| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: OpenPandemics - COVID-19 Project
Thread: OpenPandemics GPU is LIVE! |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 77
|
|
| Author |
|
|
Jorlin
Advanced Cruncher Deutschland Joined: Jan 22, 2020 Post Count: 89 Status: Offline Project Badges: 
|
I have NVIDIA and Intel jobs running at the same time.
----------------------------------------Since i haven't attached a monitor on my iGPU i had to attach a dummy plug to it to get it active (just enabling it in the BIOS isn't enough, there has to be something attached to it).  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hej Jorlin, und wie haste den Stecker konfiguriert? Brücke oder R eingelötet?
|
||
|
|
Jorlin
Advanced Cruncher Deutschland Joined: Jan 22, 2020 Post Count: 89 Status: Offline Project Badges:
|
Hej Jorlin, und wie haste den Stecker konfiguriert? Brücke oder R eingelötet? Fertig bei Amazon bestellt. Solche Lötarbeiten sind nix mehr für den alten Mann  |
||
|
|
hnapel
Advanced Cruncher Netherlands Joined: Nov 17, 2004 Post Count: 82 Status: Offline Project Badges: 
|
To all the OpenCL vs. CUDA commenters on NVIDIA: this project uses Autodock, it can potentially also leverage CUDA also
something like https://github.com/ccsb-scripps/AutoDock-GPU What they are doing is baby steps, first they test / build support for opencl and this already leads to dramatic performance improvements such that there's not even sufficient work for the donors to crunch. With opencl they have a blanket solution for both AMD and NVIDIA even if it leads to suboptimal performance on NVIDIA for all the reasons mentioned. The GPU is now production for not even a week and what a complaints! The important thing is that we are in a position to give the scientists all the power they need to combat the pandemic. It has been published by the project the GPU support will allow more complex work to be processed (larger molecules etc.) so give them some time to come up with ideas to utilize this power in a meaningful way. I'm sure all the ideas and suggestions did pop up in the project and scientists mind already before you could write them here. Software development is an intricate process. For example take the validator, when people started to make comments they learned indeed it was a long and convoluted thought-out process and nothing is ever simple. |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
I have mostly Nvidia cards and normally would love to see a CUDA implementation. But I have just compared my GTX 1060 to an RX 570 (both Ubuntu 20.04.2), and the Nvidia is over twice as fast, and more importantly for me, over twice as power efficient.
Normally I expect them to be about the same on OpenCl projects, even though the RX 570 is about a generation behind in technology. But it usually makes up for it with improved OpenCl performance. I think the scientists have done a good job optimizing for OpenCl for Nvidia, and doubt that they could squeeze much more out of it. More importantly, non-scientists don't understand the need for consistent results. They probably won't touch it unless they can see both large gains and consistent results from all versions, which will be hard to test for. I would not hold my breath for CUDA. The further optimizations for the OpenCl work units are much more likely I think. |
||
|
|
Keith Myers
Senior Cruncher USA Joined: Apr 6, 2021 Post Count: 193 Status: Offline Project Badges: 
|
Considering the developers in their github Autodock repo proclaim:
----------------------------------------The Cuda version is currently even faster than the OpenCL version. I can only surmise the project managers have decided to use the OpenCL application instead of the CUDA application simply to have more gpus of all types working on the project. But really, there is no reason why you couldn't have both applications available.   A proud member of the OFA (Old Farts Association) |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
But really, there is no reason why you couldn't have both applications available. Yes there is (or are). You would need to test for consistency between not only the GPU versions, but with the CPU version as well. They are already supporting three GPU hardware types (Intel, Nvidia and AMD) plus two CPU hardware types (AMD and Intel), and maybe ARM for all I know. Also, there are two or three OS's for each. The last thing they need is to muddy the waters further. |
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
But really, there is no reason why you couldn't have both applications available. Just because you see no reason doesn't mean there are none. Seems to me the time and effort involved in making the CUDA app run here is not worth it for only 1 make of card. Would add more work for the techs here also. Can we move on now? This horse will not come back to life. 
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.
  |
||
|
|
Ian-n-Steve C.
Senior Cruncher United States Joined: May 15, 2020 Post Count: 180 Status: Offline Project Badges: 
|
Are different device types validating against each other? Can a job run on Nvidia validate against AMD? Or any other combination. Or are WUs only being distributed to same-type devices, Nvidia/Nvidia AMD/AMD, Intel/Intel
----------------------------------------Can one of the project admins or someone who’s seen a mix-match cross validate confirm? I looked at several of my WUs, but only saw my Nvidia cards paired up with wingmen with other Nvidia cards. But I’ll admit I didn’t look at too many so it could be coincidence. Ideally if the app is providing true results, any device should be able to validate against any device. Even CPU<->GPU if the WUs weren’t different (which I know they are).  EPYC 7V12 / [5] RTX A4000 EPYC 7B12 / [5] RTX 3080Ti + [2] RTX 2080Ti EPYC 7B12 / [6] RTX 3070Ti + [2] RTX 3060 [2] EPYC 7642 / [2] RTX 2080Ti |
||
|
|
Richard Haselgrove
Senior Cruncher United Kingdom Joined: Feb 19, 2021 Post Count: 360 Status: Offline Project Badges:
|
Are different device types validating against each other? Can a job run on Nvidia validate against AMD? Or any other combination. Or are WUs only being distributed to same-type devices, Nvidia/Nvidia AMD/AMD, Intel/Intel Keith answered that one in https://www.worldcommunitygrid.org/forums/wcg...ad,43311_offset,70#654827But also read the follow-up a couple of posts later in the same thread. We don't have details of those other (non-BOINC) validation tests. |
||
|
|
|