Hello Folks!
Recently .. I'm seeing errors on all GPU tasks on my (2013) Mac Pro that has been crunching fine for over a year. The Mac has dual FirePro ATI D700 cards.
http://einsteinathome.org/host/11631928/tasks
Its not obvious what the issue is. Here is a sample log:
http://einsteinathome.org/task/499726904
Any suggestions?
Copyright © 2024 Einstein@Home. All rights reserved.
Mac Pro (2013) errors on all GPU tasks
)
You don't have any GPU tasks prior to 14th May, whilst you do have lots of CPU tasks. Did you restart GPU crunching at that time?
You are using Yosemite 10.10.3. Have you had successful GPU tasks with that version or were those previous successes done under an earlier version?
When did you do the upgrade to 10.10.3 and is it possible that the problems are associated with the upgrade?
Cheers,
Gary.
Also see this
)
Also see this discussion:
http://einsteinathome.org/node/198054&nowrap=true#140458
HB
Thanks for the prompt
)
Thanks for the prompt feedback folks. It was all OK with Yosemite, I'm certain. I think it was likely the 10.10.3 update (I wasn't watching very carefully).
Thanks H-B for pointing me to that thread; I just the video-driver update that Oliver suggested. Lets see what happens ..
RE: Thanks for the prompt
)
I saw a note today that said it's best to reinstall Boinc after upgrading to the latest version of the Mac OS.
No change after installing
)
No change after installing the video-driver update and reinstalling BOINC. Still getting all GPU tasks failing with the same errors.
Are both GPUs and the CPU
)
Are both GPUs and the CPU running tasks?
Did you keep an eye on the temperature?
The new MacPros cooling design is not made for 24/7 use under full load. It could be that the components did run too hot over a long time which is now causing errors. Hope that's not the cause though, any way to test the GPU reliabilty otherwise?
Generally I suggest to install a tool to read the temperature and manually control the fans when crunching on a Mac.
The MacPro design is just
)
The MacPro design is just fine for 24/7 load. We tested this for over a year in my lab (monitored temperatures etc).
Here is where the current discussion is now:
http://einsteinathome.org/node/198054&nowrap=true#140458
Do you have the numbers of
)
Do you have the numbers of your temperature test and watts consumed somewhere online maybe?
On full load the GPUs hover
)
On full load the GPUs hover between 80 -- 85 C and pull in the ballpark of 400W. This is on the 8-core + D700 model. The 4-core + D500 is about 10 degrees cooler.
thanks for the numbers, good
)
thanks for the numbers, good to know that not all MacPros are affected.
The fan can manage to cool about 430 Watts at normal room temperature before it reaches the thermal limit.
In the top configuration it can consume up to 450 watts, so depending on what CPU is built in and how much load the CPU gets this can be an issue for some machines.