| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: OpenPandemics - COVID-19 Project
Thread: Mandatory "Request for Project to be Ported to GPU" Thread |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 133
|
|
| Author |
|
|
Aurum
Master Cruncher The Great Basin Joined: Dec 24, 2017 Post Count: 2391 Status: Offline Project Badges: 
|
Very interesting, but they could just ditch the CPU version and do the work on the GPUs. There would be so many crunchers jumping on it that it would go faster than that justified by the present delay. And it would leave the CPU cores free for other projects. Look at this list and ask yourself do all these corporate type whales have CUDA 11 GPUs on their computers?https://www.worldcommunitygrid.org/stat/viewStatsByMemberY.do?sort=points Trying to shove both CPU & GPU down the same pipe means we completely lose our ability to control how our clients can run OpenPandemics WUs. E.g., <max_concurrent> and <project_max_concurrent> would be rendered useless. <app>Now it controls both GPU & CPU WUs. In any scenario I can think of one needs many more CPU WUs than GPU WUs. If anyone has read the new IBM WCG Terms & Conditions you'll see they're most onerous Ts & Cs BOINC has ever seen. A real turn to the dark side. I doubt IBM cares one bit what we philanthropists think.   ...KRI please cancel all shadow-banning |
||
|
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 445 Status: Offline Project Badges: 
|
I can do SOME of the work for converting an application to using a GPU, but only if someone else first separates the sections I can work on from those I cannot.
Assume that the sections I work on need to be written in either C or Fortran, and the GPU conversion can use CUDA but not OpenCL. No part of my work can require the use of Linux. I'll need the source code for both the the application and for the tests that check how well it works. I'm still looking for an online course in OpenCL, which could later allow me to do similar work with C++ and OpenCL I'll be working under Windows 10 only unless I find a way to reduce the power requirements enough to add another computer, which might allow me to learn Linux. |
||
|
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges: 
|
Very interesting, but they could just ditch the CPU version and do the work on the GPUs. There would be so many crunchers jumping on it that it would go faster than that justified by the present delay. And it would leave the CPU cores free for other projects. Sounds insane to me! The CPU one is done and working! So that's not keeping anyone from working on a GPU version! |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
Look at this list and ask yourself do all these corporate type whales have CUDA 11 GPUs on their computers? But they are doing OpenCL. In either case, the GPU would be probably at least 20 times faster than the CPU, from the ones I ran on Quarantine@Home. So the corporate whales become more or less irrelevant. They can use their CPU cores elsewhere. A lot more crunchers who are looking for GPU work would jump in to take their place. I hope it is CUDA 11. That would be best. |
||
|
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges:
|
Look at this list and ask yourself do all these corporate type whales have CUDA 11 GPUs on their computers? But they are doing OpenCL. In either case, the GPU would be probably at least 20 times faster than the CPU, from the ones I ran on Quarantine@Home. So the corporate whales become more or less irrelevant. They can use their CPU cores elsewhere. A lot more crunchers who are looking for GPU work would jump in to take their place. I hope it is CUDA 11. That would be best. Now, the cat is out of the bag! You don't want a GPU version! You want a GPU version that run only one type of GPU! |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
I hope it is CUDA 11. That would be best. Now, the cat is out of the bag! You don't want a GPU version! You want a GPU version that run only one type of GPU! I said that CUDA 11 would be best from the beginning, when I first quoted my Quarantine@Home performance levels. But OpenCL is OK if that is all they can do. However, it looks like they are delaying even that for compatibility purposes. It could be the worst of all worlds. They are coming a bit late to the game, and may have trouble being scientifically-relevant to begin with. Slowing down the process just makes them more slowly irrelevant. |
||
|
|
Aurum
Master Cruncher The Great Basin Joined: Dec 24, 2017 Post Count: 2391 Status: Offline Project Badges:
|
Now, the cat is out of the bag! You don't want a GPU version! You want a GPU version that run only one type of GPU! Nonsense. It could be CUDA or OpenCL. I don't even know the difference. I hope it runs on a wide range of high end GPUs. My point is that office PCs in companies tend to have very weak GPUs, e.g. GTX 260 or even use the graphics in a CPU.But to have a CPU project that is working great and then waste months just to force GPU WUs to run on either a CPU or GPU is a terrible waste of time. I think if they would've split CPU & GPU we'd currently have both running. I also think CPUs make more accurate calculations than GPUs. Perhaps find hits using GPUs that get within their desired bonding range and rerun those high scoring hits using CPUs. Or they can diverge and do something very different. ...KRI please cancel all shadow-banning[Edit 1 times, last edit by Aurum420 at Aug 14, 2020 2:13:43 PM] |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
I also think CPUs make more accurate calculations than GPUs. Perhaps find hits using GPUs that get within their desired bonding range and rerun those high scoring hits using CPUs. Or they can diverge and do something very different. True. That is probably at the heart of the problem they are having. It may not be solvable, except by splitting the work. Maybe they should tell us something? Though keeping it all a secret may be a good strategy too. |
||
|
|
supdood
Senior Cruncher USA Joined: Aug 6, 2015 Post Count: 333 Status: Offline Project Badges: 
|
In either case, the GPU would be probably at least 20 times faster than the CPU, from the ones I ran on Quarantine@Home. What hardware combination are you judging this by? And do you mean that you were getting 20 times more GPU results compared to a single CPU thread or compared to all of a CPU's threads? |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
In either case, the GPU would be probably at least 20 times faster than the CPU, from the ones I ran on Quarantine@Home. What hardware combination are you judging this by? And do you mean that you were getting 20 times more GPU results compared to a single CPU thread or compared to all of a CPU's threads?That was a GTX 1060, though I also tried a GTX 1070, which was correspondingly faster (on Ubuntu 18.04, since the GPU runs only on Linux). The CPU was an i7-4471 (Windows 7). I don't recall how many cores or if you had any option. You can give it a try easily enough if you can find the link. I have deleted it. I do recall that the GPU ran cool, at no more than 2/3 of it TDP rating. That is really how I compare compute efficiency. PS - I found that I still have the Windows program installed. It runs on a single core, and I don't see a way to change it. I will let it run for a while. PPS- I found the link. It would be best for you to try it and make the comparisons under whatever conditions you wish. https://quarantine.infino.me/ The GPU version on the 1060 averaged about 45 seconds as I recall. [Edit 4 times, last edit by Jim1348 at Aug 14, 2020 5:02:25 PM] |
||
|
|
|