All task "Gravitational Wave search O2 Multi-Directional GPU v2.02" are "Error while computing"
10 Dec 2019 10:51:59 UTC
Topic 220174
(moderation:
All tasks Gravitational Wave search O2 Multi-Directional GPU v2.02 () windows_x86_64 are KO ("Error while computing") and in my preferences > project i unchecked this application "Gravitational Wave search O2 Multi-Directional GPU" but it's does not work. I continue to receive these tasks.
Have you also turned "Allow
)
Have you also turned "Allow non-preferred apps:" off?
If you don't want GPU tasks at all then turn of "Use AMD GPU".
The reason for the tasks to fail was "EXIT_TIME_LIMIT_EXCEEDED", either something is making the tasks run slow or your GPU isn't fast enough to cope with these tasks.
I also took a look at your CPU tasks and they are taking a awful lot of time to complete.
The 2 completed FGRP5 task took 127010 seconds and 117824, my machine with a CPU that's a generation older than yours (i3770K vs. i4770) completes these in about 16000 seconds.
You should be able to get a lot more performance out of your machine.
Your computer tried to runs
)
Your computer tried to run those tasks with AMD FirePro W4100 GPU which has GCN 1.0 architecture.
https://www.techpowerup.com/gpu-specs/firepro-w4100.c2611
AMD GPUs with GCN 1.0 have been observed to be incompatible with GW GPU application. Tasks may seem to run, but they don't run properly and they end up erroring out. A newer architecture would be required. At least some cards with GCN 2.0 are able to run these tasks properly, but a GPU from even newer generations would probably be a safer choice in general.
Your host has also Intel GPU, but there is no GW GPU app variant for Intel chips available. 'Gamma-ray pulsar binary search #1' tasks (FGRBP1G) could be okay for FirePro W4100.
It's ok with "Allow
)
It's ok with "Allow non-preferred apps:" off. Thank
For FGPR5 i don't know. Here are my settings :
neofantome wrote:For FGPR5 i
)
Not sure if the following will lead to a solution but it might be worth experimenting to try and get things moving.
Your CPU has 4 cores that when using hyperthreading shows up as 8 threads.
Looking at the pictures your running a combination of Einstein and Seti work. At the time the picture was taken you had 3 Seti CPU tasks and 5 Einstein FGRP5 tasks running alongside 2 Seti GPU tasks. Your CPU is over committed and has to switch between tasks, normally this is not a big problem and should only lead to a minor slowdown.
I also see that you limit Boinc to using 75% if the CPU time, this works by starting and stopping the tasks. E.g. run tasks for 7,5 seconds then pause for 2,5 seconds, in reality it might be on another timescale but the principle is the same.
Instead you could try to lower the number of cores (processors) Boinc is allowed to use.
As a starting point you could try "Use at most: 50% of the processors" and reset the CPU time setting to 100%.
This will only run 1 CPU task for every physical core and should speed things up. Let it run for a few hours while keeping an eye on the runtimes. If things start to look good then your could experiment with increasing or decreasing the number of cores, 12,5% in either direction will add or subtract 1 core.
OK, thanks. We'll see what
)
OK, thanks. We'll see what happens in time.
Hi. The new setting gives
)
Hi. The new setting gives much better results thanks.