rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
Maybe we have found a solution with commandline without hardcoded.
The GPU_limit issue is in progress
Testing with v0.08
update
and failed...
reported for latest wrapper code only, will update and retry
v0.10 with updated wrapper
/ 
Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
We need a longer run to test checkpoints.
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
batch cptest1:
- max_tasks_in_progress = 2
- 14min on a 5500XT, faster on better GPUs
- deadline 4h + 1d grace_period
Checkpoint test
Go to your slot folder and check the stderr.txt
After 5min or you see an entry of an iteration stop BOINC and resume the test, the iteration line must be continue after the last line.
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
The same behavior as TF, the CPU time is resetted to 0 but all is working as it should.
/ Reply
Quote
They don't run 5 minutes on my system and I'm not sure what an iteration looks like. They have all completed now. Maybe someone else was able to test the checkpoint function.
/ Reply
Quote
Checkpoint test
Go to your slot folder and check the stderr.txt
After 5min or you see an entry of an iteration stop BOINC and resume the test, the iteration line must be continue after the last line.
Have you hard-coded it to have a 5 minute checkpoint or are you relying on the Boinc setting to checkpoint as set by a user ?
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
Checkpoint test
Go to your slot folder and check the stderr.txt
After 5min or you see an entry of an iteration stop BOINC and resume the test, the iteration line must be continue after the last line.
Have you hard-coded it to have a 5 minute checkpoint or are you relying on the Boinc setting to checkpoint as set by a user ?
could be the
-iters
option
Could be earlier on faster cards to write / reach the iteration
I like the standard values. Nothing to setup by user
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
If you have multi-GPUs in your system pls test it to make sure that you are getting 1 WU per GPU. The tests are short.
/ Reply
Quote
With this new batch of tasks, I am only getting 1 task on the PC which has 2 GPUs. Earlier today, I got 2 tasks, one per GPU.
The PC is asking for more work, but the server says no tasks available, even tho I see there are ones ready to send on the Project status page.
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
With this new batch of tasks, I am only getting 1 task on the PC which has 2 GPUs. Earlier today, I got 2 tasks, one per GPU.
yes, set back to 1 if someone else getting more.
/ Reply
Quote
If you have multi-GPUs in your system pls test it to make sure that you are getting 1 WU per GPU. The tests are short.
So, the answer is no, I am not getting 1 WU per GPU, I am getting 0.5 WU per GPU.
/ Reply
Quote
If you have multi-GPUs in your system pls test it to make sure that you are getting 1 WU per GPU. The tests are short.
This host with two GPUs only gets one WU. https://srbase.my-firewall.org/sr5/show_host_detail.php?hostid=222234
BM reports two 5070 Ti but it has one 4070 Ti and one 5070 Ti installed.
Another one with 5070 Ti and Pro VII is getting one WU each.
https://srbase.my-firewall.org/sr5/show_host_detail.php?hostid=226930
/ Reply
Quote
yes, set back to 1 if someone else getting more.
That's why :D
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
yes, set back to 1 if someone else getting more.
That's why :D
That setting is 1 WU per GPU but could be bugged. If you say the other host got 1 each was this for LL-dc or TF?
/ Reply
Quote
The setting is bugged for me. 1 PC, 2 GPUs, 1 WU only.
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
The setting is bugged for me. 1 PC, 2 GPUs, 1 WU only.
@FritzB
/ Reply
Quote
Only LL-dc
I've disabled all other apps in the Einstellungen :)
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
ok, sent the observation to the boinc dev.
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
There is another config added to test if you are getting 1 task per GPU.
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 8486
Credit: 187,852,813
RAC: 74
The test was working. If you have multi-GPU you will get 1 task per GPU.
/ Reply
Quote