AMD GPU Task Turns Computer Off Immediately

Message boards : GPUs : AMD GPU Task Turns Computer Off Immediately
Message board moderation

To post messages, you must log in.

AuthorMessage
SSendra

Send message
Joined: 10 Sep 19
Posts: 2
United States
Message 92769 - Posted: 11 Sep 2019, 2:13:12 UTC

Computer Specs:
CPU: Intel Core i7-4810MQ
1st GPU: Intel HD 4600
2nd GPU: AMD FirePro W4170ML
Memory: 16GB
Boinc Version: 7.14.2 (x64)
I updated the intel and amd graphics drivers to the latest versions. It looks like both GPU support OpenCL.

I have a laptop with the specs above. I am able to run tasks that use CPU only, CPU + intel GPU. I am unable to run tasks with the CPU + AMD GPU. When a tasks starts that requires the AMD GPU but computer turns off immediately as soon as the AMD tasks starts. No blue screen, no errors, etc. I have seen a few times when restarting the computer I get an error of "Your system automatically powered off because it became warmer than expected." The laptop has a switchable graphics and most applications are using the Intel GPU so the AMD GPU is not really used. The AMD GPU is working, because when I use applications that require it, I can see usage, even while the boinc is using the CPU and Intel for tasks. When my other application is using the AMD GPU, the computer does not shutdown. So this tells me it is not a heat issue, I would expect the computer to run for a little while before shutting down. As mentioned before the computer turns off immediately as soon as the AMD GPU tasks starts in boinc.

I would like to use the AMD GPU in the laptop because it is a higher performance than the Intel and would like to contribute more to the scientific community.

Any thought or help would be greatly appreciated.
ID: 92769 · Report as offensive
Gary Roberts

Send message
Joined: 7 Sep 05
Posts: 125
Australia
Message 92770 - Posted: 11 Sep 2019, 6:56:19 UTC - in response to Message 92769.  

.... So this tells me it is not a heat issue, I would expect the computer to run for a little while before shutting down.
Whilst you would not expect the temperature to rise instantly and therefore might expect to see a bit of a delay, maybe the firmware is using something other than a temperature change to invoke a protection mechanism. I have no idea if this is ever done but perhaps it might be current draw that triggers the response. You don't mention what project is supplying the GPU work but if that work is really compute intensive, maybe some current limit is being tripped. I could imagine that happening quite quickly - almost instantly.
Cheers,
Gary.
ID: 92770 · Report as offensive
swiftmallard

Send message
Joined: 18 Aug 19
Posts: 1
United States
Message 92774 - Posted: 11 Sep 2019, 18:28:25 UTC

Your GPU work is coming from Milky Way, yes? Are you absolutely certain that your graphics card supports the double precision the project requires? I don't know of any mobile cards from 2013 that will.
ID: 92774 · Report as offensive
Profile Dave

Send message
Joined: 28 Jun 10
Posts: 877
United Kingdom
Message 92779 - Posted: 12 Sep 2019, 7:21:54 UTC

From what I have seen of many laptops running BOINC I would be wary of adding the extra of graphics cards running flat out if you are also running tasks using all the cores of the CPU. Especially if ambient temperatures are high. (My current laptop is much happier running all four cores now that temperatures are typically around 20C rather than 30!)
ID: 92779 · Report as offensive
Profile Joseph Stateson
Volunteer tester
Avatar

Send message
Joined: 27 Jun 08
Posts: 538
United States
Message 92805 - Posted: 15 Sep 2019, 2:25:29 UTC - in response to Message 92770.  

.... So this tells me it is not a heat issue, I would expect the computer to run for a little while before shutting down.
Whilst you would not expect the temperature to rise instantly and therefore might expect to see a bit of a delay, maybe the firmware is using something other than a temperature change to invoke a protection mechanism. I have no idea if this is ever done but perhaps it might be current draw that triggers the response. You don't mention what project is supplying the GPU work but if that work is really compute intensive, maybe some current limit is being tripped. I could imagine that happening quite quickly - almost instantly.



Try underclocking the GPU using msi afterburner or, if supported, AMD's wattman. If it works at low speed then that current surge could be the problem.
ID: 92805 · Report as offensive

Message boards : GPUs : AMD GPU Task Turns Computer Off Immediately

Copyright © 2020 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.