100% GPU load after BOINC is closed.

Message boards : Questions and problems : 100% GPU load after BOINC is closed.
Message board moderation

To post messages, you must log in.

AuthorMessage
axzxc1236

Send message
Joined: 9 Mar 18
Posts: 6
Taiwan
Message 85068 - Posted: 9 Mar 2018, 11:48:19 UTC
Last modified: 9 Mar 2018, 11:53:13 UTC

GPU-Z shows that BOINC can cause 100% GPU load even after BOINC is closed and not BOINC related process shows up in task manager.

My Steup:

    OS: Windows 10 pro 64bit 1709 with all Windows patches installed.
    CPU: Ryzen 5 1600 overclocked to all core 3.6Ghz
    GPU: MSI RX 560 AERO ITX OC 4G (with AMD GPU driver 18.3.1, which is latest version)
    BOINC:7.8.3
    GPU-Z:2.8.0
    BOINC projects: Enigma@Home, Milkeyway@Home, NFS@Home, SETI@Home and World Community Grid



How to reproduce:


    0.Set your BOINC not run to startup
    1.Restart your computer
    2.open GPU-Z and switch to sensors tab (and you should observe low GPU load at the moment)
    3.Open BOINC
    4.Projects are running, can observe that GPU clock and GPU memory clock goes up.
    5.Close BOINC (or set BOINC to not use GPU)
    6.Observe that GPU clock is still max out but GPU memory clock goes down, GPU load stops at 100%. (which is not expected result)
    This makes me worry because it draws more power, possibly affects performance of gaming and electricity where I live is not cheap. (but I am willing to do contribution to science projects from time to time)



Restart your computer solves the problem until you open BOINC again.

ID: 85068 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5078
United Kingdom
Message 85070 - Posted: 9 Mar 2018, 12:04:21 UTC - in response to Message 85068.  

It would be worth looking at BOINC Manager just before you shut down, to see which project's science application(s) are running on your GPU when this happens.

Then, after shutting down BOINC, use Task Manager (or Process Explorer) to see whether that particular science app is still in the task list.

I know that at least one of the SETI@Home science apps for GPU does take an appreciable amount of time to notice that BOINC has requested a closedown (although my experience is with NVidia cards, not AMD).

But once both BOINC and every related project application have closed and vanished from the list of running processes, there would be nothing left under our control to keep the GPU active. If a project science app does remain active for more than a minute (and that's being generous) after BOINC itself closes, you should take it up directly with the project concerned.
ID: 85070 · Report as offensive
axzxc1236

Send message
Joined: 9 Mar 18
Posts: 6
Taiwan
Message 85075 - Posted: 9 Mar 2018, 13:24:58 UTC

I tried to isolate the issue by only enable one project to run at a time, but all the projects that uses GPU (Enigma@Home, Milkeyway@Home and SETI@Home) seems to cause the GPU load problem.
Should I report to all these projects?...

From my observation, all research related process just closes after BOINC manager is closed, nothing stays in RAM.

Also, logout and login again seems to fix the problem aswell.
ID: 85075 · Report as offensive
axzxc1236

Send message
Joined: 9 Mar 18
Posts: 6
Taiwan
Message 85076 - Posted: 9 Mar 2018, 13:51:10 UTC

I tried to run other projects that utilize GPU as well. (Amicable Numbers, Einstein@home and PrimeGrid)
All BOINC projects utilize GPU will trigger the problem even if they run on my computer.
I think it is safe to say that BOINC caused it, but I don't know why or how.
ID: 85076 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5078
United Kingdom
Message 85078 - Posted: 9 Mar 2018, 14:44:24 UTC

OK, I think we can assume (for the time being) that it isn't the science applications which are causing the problem.

The next thing to look at would be - when does the GPU usage start to ramp up? The only thing which BOINC itself does to/with a GPU is to query at startup what GPUs are fitted and what their specifications - memory, speed, driver version, supported languages - are. We did have a similar problem some years ago (but I think it was only present in macbooks fitted with dual GPUs - one low-power to prolong battery life on the road, the other high-powered for desktop power users). In that case, the 'capability test' triggered the high-power GPU to fire up and it wouldn't let go again.

If you could test that, please? Suspend all projects (or finish all worked currently on the machine and set 'no new tasks' so nothing new is downloaded), then do whatever is needed (reboot, log off/on) to ensure the GPU shows as idle. Then, start the BOINC Client and the BOINC Manager (separately if you can), and watch to see if the GPU usage spikes at that stage - and if so, how long for.

