Gromacs GPU work units now available

Message boards : Number crunching : Gromacs GPU work units now available

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 · Next

AuthorMessage
PappaLitto

Send message
Joined: 19 Oct 17
Posts: 6
Credit: 281,670
RAC: 950
Message 5091 - Posted: 19 Oct 2017, 21:03:45 UTC - in response to Message 5089.  
Last modified: 19 Oct 2017, 21:10:50 UTC

By editing -nt 1 parameter in job file (in project folder) you can change CPU threads for WU. In my Windows PC with Xeon v2 and GTX1060 6GB using -nt 3 GPU usage jumps to 80%.
It should also dramatically decrease WU computing time.

can it be automised (in account settings may be?)


I cannot get this to change my GPU utilization with a 1080ti. I am sitting at 50% utilization at a much lower freq. (1569mhz) This is actually ~30% utilization at my max boost of 1999mhz. I have changed the -nt 1 to -nt 3. The application is now using 32% of my 6 core 12 thread cpu clocked at 4.4ghz. Do you have any suggestion of how to improve this utilization?
ID: 5091 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Krzysztof Piszczek - wspieram Polski Projekt Boinc
Project administrator
Project developer
Project tester
Avatar

Send message
Joined: 8 Nov 10
Posts: 125
Credit: 5,508,546
RAC: 3,756
Message 5092 - Posted: 19 Oct 2017, 21:14:36 UTC - in response to Message 5091.  

As you have different configuration then I have you must just try which settings are best fit your machine.

Currently (for testing) I use 3 thread for WU and 2 WU at the same time (by editing app_config.xml file).
Krzysztof 'krzyszp' Piszczek

Member of Radioactive@Home project team and Universe@Home admin.
ID: 5092 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
PappaLitto

Send message
Joined: 19 Oct 17
Posts: 6
Credit: 281,670
RAC: 950
Message 5093 - Posted: 19 Oct 2017, 21:23:01 UTC - in response to Message 5092.  
Last modified: 19 Oct 2017, 21:35:02 UTC

As you have different configuration then I have you must just try which settings are best fit your machine.

Currently (for testing) I use 3 thread for WU and 2 WU at the same time (by editing app_config.xml file).


-nt 2 seems to be the same as -nt 3 in terms of GPU utilization. Whether I have 2 or 3 GPU applications running concurrently I get the same GPU utilization for some reason, ~60% at 1999mhz. I use this in my app_config.xml file I made:

<app_config>
<app>
<name>gmx</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config>
ID: 5093 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Krzysztof Piszczek - wspieram Polski Projekt Boinc
Project administrator
Project developer
Project tester
Avatar

Send message
Joined: 8 Nov 10
Posts: 125
Credit: 5,508,546
RAC: 3,756
Message 5094 - Posted: 19 Oct 2017, 21:32:27 UTC - in response to Message 5093.  

<cpu_usage>0.2</cpu_usage>

You have set that single work unit can use only 20% of CPU thread. This value should by "1".
Krzysztof 'krzyszp' Piszczek

Member of Radioactive@Home project team and Universe@Home admin.
ID: 5094 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
PappaLitto

Send message
Joined: 19 Oct 17
Posts: 6
Credit: 281,670
RAC: 950
Message 5095 - Posted: 19 Oct 2017, 21:34:51 UTC - in response to Message 5094.  
Last modified: 19 Oct 2017, 21:36:38 UTC

<cpu_usage>0.2</cpu_usage>

You have set that single work unit can use only 20% of CPU thread. This value should by "1".

Right, I'll edit that. Also, did you update the application to v24 recently? What did you change?
ID: 5095 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Krzysztof Piszczek - wspieram Polski Projekt Boinc
Project administrator
Project developer
Project tester
Avatar

Send message
Joined: 8 Nov 10
Posts: 125
Credit: 5,508,546
RAC: 3,756
Message 5097 - Posted: 19 Oct 2017, 21:50:55 UTC - in response to Message 5095.  

