| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: OpenPandemics - COVID-19 Project
Thread: Well it's about time for GPU work for years of delay. |
| No member browsing this thread |
|
Thread Status: Locked Total posts in this thread: 136
|
|
| Author |
|
|
[P3D]Magiceye04
Cruncher Joined: Jul 5, 2008 Post Count: 38 Status: Offline Project Badges: 
|
Just wait until the first WUs are available.
----------------------------------------Some projects run better in AMD, some on Nvidia. Some run also ONLY on AMD or Nvidia. For Folding@Home my Gefoce1070ti is running at the same performance-level like my Radeon VII. Some years ago GPUs were really much faster then CPUs. This was the time of dual-core or Quad-Core-CPUs. But now everyone can have a 128-Thread-CPU-Monster at Home. And this is hard to beat for a GPU. [Edit 1 times, last edit by magiceye04 at Sep 25, 2020 8:01:37 PM] |
||
|
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges: 
|
Just wait until the first WUs are available. Some projects run better in AMD, some on Nvidia. Some run also ONLY on AMD or Nvidia. For Folding@Home my Gefoce1070ti is running at the same performance-level like my Radeon VII. Some years ago GPUs were really much faster then CPUs. This was the time of dual-core or Quad-Core-CPUs. But now everyone can have a 128-Thread-CPU-Monster at Home. And this is hard to beat for a GPU. I haven't been out for awhile, but I don't think I can get one of those at Walmart or Best Buy. |
||
|
|
cristipurdel
Senior Cruncher Joined: Dec 13, 2008 Post Count: 158 Status: Offline Project Badges: 
|
---------------------------------------- |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12432 Status: Recently Active Project Badges:
|
The problem with running Einstein on GPU is that most of their projects require a CPU thread as well as GPU. However, their GRP4 does not. With GRP4 you can run it on GPU and all CPU threads on something else. Nominally it says it requires 0.5 CPU thread as well but I think that must be nominal because none of the CPU threads seem to be slowed by it.
Mike |
||
|
|
hnapel
Advanced Cruncher Netherlands Joined: Nov 17, 2004 Post Count: 82 Status: Offline Project Badges: 
|
What Openpandemics plans on WCG is fairly unique, namely trying to process the same type of WU's on both CPU and GPU, on F@H there are distinct project numbers that run either on GPU or CPU, for example the latest 'sprints' for the Covid Moonshot run on GPU only. If WCG starts with the GPU jobs I will be able to do some comparisons (apples & pears come to mind...) Also note so far CPU work on WCG is single threaded, so if you have more cores you simply run more individual WUs in parallel, it could be more efficient to parallelize individual jobs also. The latest Intel processors have on-board GPU also which sits idle if you have a addon graphics card, I would try to use the power of those too, there's a lot of spare resources there I think.
|
||
|
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges:
|
The problem with running Einstein on GPU is that most of their projects require a CPU thread as well as GPU. However, their GRP4 does not. With GRP4 you can run it on GPU and all CPU threads on something else. Nominally it says it requires 0.5 CPU thread as well but I think that must be nominal because none of the CPU threads seem to be slowed by it. Mike For me anything less than 1 (ie. 0.9 CPU), doesn't take a thread slot. |
||
|
|
Aurum
Master Cruncher The Great Basin Joined: Dec 24, 2017 Post Count: 2387 Status: Offline Project Badges: 
|
For me anything less than 1 (ie. 0.9 CPU), doesn't take a thread slot. But it does. If you don't budget a CPU it must share with other CPUs and other WUs wait to get access to a CPU. It's easy to watch on Linux with System Monitor.  ...KRI please cancel all shadow-banning[Edit 1 times, last edit by Aurum420 at Sep 27, 2020 2:24:49 PM] |
||
|
|
Aurum
Master Cruncher The Great Basin Joined: Dec 24, 2017 Post Count: 2387 Status: Offline Project Badges:
|
The latest Intel processors have on-board GPU also which sits idle if you have a addon graphics card, I would try to use the power of those too, there's a lot of spare resources there I think. The "latest" Intel CPUs? I thought it was more like the oldest. I haven't had one of those since the nineties. What CPU model are you referring to? If one did utilize the onboard GPU how much of the CPU must be dedicated to support it?I guess they have their place in a laptop: https://www.tomshardware.com/features/amd-vs-intel-integrated-graphics ...KRI please cancel all shadow-banning[Edit 1 times, last edit by Aurum420 at Sep 27, 2020 2:24:19 PM] |
||
|
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges:
|
The latest Intel processors have on-board GPU also which sits idle if you have a addon graphics card, I would try to use the power of those too, there's a lot of spare resources there I think. The "latest" Intel CPUs? I thought it was more like the oldest. I haven't had one of those since the nineties. What CPU model are you referring to? If one did utilize the onboard GPU how much of the CPU must be dedicated to support it?I guess they have their place in a laptop: https://www.tomshardware.com/features/amd-vs-intel-integrated-graphics Intel desktop CPUs have included integrated graphics for years...until this year some new CPUs no longer include integrated graphics. |
||
|
|
flynryan
Senior Cruncher United States Joined: Aug 15, 2006 Post Count: 235 Status: Offline Project Badges:
|
years ago GPUs were really much faster then CPUs. This was the time of dual-core or Quad-Core-CPUs. But now everyone can have a 128-Thread-CPU-Monster at Home. And this is hard to beat for a GPU. GPU performance has also gone up quite significantly, at least keeping pace with CPU's. The latest GPU's have around 10,000 "cores". Of course not all cores are created equal, but the top of the line GPU's have significantly more compute power than the high core count CPU's. For a rough idea, my Ryzen 3950x CPU with 16c/32t has around 3.8 Billion transistors. The newly released Nvidia RTX3090 has 28 Billion. Transistor count isn't everything but it gives some clue as to the capability of a well optimized system. |
||
|
|
|