Remember to un-suspend the projects, or to fetch new work, when you're done.
ID: 85078 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15478
Netherlands
Message 85081 - Posted: 9 Mar 2018, 15:47:52 UTC - in response to Message 85076.  

I've tested this with BOINC 7.9.2, GPU-Z 2.8.0, my Sapphire RX 470 - 8GB, running the Adrenalin 17.12.2 driver.
Running Seti@Home, solely on GPU, one task at a time.
Using the GPU-Z option to log to file.

GPU clock when idle, 300MHz.
GPU clock when under load, 1260MHz.

Ran the Seti task for 3 minutes.
GPU load between 99 - 100%
Exited BOINC.
Memory clock immediately drops to 300MHz.

Date                , GPU Core Clock [MHz] , GPU Memory Clock [MHz] , GPU Temperature [°C] , Fan Speed (%) [%] , Fan Speed (RPM) [RPM] , GPU Load [%] , Memory Controller Load [%] , GPU only Power Draw [W] , VDDC Power Draw [W] , VDDCI Power Draw [W] , VRM Efficiency [%] , Memory Used [MB] , VDDC [V] ,
2018-03-09 16:42:29 ,             1259.2   ,               2000.0   ,               70.0   ,              19   ,                 920   ,         98   ,                       76   ,                  93.3   ,              72.3   ,               13.6   ,             93.5   ,            439   , 1.1750   ,
2018-03-09 16:42:30 ,             1259.1   ,               2000.0   ,               70.0   ,              19   ,                 921   ,        100   ,                       52   ,                  83.4   ,              65.2   ,               13.6   ,             93.1   ,            439   , 1.1750   ,
2018-03-09 16:42:31 ,             1260.0   ,               2000.0   ,               70.0   ,              19   ,                 921   ,         97   ,                       49   ,                  77.8   ,              58.5   ,               13.6   ,             92.6   ,            439   , 1.1750   ,
2018-03-09 16:42:32 ,             1260.0   ,               2000.0   ,               70.0   ,              19   ,                 920   ,         84   ,                       40   ,                  84.5   ,              64.5   ,               13.6   ,             92.9   ,            439   , 1.1750   ,
2018-03-09 16:42:33 ,             1260.0   ,               2000.0   ,               70.0   ,              19   ,                 920   ,         86   ,                       55   ,                  79.9   ,              68.7   ,               13.6   ,             93.3   ,            439   , 1.1750   ,
2018-03-09 16:42:34 ,             1260.0   ,               2000.0   ,               71.0   ,              19   ,                 920   ,         99   ,                       61   ,                  88.7   ,              63.4   ,               13.6   ,             93.2   ,            439   , 1.1750   ,
2018-03-09 16:42:35 ,             1259.2   ,               2000.0   ,               71.0   ,              19   ,                 920   ,        100   ,                       54   ,                  89.0   ,              61.0   ,               13.6   ,             93.2   ,            439   , 1.1750   ,
2018-03-09 16:42:36 ,             1260.0   ,               2000.0   ,               71.0   ,              19   ,                 920   ,        100   ,                       59   ,                  83.8   ,              65.9   ,               13.6   ,             93.3   ,            439   , 1.1750   ,
2018-03-09 16:42:37 ,             1260.0   ,               2000.0   ,               71.0   ,              19   ,                 920   ,         98   ,                       60   ,                  47.5   ,              14.8   ,               12.2   ,             78.6   ,            440   , 1.0938   ,
2018-03-09 16:42:38 ,             1260.0   ,               2000.0   ,               71.0   ,              20   ,                 932   ,         99   ,                       60   ,                  83.1   ,              66.6   ,               13.6   ,             93.0   ,            440   , 1.1750   ,
2018-03-09 16:42:39 ,             1260.0   ,               2000.0   ,               71.0   ,              20   ,                 949   ,         99   ,                       59   ,                  83.8   ,              64.9   ,               13.6   ,             92.7   ,            436   , 1.1750   ,
2018-03-09 16:42:41 ,             1206.9   ,               2000.0   ,               68.0   ,              20   ,                 954   ,          0   ,                        0   ,                  28.1   ,              10.5   ,               12.2   ,             78.6   ,            436   , 1.0938   ,
2018-03-09 16:42:42 ,              346.7   ,               2000.0   ,               66.0   ,              20   ,                 960   ,          0   ,                        0   ,                  18.2   ,               0.0   ,               12.2   ,             67.1   ,            225   , 1.0250   ,
2018-03-09 16:42:43 ,              388.6   ,               2000.0   ,               65.0   ,              19   ,                 955   ,          0   ,                        0   ,                  21.9   ,               3.4   ,               12.2   ,             73.0   ,            225   , 1.0250   ,
2018-03-09 16:42:44 ,              326.8   ,               2000.0   ,               65.0   ,              19   ,                 936   ,          1   ,                        0   ,                  18.2   ,               0.0   ,               12.2   ,             67.1   ,            225   , 1.0250   ,
2018-03-09 16:42:45 ,              300.0   ,               2000.0   ,               64.0   ,              19   ,                 931   ,          0   ,                        0   ,                  18.2   ,               0.0   ,               12.2   ,             67.1   ,            225   , 1.0250   ,
2018-03-09 16:42:46 ,              480.2   ,               2000.0   ,               64.0   ,              19   ,                 930   ,          0   ,                        0   ,                  18.2   ,               0.0   ,               12.2   ,             67.1   ,            225   , 1.0250   ,
2018-03-09 16:42:47 ,              409.6   ,               2000.0   ,               64.0   ,              19   ,                 929   ,          0   ,                        0   ,                  19.7   ,               0.9   ,               12.2   ,             67.1   ,            225   , 1.0250   ,
2018-03-09 16:42:48 ,              300.0   ,               2000.0   ,               63.0   ,              19   ,                 908   ,          0   ,                        0   ,                  18.2   ,               0.0   ,               12.2   ,             67.1   ,            225   , 1.0250   ,
2018-03-09 16:42:49 ,              300.0   ,               2000.0   ,               63.0   ,              19   ,                 898   ,          0   ,                        0   ,                  18.2   ,               0.0   ,               12.2   ,             67.1   ,            225   , 1.0250   ,
2018-03-09 16:42:50 ,              300.0   ,               2000.0   ,               63.0   ,              19   ,                 894   ,          0   ,                        0   ,                  18.2   ,               0.0   ,               12.2   ,             67.1   ,            225   , 1.0250   ,