I just switched off some debug output as it makes stderr.txt full of crap ;)

But, probably tomorrow I will make update for wrapper with some more functionality.
Krzysztof 'krzyszp' Piszczek

Member of Radioactive@Home project team and Universe@Home admin.
ID: 5097 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mmonnin

Send message
Joined: 25 Jan 17
Posts: 34
Credit: 6,961,791
RAC: 2,206
Message 5099 - Posted: 19 Oct 2017, 22:14:05 UTC - in response to Message 5094.  

<cpu_usage>0.2</cpu_usage>

You have set that single work unit can use only 20% of CPU thread. This value should by "1".


This has absolutely zero impact on how much CPU is used to feed the GPU. It only changes how many CPU threads are removed from doing CPU work. If its 1 then your 8 core CPU can only run 7 CPU tasks. If anything less than 1 BOINC will still run CPU tasks along with 1 GPU task. Then its up to the OS to give priority to feed the GPU. Process Lasso in windows and Linux has no issues properly feed GPUs.
ID: 5099 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
STE\/E

Send message
Joined: 1 May 09
Posts: 37
Credit: 720,416
RAC: 1,368
Message 5120 - Posted: 21 Oct 2017, 10:00:45 UTC
Last modified: 21 Oct 2017, 10:29:44 UTC

Up to 40 Hr's on a GTX 960 with a v23 Wu, BoincTasks says it's using 91% of the CPU but I know it isn't because none of the regular CPU Wu's would ever finish if it was using 91% ...

I set the app file to -nt 3 before the Wu started & it took off real fast & climbed fast % wise but it's been at 100% done now for 20 Hr's or so without finishing.
ID: 5120 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Gunde

Send message
Joined: 1 Jan 17
Posts: 15
Credit: 71,613,008
RAC: 46,290
Message 5121 - Posted: 21 Oct 2017, 11:39:45 UTC
Last modified: 21 Oct 2017, 11:41:03 UTC

Started to fail when it got gromacs v0.18 (cuda23) x86_64-pc-linux-gnu

Exit status 195 (0x000000C3) EXIT_CHILD_FAILED
<![CDATA[
<message>
process exited with code 195 (0xc3, -61)
</message>
https://boinc.drugdiscoveryathome.com/result.php?resultid=13770031

Change nt did not change anything and been trying with and without app_config, this happen when it got latest version.
ID: 5121 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Gunde

Send message
Joined: 1 Jan 17
Posts: 15
Credit: 71,613,008
RAC: 46,290
Message 5122 - Posted: 21 Oct 2017, 19:24:24 UTC

Thanks Krzysztof is is back to 0.16 and running fine again.
ID: 5122 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mmonnin

Send message
Joined: 25 Jan 17
Posts: 34
Credit: 6,961,791
RAC: 2,206
Message 5126 - Posted: 22 Oct 2017, 0:54:07 UTC

Whats with the huge disparity in points?
12480736 11644869 18 Oct 2017, 15:35:17 UTC 21 Oct 2017, 2:30:45 UTC Completed and validated 20,386.33 20,250.11 830,742.42 gromacs v0.23 (cuda23)
windows_x86_64
12479110 11649199 18 Oct 2017, 15:30:02 UTC 21 Oct 2017, 13:34:58 UTC Completed and validated 20,364.12 20,222.16 22,291.11 gromacs v0.23 (cuda23)
windows_x86_64
ID: 5126 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
PAK-FA

Send message
Joined: 24 Jan 17
Posts: 34
Credit: 18,829,318
RAC: 29,488
Message 5127 - Posted: 22 Oct 2017, 3:36:57 UTC - in response to Message 5126.  

