I recently brought a tower unit back on line. It runs Ubu18 and a Nvidia GeForce GTX 770. Using Boinc manager I clicked on Update for E@H. As expected several work units were down loaded. For the project I had configured for 2 simultaneous GPU WUs to run. So far all is as expected. Then on the Einstein website I selected the "Location" for this pc and changed the GPU utilization factor for FGRP from .5 to .33. After doing this all WU on the pc started to fail with "computation errors. Of interest the two current active WUs continued to completion but the others as they "became active" failed (the third active WU). This was a quick death for these WUs. I dont recall seeing behavior like this before.
Copyright © 2024 Einstein@Home. All rights reserved.
robl wrote:I recently brought
)
Not enough memory on GPU?
2GB GPU mem wasn't enough to
)
2GB GPU mem wasn't enough to run tasks 3x at least in the past. But have you tried to stop Boinc completely and see what happens if you restart it (with that 3x setting). Would one of the three tasks fail... or more?
Richie wrote:2GB GPU mem
)
two tasks would run with a 3rd task trying. it would run for a few seconds and fail, then the next task above two would run and it too would fail. this behavior would continue until all tasks above two concurrent tasks failed. then I would be locked out from further downloads for several hours as expected.
Thanks to you both for taking the time to reply. It seems you nailed it.