Advanced search

Message boards : Number crunching : Not getting work for my GPU anymore

Author Message
Dayle Diamond
Send message
Joined: 5 Dec 12
Posts: 84
Credit: 1,587,840,915
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 31720 - Posted: 23 Jul 2013 | 14:51:06 UTC

Starting last night my computer hasn't been assigned any tasks.
I've upped my resource share to 1000 (78%).
I've allowed the server to assign tasks from the short queue.
I've increased the work buffer from 0.6 days to 1.8 days [usually takes 25 hours to run a long task on my GTX 460].

My CPU projects, World Community Grid & FightMalaria@Home, are crunching away.

Any diagnosis would be appreciated.

Profile dskagcommunity
Avatar
Send message
Joined: 28 Apr 11
Posts: 456
Credit: 817,865,789
RAC: 0
Level
Glu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 31721 - Posted: 23 Jul 2013 | 16:04:59 UTC
Last modified: 23 Jul 2013 | 16:13:22 UTC

You should not set the buffer so high (only 0,1 or less?), that costs you only points because you miss the 125% Bonus for sending the units back inner 2 Days. I see your only one from todays is validated. The ones before are mostly erroring, so i would say, the server did not send you more units on that days because you produced so much errors. The second thing i would try to stop cpu work for two days only to test if it reduce perhaps the errorrate or leave one cpu core free with no work. When i does not work, try to set 25mV more GPU Voltage and perhaps the GPU Clock 50-100Mhz lower when it doesnt help. The third thing because i see your computer is running Win7 is, do you run the units in one piece without reboot, logout or user switching? Because these three Events can crash the units too, in this case do do not need to change anything on your system, only let the units compute in one piece.
____________
DSKAG Austria Research Team: http://www.research.dskag.at



Crunching for my deceased Dog who had "good" Braincancer..

Dayle Diamond
Send message
Joined: 5 Dec 12
Posts: 84
Credit: 1,587,840,915
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 31722 - Posted: 23 Jul 2013 | 16:28:37 UTC

Thanks for the quick advice, but there's a few things I really don't know how to implement. How do I leave one CPU free, and why would that help? Would telling GPU grid to use up to 100% of a CPU do the same effect?

How do I send more voltage to the GPU?

I'm trying to reboot as little as possible (I'm the only user). Sometimes after running GPUgrid tasks for a few hours, the PC will become unresponsive - the screen stays black and does not respond to mouse and keyboard input, but I can hear the fans running full crunching power.

During those times I tend to reboot, and when I come back, 60% of the time the workunit will have failed.

Temp on the card isn't going past ~75 C.

TJ
Send message
Joined: 26 Jun 09
Posts: 815
Credit: 1,470,385,294
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 31735 - Posted: 24 Jul 2013 | 22:18:51 UTC - in response to Message 31722.

Thanks for the quick advice, but there's a few things I really don't know how to implement. How do I leave one CPU free, and why would that help? Would telling GPU grid to use up to 100% of a CPU do the same effect?

How do I send more voltage to the GPU?

I'm trying to reboot as little as possible (I'm the only user). Sometimes after running GPUgrid tasks for a few hours, the PC will become unresponsive - the screen stays black and does not respond to mouse and keyboard input, but I can hear the fans running full crunching power.

During those times I tend to reboot, and when I come back, 60% of the time the workunit will have failed.

Temp on the card isn't going past ~75 C.

The CPU feeds the GPU and is only little time active. With the new long runs CPU time is almost the same as GPU time, so the CPU is important. If all cores are crunching then the GPU load will drop dramatically.
You can change CPU usage via the BOINC Manager, then to the tab "processor usage", and then at the bottom "On multiprocessor systems, use at most ... % of the processors". You can set that at 99% but I have bad experience with that setting.
If you have 8 cores/threads, then every 12.5% is one core/thread.
I have set it to 80%.

Giving the graphics card more voltage or clock rate you can use utility software from the manufacturer of the card, or you can download MSI Afterburner, which is free and heavily used in this project by many crunchers.

However as dskagcommunity already said, first try to run without any tasks on the CPU. Set new work to no and let them finish and then see how it goes.
And work for 0.1 or 0.15 is enough.
You can also try short runs only first.

One tip: if you need to reboot your system you can first set the work to "suspend", via BOINC Manager and then reboot. Afterwards you can resume the task(s). In many cases this worked at my system.
____________
Greetings from TJ

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 31759 - Posted: 27 Jul 2013 | 13:26:17 UTC

The first thing to check is whether BOINC requests GPU at all. To do this set BOINC manager to advanced mode, then choose extras -> messages. Initiate a manual project update for GPU-Grid and report the lines appearing in the log.

MrS
____________
Scanning for our furry friends since Jan 2002

Dayle Diamond
Send message
Joined: 5 Dec 12
Posts: 84
Credit: 1,587,840,915
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 31848 - Posted: 6 Aug 2013 | 16:22:29 UTC - in response to Message 31759.

No new tasks since my first post.
I was able to run the GPU version of Einstein@Home, so I know BOINC is recognizing my card. And there haven't been any crashes.

I went back in the event log, and I've got days and days of this.

8/6/2013 9:13:43 AM | GPUGRID | update requested by user
8/6/2013 9:13:45 AM | GPUGRID | Sending scheduler request: Requested by user.
8/6/2013 9:13:45 AM | GPUGRID | Requesting new tasks for NVIDIA
8/6/2013 9:14:14 AM | GPUGRID | Scheduler request completed: got 0 new tasks
8/6/2013 9:14:14 AM | GPUGRID | No tasks sent

Stopped all new work and stopped all other projects from updating. Temporarily ramped the work buffer to five days just to rule that out as a cause. The last hour or so of tasks is finishing now.

And still...

8/6/2013 9:20:44 AM | GPUGRID | Sending scheduler request: Requested by user.
8/6/2013 9:20:44 AM | GPUGRID | Requesting new tasks for NVIDIA
8/6/2013 9:21:00 AM | GPUGRID | Scheduler request completed: got 0 new tasks
8/6/2013 9:21:00 AM | GPUGRID | No tasks sent

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 31880 - Posted: 7 Aug 2013 | 18:37:41 UTC - in response to Message 31848.

Strange.. but I see you got one today. Let's how this works out!

MrS
____________
Scanning for our furry friends since Jan 2002

Post to thread

Message boards : Number crunching : Not getting work for my GPU anymore

//