Pause GPU - computation error
log in

Advanced search

Message boards : Number crunching : Pause GPU - computation error

Author Message
Peter Hucker
Send message
Joined: 5 Jul 11
Posts: 4
Credit: 20,879,539
RAC: 0
Message 24504 - Posted: 21 Nov 2017, 13:45:42 UTC

Collatz works fine on my Radeon R9 290 GPU, until it gets paused (I've set it to stop when the computer is in use, otherwise the interface is terribly sluggish). This always results in a computation error when it tries to continue. Anything I (or the developers) can do to fix this?

Profile step2000
Avatar
Send message
Joined: 1 Aug 13
Posts: 96
Credit: 1,479,949,115
RAC: 1,953,650
Message 24507 - Posted: 21 Nov 2017, 18:44:44 UTC

4 year old and still running is good. The issue is the AMD driver and only supporting older tech. I would look for newer drivers and if that doesn't work then when slow or you need it do a SHUT down on the project and exit the software and once all is good relaunch.
____________
Retired Business Owner/Developer
Working toward a real solution to why programming takes so many versions to get the end product that just keeps getting better with each version. Do Loop of Products!

Peter Hucker
Send message
Joined: 5 Jul 11
Posts: 4
Credit: 20,879,539
RAC: 0
Message 24515 - Posted: 22 Nov 2017, 17:25:34 UTC - in response to Message 24507.
Last modified: 22 Nov 2017, 17:26:04 UTC

I have the latest driver for my graphics card.
I don't want to manually shut it off each time, and would forget to put it back on.
I've changed it to wait longer before coming back on, so the chances of corrupting are lower if I'm using the computer intermittently. It only seems to corrupt 1 in 4 times.

Profile step2000
Avatar
Send message
Joined: 1 Aug 13
Posts: 96
Credit: 1,479,949,115
RAC: 1,953,650
Message 24587 - Posted: 12 Dec 2017, 14:34:44 UTC

Sounds like a write error to the hard drive. Check the HD and update on those drivers maybe.
____________
Retired Business Owner/Developer
Working toward a real solution to why programming takes so many versions to get the end product that just keeps getting better with each version. Do Loop of Products!

Peter Hucker
Send message
Joined: 5 Jul 11
Posts: 4
Credit: 20,879,539
RAC: 0
Message 24613 - Posted: 16 Dec 2017, 19:05:23 UTC - in response to Message 24587.

Nothing wrong with the hard drive or drivers. Running mirrored 256GB SSDs and mirrored 3TB rotary drives using onboard Intel controller with latest drivers.

Profile step2000
Avatar
Send message
Joined: 1 Aug 13
Posts: 96
Credit: 1,479,949,115
RAC: 1,953,650
Message 24626 - Posted: 20 Dec 2017, 15:45:09 UTC

One last idea and this is a long shot but it could be Voltage on the card. The PSU could be right at the edge of voltage use needs. I have 750 psu on all my units and use 80 Gold standard to prevent issues but have had times when a rail on a PSU is acting up and a replacement brings all back. It is hard to track these types of issues down for sure and the compiled code used is working but a pause could be that the voltage drop at the GPU is freaking the system out at the pause point is all I can think of at this point.

While this is just an idea maybe others have seen this and can add to the thread.
____________
Retired Business Owner/Developer
Working toward a real solution to why programming takes so many versions to get the end product that just keeps getting better with each version. Do Loop of Products!

Peter Hucker
Send message
Joined: 5 Jul 11
Posts: 4
Credit: 20,879,539
RAC: 0
Message 24641 - Posted: 22 Dec 2017, 20:23:26 UTC - in response to Message 24626.
Last modified: 22 Dec 2017, 20:24:09 UTC

While I'd accept the graphics card may be dodgy (it's 3.5 years old and has lost a couple of it's monitor outputs, although it never actually crashes), I can't blame the PSU. It's a big beefy Corsair or Crucial (can't remember which) with twice the power it needs. I've used the same supplies on machines with two graphics cards the same as mine, running 24/7 on BOINC or bitcoins.

Anyway, I've changed the settings now, and it only pauses when I use certain applications, so the number of screwed up tasks is kept to a minimum.


Post to thread

Message boards : Number crunching : Pause GPU - computation error


Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.