarish at worst. 2] The opportunity for exploitation of the power-efficiency, compute-efficiency, and over-all higher level of capability of mainstream GPU-cards (and designed with GPU-computing in mind) would not be maximized. ; ; edit1_2012.07.28Sa.1517.utc: ; -- readability improvements ; [Edit 1 times, last edit by Former Member at Jul 28, 2012 3:17:16 PM] |
||
|
|
|