Whats with the huge disparity in points?
12480736 11644869 18 Oct 2017, 15:35:17 UTC 21 Oct 2017, 2:30:45 UTC Completed and validated 20,386.33 20,250.11 830,742.42 gromacs v0.23 (cuda23)
windows_x86_64
12479110 11649199 18 Oct 2017, 15:30:02 UTC 21 Oct 2017, 13:34:58 UTC Completed and validated 20,364.12 20,222.16 22,291.11 gromacs v0.23 (cuda23)
windows_x86_64

yes points assignment are very strange
ID: 5127 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
LELE1964

Send message
Joined: 21 Feb 17
Posts: 4
Credit: 2,099,670
RAC: 6,480
Message 5129 - Posted: 22 Oct 2017, 4:40:48 UTC

Gromacs v2 why errors after 9 hours of work, because problem with the upload?
ID: 5129 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
STE\/E

Send message
Joined: 1 May 09
Posts: 37
Credit: 720,416
RAC: 1,368
Message 5130 - Posted: 22 Oct 2017, 8:37:10 UTC
Last modified: 22 Oct 2017, 8:39:05 UTC

https://boinc.drugdiscoveryathome.com/result.php?resultid=13959958

Same, error after about 8 Hr's run time ... over 100 hr's of run time without a single Wu finishing.
ID: 5130 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mmonnin

Send message
Joined: 25 Jan 17
Posts: 34
Credit: 6,961,791
RAC: 2,206
Message 5131 - Posted: 22 Oct 2017, 13:28:45 UTC

Wonder whats different with this version to make it a new app name.
ID: 5131 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Krzysztof Piszczek - wspieram Polski Projekt Boinc
Project administrator
Project developer
Project tester
Avatar

Send message
Joined: 8 Nov 10
Posts: 125
Credit: 5,508,546
RAC: 3,756
Message 5132 - Posted: 22 Oct 2017, 16:33:50 UTC - in response to Message 5131.  

Wonder whats different with this version to make it a new app name.

v1 execute one task, v2 is 6 tasks (5 very short, prepare data for sixth) (look at job.xml).
Krzysztof 'krzyszp' Piszczek

Member of Radioactive@Home project team and Universe@Home admin.
ID: 5132 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mmonnin

Send message
Joined: 25 Jan 17
Posts: 34
Credit: 6,961,791
RAC: 2,206
Message 5142 - Posted: 27 Oct 2017, 22:17:02 UTC - in response to Message 5132.  

Wonder whats different with this version to make it a new app name.

v1 execute one task, v2 is 6 tasks (5 very short, prepare data for sixth) (look at job.xml).


Kind of hard to look at job.xml if I can't get any. They were only sent to just several hosts like a week ago and still aren't finished.
ID: 5142 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Bill F

Send message
Joined: 15 Aug 17
Posts: 7
Credit: 42,654
RAC: 130
Message 5143 - Posted: 29 Oct 2017, 4:40:44 UTC - in response to Message 5142.  

I received 25 of the newest WU's and 6 coded out as Completed, validation inconclusive. Seems like a high percentage, but I did not write the Code and I will defer to the Project staff to decide if that is acceptable.

Thanks
Bil F
ID: 5143 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
noxcivi

Send message
Joined: 25 Sep 17
Posts: 9
Credit: 66,637
RAC: 58
Message 5144 - Posted: 30 Oct 2017, 9:22:06 UTC

@Bill F: Did you use a hidden PC for that? I do not see new gromacsV2 jobs listed for any of your three computers.
ID: 5144 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Yavanius
Avatar

Send message
Joined: 17 Jan 17
Posts: 11
Credit: 16,155
RAC: 34
Message 5158 - Posted: 10 Nov 2017, 2:01:34 UTC

We got any kind of consensus on how long these are running. I aborted after 12+ hours with the time remaining beyond 1 day and still incrementing. :/
ID: 5158 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 · 2 · 3 · 4 · 5 · 6 · Next

Message boards : Number crunching : Gromacs GPU work units now available


©2017 All rights reserved | Design by Digital BioPharm Ltd