| 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 Stress Test |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 781
|
|
| Author |
|
|
davebodger
Cruncher Joined: Sep 19, 2011 Post Count: 3 Status: Offline Project Badges: 
|
I've not received any yet. All my tasks are OPN1 :-(
I have two Nvida cards sitting idle at present.....waiting..... |
||
|
|
Dayle Diamond
Senior Cruncher Joined: Jan 31, 2013 Post Count: 452 Status: Offline Project Badges: 
|
Uplinger, WCG staff, thank you for running this stress test.
Taking this second experiment, which would have run in a year, and running it over the course of a day is probably the most exciting development I've seen since I joined in 2013. Good luck and enjoy the deluge! |
||
|
|
Jorlin
Advanced Cruncher Deutschland Joined: Jan 22, 2020 Post Count: 90 Status: Offline Project Badges: 
|
I've not received any yet. All my tasks are OPN1 :-( I have two Nvida cards sitting idle at present.....waiting..... Did you enable GPU crunching? Just asking because there is a constant flow of tasks right now.  |
||
|
|
KoopaTroopa
Cruncher Joined: May 22, 2016 Post Count: 8 Status: Offline Project Badges: 
|
Would probably be better off queuing a bunch of GPU work in RAM and having the GPU work it instead of using the disk so much. Not really an issue if the task only takes a couple minutes. If one is lost, you could restart it no problem. I have a SSD and while fast, it only has so many write cycles. Might as well keep the whole task in RAM till it's done then write out at the end as it is really doing a lot of picking things up and putting them down again when it doesn't need to.
----------------------------------------[Edit 1 times, last edit by KoopaTroopa at Apr 26, 2021 9:07:38 PM] |
||
|
|
biini
Senior Cruncher Finland Joined: Jan 25, 2007 Post Count: 334 Status: Offline Project Badges:
|
I've not received any yet. All my tasks are OPN1 :-( I have two Nvida cards sitting idle at present.....waiting..... Have you enabled GPU computing from the settings -> device manager -> device profiles? |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges: 
|
I've not received any yet. All my tasks are OPN1 :-( I have two Nvida cards sitting idle at present.....waiting..... You should be able to get work. Please check your settings and what messages you are seeing on your client. These will both help debug why you have not received any work. 1. Go to here: https://www.worldcommunitygrid.org/ms/device/viewDevices.do 2. Then check out the profile your machine is using. 3. Click on Custom radio button 4. Scroll down to GPU dropdowns to make sure you have NVIDIA turned on 5. if you made a change, press save at bottom of the page. (Note: it will take 2 project updates for these to propagate to your machine. click projects tab, select wcg and then press update. Wait 2 minutes and press update again.) As for messages, please capture a recent request for work on your client in the event logs. Thanks, -Uplinger |
||
|
|
Dayle Diamond
Senior Cruncher Joined: Jan 31, 2013 Post Count: 452 Status: Offline Project Badges:
|
Agreed with Koopa. Hopefully lessening the damage to SSDs is the next priority.
My new SSD has 6.6 TB written on it and it's almost all from World Community Grid. |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges:
|
Would probably be better off queuing a bunch of GPU work in RAM and having the GPU work it instead of using the disk so much. Not really an issue if the task only takes a couple minutes. If one is lost, you could restart it no problem. I have a SSD and while fast, it only has so many write cycles. Might as well keep the whole task in RAM till it's done then write out at the end as it is really doing a lot of picking things up and putting them down again when it doesn't need to. I have something similar in the queue of enhancements to the application. However, at the moment, we have more GPUs than we can properly fill. Thanks for the suggestion! -Uplinger |
||
|
|
KoopaTroopa
Cruncher Joined: May 22, 2016 Post Count: 8 Status: Offline Project Badges:
|
Even if you loaded the entire task at hand to ram and operated out of a ram drive or something similar stored in RAM, it would cut down on disk writes a lot. I would not want to run the GPU task since it's cache thrashing pretty bad.
I never see high RAM usage in WCG other than Africa Rainfall and that's only ~ 1GB per task. I manage that by only enabling like 6 Rainfall tasks and it solves it. For a task that can only run 1 at a time, high RAM usage shouldn't be a problem unless it's a few gigs and even then, most wouldn't care. |
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
Agreed with Koopa. Hopefully lessening the damage to SSDs is the next priority. My new SSD has 6.6 TB written on it and it's almost all from World Community Grid. For a dedicated cruncher spend the extra coin for a server grade SSD then you won't have to worry about it.
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.
  |
||
|
|
|