Tasks list.

San-Fernando-Valley
San-Fernando-Valley
Joined: 16 Mar 16
Posts: 260
Credit: 6915611637
RAC: 20352791

... like I already pointed

... like I already pointed out:  GW usually needs a GPU with more than 2GB VRAM ...

so, please !!

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3681
Credit: 33843005949
RAC: 36874841

San-Fernando-Valley

San-Fernando-Valley wrote:

... like I already pointed out:  GW usually needs a GPU with more than 2GB VRAM ...

so, please !!

 

I can't comment on the more recent O3 tasks since I havent run any. But I know before that some tasks required less than 2GB and could be completed with less than 2GB GPUs in this case. I also thought that there was some point in old AMD GPUs where certain ones were just too old because of the architecture. I think the OP is in this range, but I'm not sure.

 

The OP hasn't given enough specifics about what card he has exactly (other than the arch name "Pitcairn" which can be seen in his tasks), and how exactly he's trying to run them.

 

If his arch isnt too old, and the tasks are using less than 2GB, he could still be failing them if he's trying to run 2 at a time. lots of possible configurations to iron out.

 

but it's best if he just disables GW tasks altogether as that will reduce the frustration of some tasks completing and some tasks not.

 

Probably a good idea for the project to never turn GW-GPU on by default. it's clear that their scheduler still sends work to unsuitable GPUs and causes frustration for users who just sign up and never know to  change the settings manually.

_________________________________________________________________________

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109407907901
RAC: 35283703

adrianxw wrote:I've now got

adrianxw wrote:
I've now got one, a GW, which normally run to completion in ~15 minutes, which has run for 4:47:40, with a remaining at suspension of 6d 17:05:54 increasing. It is showing 2.910% done.

Your earlier report gave a link to a GRP GPU task.  I gave you a detailed explanation of what the problem was with that task.  For GRP GPU tasks, you have 126 in total of which only 1 shows as an error.  Your GPU can do these quite well.

For GW GPU tasks (received more recently by your host) you have just 2, (none completed) so how you can claim a 'normal' time of ~15 mins is beyond me.  If you look at the stderr output for the one that did fail, you can see the exit status reported as "EXIT_TIME_LIMIT_EXCEEDED".

If you use the forum search function with the string "time_limit_exceeded AND pitcairn", (2 terms to limit the results) you will find a previous post of mine back in 2019 where the problem with GCN 1st generation GPUs (like Pitcairn and Tahiti series) was described.  Those 1st gen GPUs cannot handle the GW GPU app.  There have been many further examples since that time so the problem is documented enough for a search on just the exit status to find many others, if you want further proof.

If you want to run Einstein GPU tasks, just choose the GRP GPU app and accept that (fairly rarely these days) the old Pitcairn GPU may stop making normal progress and need a reboot to get it going again.  If you see a GRP task not making normal progress, my experience is that after a reboot, the task will restart from the last saved checkpoint and then complete normally.

Cheers,
Gary.

adrianxw
adrianxw
Joined: 21 Feb 05
Posts: 242
Credit: 322654862
RAC: 0

I changed my settings and

I changed my settings and enabled Einstein again, it downloaded a shed load of GRP units and is churning through them now. So I am hopeful that normal service can be resumed. I don't have Einstein on my other machine, the GPU in that one started to suffer, the machine really wants a new graphics card, but it happily turns out Rosetta, TN, SiDock etc.

The way it seems set up is the settings apply to the user, thus I cannot set a "per machine" work fetch policy.

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

Harri Liljeroos
Harri Liljeroos
Joined: 10 Dec 05
Posts: 3611
Credit: 2902425571
RAC: 1042000

You can have different

You can have different preference sets (generic, work, school, home). They used to be called venues (and still are on other Boinc projects). You can select a different venue for each of your hosts and thus can have different preferences for different hosts. https://einsteinathome.org/account/prefs/project/combined

adrianxw
adrianxw
Joined: 21 Feb 05
Posts: 242
Credit: 322654862
RAC: 0

Yes, thats true Harri, I'd

Yes, thats true Harri, I'd not thought of that, I'll have a look at it. Thanks.

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

earthbilly
earthbilly
Joined: 4 Apr 18
Posts: 59
Credit: 1140229967
RAC: 0

I have had this issue when a

I have had this issue when a PCI-e riser bites the dust. Replacement is the test for a cure if nothing else raises eyebrows, and if persist after cleaning all AMD stuff out with the AMD driver cleanout apptool and reinstall latest AMD Enterprise drivers, set to 'compute'.

Work runs fine on Bosons reacted into Fermions,

Sunny regards,

earthbilly

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.