09-Mar-2018 16:42:32 [SETI@home] [checkpoint] result blc16_2bit_guppi_58137_29913_HIP46580_0021.20186.409.22.45.159.vlar_1 checkpointed
09-Mar-2018 16:42:40 [---] Exiting
09-Mar-2018 16:42:40 [---] [task_debug] requesting tasks to exit
09-Mar-2018 16:42:40 [SETI@home] [task] task_state=QUIT_PENDING for blc16_2bit_guppi_58137_29913_HIP46580_0021.20186.409.22.45.159.vlar_1 from request_exit()
09-Mar-2018 16:42:40 [---] request_exit(): PID 9028 has 0 descendants

ID: 85081 · Report as offensive
axzxc1236

Send message
Joined: 9 Mar 18
Posts: 6
Taiwan
Message 85082 - Posted: 9 Mar 2018, 16:00:22 UTC
Last modified: 9 Mar 2018, 16:02:39 UTC

I just recorded the issue with ShareX (I can't record with OBS because it also utilize GPU, sorry if the quality is bad.)
ID: 85082 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15478
Netherlands
Message 85083 - Posted: 9 Mar 2018, 16:10:22 UTC - in response to Message 85082.  
Last modified: 9 Mar 2018, 16:14:14 UTC

Enigma@Home runs only on the CPU, not on the GPU. So suspending Enigma@Home does nothing for the GPU clock. Any task still running on the GPU will continue to run on the GPU. Gridcoin perhaps?

Check the Tasks tab, Status column. It'll show there which resource is used. If GPU it will say something like Running (0.04 CPUs + 1 AMD/ATI GPU). If just CPU, it'll state Running.

Edit, you can check at https://boinc.berkeley.edu/projects.php which projects use the AMD GPU.
ID: 85083 · Report as offensive
mmonnin

Send message
Joined: 1 Jul 16
Posts: 146
United States
Message 85087 - Posted: 9 Mar 2018, 20:48:58 UTC - in response to Message 85083.  

Enigma@Home runs only on the CPU, not on the GPU. So suspending Enigma@Home does nothing for the GPU clock. Any task still running on the GPU will continue to run on the GPU. Gridcoin perhaps?

Check the Tasks tab, Status column. It'll show there which resource is used. If GPU it will say something like Running (0.04 CPUs + 1 AMD/ATI GPU). If just CPU, it'll state Running.

Edit, you can check at https://boinc.berkeley.edu/projects.php which projects use the AMD GPU.


Enigma has a GPU app now. It's been around a few months. Also if it didn't, starting it would not make the GPU clocks jump. The https://boinc.berkeley.edu/projects.php page is not kept up to date with new projects and the OS/devices a project supports.

