This is more likely a BOINC problem than an Einstein problem. Einstein definitely uses more than one GPU on many of our systems (including three of mine). Yours might be a case where the system needs you to employ a cc_config.xml file with the use_all_gpus option set to on.
I have three NVIDIA 750Ti GPUs but boinc-manager shows only one running. Is this normal for Einstein?
The key is when Boinc first starts up it tells you about the gpu's it finds near the top of the Event Log, the easiest thing to do is to exit Boinc wait about 10 seconds and then restart Boinc and open the Event Log under Tools in the Boinc Manager. Go to the top of the list and start looking for your 750Ti's, it should list them as gpu0, gpu1 and gpu2...if that's what you see a simple cc_config.xml will fix your problem.
This tells Boinc to use all the gpu's it finds which should allow you to run 3 workunits at the same time each on it's own gpu. Use NOTEPAD to copy the above file to c:\programdata\boinc with the name cc_config.xml Yes NOTEPAD will complain but save it as a .txt type file with the extension .xml so Boinc can read it. You will need to stop and restart Boinc again but this time it should start running 3 tasks once it gets done doing it's thing. One problem you may have is keeping the gpu's fed so they crunch as fast as possible though, you may need to reserve a cpu core or two just to do that so the gpu's run at full speed. If you don't know how to do that, and the above works, come back here and we can help you with that too.
I notice when a GPU task is running, and the progress bar reaches 89.997%, it freezes and stays there for maybe a minute. Elapsed time is still running up, but remaining time is frozen @ 3:08 every time. After a minute or so, the task will complete to 100%, and move onto the next one. ???
I'm plagued with multi-GPU rigs that only run one E@H WU at a time while the other 2 or 3 GPUs sit idle. They all used to run all GPUs but recently some dropped down to just one. These are 1060, 1070, 1070 Ti, 1080 & 1080 Ti cards. I recently upgraded the driver to 391.35 for Win7-64 and 396.18 for Linux Mint 18.3.
I tried everything here and a few other things. I can find no reason for. Some rigs run all GPUs.
On 1-of-4 rigs the first two options in my cc_config can turn off GPU processing but they cannot get more than one GPU to run.
If you restart Boinc on one of the problematic machines and then examine the Event log, what does it say when detecting the GPUs? And what does it say when starting Einstein@home tasks?
If your unsure about what lines might be important then copy and past a few more than what seems to be necessary
This is more likely a BOINC
)
This is more likely a BOINC problem than an Einstein problem. Einstein definitely uses more than one GPU on many of our systems (including three of mine). Yours might be a case where the system needs you to employ a cc_config.xml file with the use_all_gpus option set to on.
For details on how to implement that see:
https://boinc.mundayweb.com/wiki/index.php?title=I_have_added_two_GPUs,_but_BOINC_only_uses_one
There are doubtless other possibilities, but if you don't already have this option activated, I'd try that first.
Malcolm Beeson_2 wrote:I have
)
The key is when Boinc first starts up it tells you about the gpu's it finds near the top of the Event Log, the easiest thing to do is to exit Boinc wait about 10 seconds and then restart Boinc and open the Event Log under Tools in the Boinc Manager. Go to the top of the list and start looking for your 750Ti's, it should list them as gpu0, gpu1 and gpu2...if that's what you see a simple cc_config.xml will fix your problem.
<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>
This tells Boinc to use all the gpu's it finds which should allow you to run 3 workunits at the same time each on it's own gpu. Use NOTEPAD to copy the above file to c:\programdata\boinc with the name cc_config.xml Yes NOTEPAD will complain but save it as a .txt type file with the extension .xml so Boinc can read it. You will need to stop and restart Boinc again but this time it should start running 3 tasks once it gets done doing it's thing. One problem you may have is keeping the gpu's fed so they crunch as fast as possible though, you may need to reserve a cpu core or two just to do that so the gpu's run at full speed. If you don't know how to do that, and the above works, come back here and we can help you with that too.
I had this error for about 3
)
I had this error for about 3 days on my ATI cards where only card 0 would run a task. I seemed to have fixed it.
Here is what worked for me: I suspended the project, then via BOINC Projects tab, I 'Updated' the tasks to report and clear out completed ones.
I then hit 'Reset Project' under BOINC projects tab. All tasks were deleted and a new batch was downloaded.
CPU tasks downloaded and started running first, then the GPU tasks downloaded. Now both GPU cards are running tasks again.
Happy times!
I suspect a task got frozen or something in the scheduler for GPU 1? Any higher level nerds around here have an idea as to why?
Thanks!
I notice when a GPU task is
)
I notice when a GPU task is running, and the progress bar reaches 89.997%, it freezes and stays there for maybe a minute. Elapsed time is still running up, but remaining time is frozen @ 3:08 every time. After a minute or so, the task will complete to 100%, and move onto the next one. ???
The executable running is:
hsgamma_FGRPB1G_1.18_windows_x86_64__FGRPopencl1K-ati.exe
Is this normal?
Thanks!
Try reading this thread.
)
Try reading this thread.
Cheers,
Gary.
I'm plagued with multi-GPU
)
I'm plagued with multi-GPU rigs that only run one E@H WU at a time while the other 2 or 3 GPUs sit idle. They all used to run all GPUs but recently some dropped down to just one. These are 1060, 1070, 1070 Ti, 1080 & 1080 Ti cards. I recently upgraded the driver to 391.35 for Win7-64 and 396.18 for Linux Mint 18.3.
I tried everything here and a few other things. I can find no reason for. Some rigs run all GPUs.
On 1-of-4 rigs the first two options in my cc_config can turn off GPU processing but they cannot get more than one GPU to run.
<cc_config>
<options>
<no_gpus>0</no_gpus>
<use_all_gpus>1</use_all_gpus>
<no_opencl>0</no_opencl>
<max_file_xfers_per_project>8</max_file_xfers_per_project>
<process_priority>2</process_priority>
<process_priority_special>2</process_priority_special>
</options>
</cc_config>
I give up. Rigs that don't run all GPUs are moving off E@H to other DC projects.
If you restart Boinc on one
)
If you restart Boinc on one of the problematic machines and then examine the Event log, what does it say when detecting the GPUs? And what does it say when starting Einstein@home tasks?
If your unsure about what lines might be important then copy and past a few more than what seems to be necessary