| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: OpenPandemics - COVID-19 Project
Thread: OpenPandemic GPU work is running strange. |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 13
|
|
| Author |
|
|
Sandman192
Cruncher United States Joined: Apr 26, 2007 Post Count: 20 Status: Offline Project Badges: 
|
I notice that the GPU work for 2 of my GPUs not showing much GPU usage. On my 1080 Ti that runs at 54c to 80c on other projects. Open Pandemic is running it at 42c. They do finish but how when I'm seeing 2% GPU usage if that.
GPU usage from Task Manger and Precision X1. |
||
|
|
Ian-n-Steve C.
Senior Cruncher United States Joined: May 15, 2020 Post Count: 180 Status: Offline Project Badges: 
|
I notice that the GPU work for 2 of my GPUs not showing much GPU usage. On my 1080 Ti that runs at 54c to 80c on other projects. Open Pandemic is running it at 42c. They do finish but how when I'm seeing 2% GPU usage if that. GPU usage from Task Manger and Precision X1. Check with GPUz.  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 |
||
|
|
Keith Myers
Senior Cruncher USA Joined: Apr 6, 2021 Post Count: 193 Status: Offline Project Badges: 
|
Yes, the application is very poorly optimized. Only 3.8% utilization. If you want to see an analysis and graphs of how a gpu runs a task look at the Seti Orphans thread in Number Crunching where our Seti OpenCL developer investigates the application.
----------------------------------------https://setiathome.berkeley.edu/forum_thread.php?id=85283&postid=2073170   A proud member of the OFA (Old Farts Association) |
||
|
|
Jake1402
Senior Cruncher USA Joined: Dec 30, 2005 Post Count: 181 Status: Offline Project Badges: 
|
On my 2 win 10 boxes with AMD R9 270X cards running 2 tasks per card, my GPUz shows 99% utilization at 60c.
----------------------------------------
Join the Chicago-IL-USA team!
2 AMD FX 8320/AMD R9 270X/Win 10 2 AMD FX 8320/AMD RX 560/Linux Mint 20.3 (both computers DOA) Intel Pentium G240/Win 10 |
||
|
|
ericcui1
Cruncher Joined: Aug 26, 2017 Post Count: 5 Status: Offline Project Badges: 
|
There seems to be some bottleneck on the CPU or IO side. GPU utilization often drop sharply for a few second and then back up. Not sure it's a problem, could be just the nature of the work.
|
||
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3295 Status: Offline Project Badges: 
|
There seems to be some bottleneck on the CPU or IO side. GPU utilization often drop sharply for a few second and then back up. Not sure it's a problem, could be just the nature of the work. Once the GPU finishes a job (each task has several of these, could be dozens or even more), it pauses while preparing to run the next job within that task. So you should see several of these ups and downs in GPU utilization which will match with the number of jobs inside the task you were running. AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
Ian-n-Steve C.
Senior Cruncher United States Joined: May 15, 2020 Post Count: 180 Status: Offline Project Badges:
|
There seems to be some bottleneck on the CPU or IO side. GPU utilization often drop sharply for a few second and then back up. Not sure it's a problem, could be just the nature of the work. Once the GPU finishes a job (each task has several of these, could be dozens or even more), it pauses while preparing to run the next job within that task. So you should see several of these ups and downs in GPU utilization which will match with the number of jobs inside the task you were running. a mutex lock to pre-load the next job while one job is running would solve that. if the developers know how to implement that. if they did, hand-off from one job to the next would be pretty seamless and you wouldn't have this lag between jobs. 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 |
||
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3295 Status: Offline Project Badges:
|
There seems to be some bottleneck on the CPU or IO side. GPU utilization often drop sharply for a few second and then back up. Not sure it's a problem, could be just the nature of the work. Once the GPU finishes a job (each task has several of these, could be dozens or even more), it pauses while preparing to run the next job within that task. So you should see several of these ups and downs in GPU utilization which will match with the number of jobs inside the task you were running. a mutex lock to pre-load the next job while one job is running would solve that. if the developers know how to implement that. if they did, hand-off from one job to the next would be pretty seamless and you wouldn't have this lag between jobs. I'm not a developer but that sounds like a great idea if it worked. Feel free to e-mail WCG with that suggestion. AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
Keith Myers
Senior Cruncher USA Joined: Apr 6, 2021 Post Count: 193 Status: Offline Project Badges:
|
It works. We made a mutex lock version of our special app at Seti.
----------------------------------------Very useful for fast running tasks on a gpu to preload the next job. Reduces the wait/loading time. A proud member of the OFA (Old Farts Association) |
||
|
|
hnapel
Advanced Cruncher Netherlands Joined: Nov 17, 2004 Post Count: 82 Status: Offline Project Badges: 
|
It works. We made a mutex lock version of our special app at Seti. Very useful for fast running tasks on a gpu to preload the next job. Reduces the wait/loading time. If it is not requiring serialization. [Edit 1 times, last edit by hnapel at Apr 21, 2021 6:05:09 PM] |
||
|
|
|