Anyway, the clocks are 2D vs 3D clocks. 2d for idle etc when just at a desktop and not in a game/computing. Some apps will display this as a p-state, like nvidia inspector. So something is making the OS/driver think that it should still be in 3D clocks like if the task is still in memory. Does the task exit or become idle and use 0% CPU?
ID: 85087 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15478
Netherlands
Message 85088 - Posted: 9 Mar 2018, 22:36:31 UTC - in response to Message 85087.  
Last modified: 10 Mar 2018, 0:13:57 UTC

Enigma has a GPU app now. It's been around a few months.
Ah, thanks for that bit of information. I'll pass the info on to David.
But do know that it's normally up to the project admin to tell that they've changed their hardware requirements.

The https://boinc.berkeley.edu/projects.php page is not kept up to date with new projects and the OS/devices a project supports.
No, it's kept up to date, but not all projects are eligible to reach the list. Only production projects, and then only those that actually do good science (and aren't just there to give credit for absolutely nothing), or are (covert) commercial doohickeys. And as said, when the project admin don't say anything about a change in status of their project, there's no way that the list can be updated.
ID: 85088 · Report as offensive
axzxc1236

Send message
Joined: 9 Mar 18
Posts: 6
Taiwan
Message 85093 - Posted: 10 Mar 2018, 3:12:05 UTC - in response to Message 85087.  
Last modified: 10 Mar 2018, 3:58:07 UTC

I tried to disable everything (including Kaspersky Antivirus program) but this problem still exists.
(I tried to produce the same issue in safe mode, but I can't use GPU-Z in safe mode.)
All BOINC related processes "vanishes" from task manager after I close BOINC manager (also vanishes from process explorer), so it's safe to say the tasks exited.

I found out reset global graphical settings also temporary fixes the problem, but the problem happens again after I enable Enigma@Home, maybe it's a driver issue.

Edit: I just reinstalled driver version 17.12.1, the problem still exists.
ID: 85093 · Report as offensive
Juha
Volunteer developer
Volunteer tester
Help desk expert

Send message
Joined: 20 Nov 12
Posts: 801
Finland
Message 85099 - Posted: 10 Mar 2018, 18:06:06 UTC - in response to Message 85093.  

See if you can find out which process is keeping the GPU usage up. Nvidia has a notification area icon that shows which processes are using GPU. Maybe AMD has something like it.

Process Explorer can show GPU usage. Go to View -> Select Columns -> Process GPU and select GPU Usage. Next go to View -> System Information -> GPU -> Engines and select all of the engines. While there, check that there is some activity in one or some of the engines.

Now the main view should show GPU usage. Process Explorer computes the GPU usage with some weird formula so don't expect 100% usage there. I get 12% usage for a GPU task that is using the GPU 95+%.

Windows Task Manager also show GPU usage but for me it's using non-compute engine and I can't find a way to change it.
ID: 85099 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5078
United Kingdom
Message 85100 - Posted: 10 Mar 2018, 18:55:55 UTC - in response to Message 85099.  

Windows Task Manager also show GPU usage but for me it's using non-compute engine and I can't find a way to change it.
For me, I found evidence of activity like this, on the Performance tab:


Note that each of the four mini-graphs has a drop-down chooser above left: this shows activity for 3D and Copy, which I think was for an OpenCL task. It's not easy to match the application with the activity in Task Manager, though.
ID: 85100 · Report as offensive
Juha
Volunteer developer
Volunteer tester
Help desk expert

Send message
Joined: 20 Nov 12
Posts: 801
Finland
Message 85104 - Posted: 10 Mar 2018, 20:11:49 UTC - in response to Message 85100.  

You can add GPU columns to Processes and Details tab by right-clicking the column headings. For me it seems to show only the first engine of either GPU. I use Intel GPU for desktop and Nvidia is compute only. On my Nvidia computing runs in Compute_0 engine and is at about 98% usage, the 3D engine shows 0% usage.
ID: 85104 · Report as offensive
axzxc1236

Send message
Joined: 9 Mar 18
Posts: 6
Taiwan
Message 85110 - Posted: 11 Mar 2018, 3:11:27 UTC
Last modified: 11 Mar 2018, 3:12:16 UTC

#85099 #85100 #85104
I've done all these before and it shows no GPU usage (0%)(the 00 Column on the right) when Enigma@Home (or whatever research uses GPU) is running.
Thank you for advice though.
ID: 85110 · Report as offensive

Message boards : Questions and problems : 100% GPU load after BOINC is closed.

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.