Advanced search

Message boards : Number crunching : GPU Computing Issue - Need advice

Post to thread Subscribe Sort
Author Message
nedmanjo
Avatar
Send message
Joined: 29 Jan 16
Posts: 3
Credit: 153,854,015
RAC: 0
Level
Ile
Scientific publications
watwat
Message 47799 - Posted: 22 Aug 2017 | 0:12:29 UTC
Last modified: 22 Aug 2017 | 0:13:36 UTC

Have been unable to run GPUGRID consistently. I used to run it for a fair amount of time but then my system suddenly began to have power off events. I run a number of other GPU intensive projects without issue but can't seem to run GPUGRID. I have the following applications settings:

ACEMD short runs (2-3 hours on fastest card): yes
ACEMD long runs (8-12 hours on fastest GPU): yes
ACEMD Beta: no
Python Runtime: yes

Any advice on getting to the bottom of this would be appreciated.

Windows 10 Pro 64Bit
[2] Intel(R) Xeon(R) CPU X5690 @ 3.47GHz
[2] NVIDIA GeForce GTX TITAN Black, Driver: 381.65 OpenCL: 1.2
____________

ID: 47799 | Rating: 0 | rate: Rate + / Rate - Report as offensive    Reply Quote
Jim1348
Send message
Joined: 28 Jul 12
Posts: 819
Credit: 1,591,285,971
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 47801 - Posted: 22 Aug 2017 | 14:11:17 UTC - in response to Message 47799.

I would detach from GPUGrid and then re-attach. There might be something corrupted in BOINC due to the power outages.

ID: 47801 | Rating: 0 | rate: Rate + / Rate - Report as offensive    Reply Quote
Jacob Klein
Send message
Joined: 11 Oct 08
Posts: 1127
Credit: 1,901,927,545
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 47835 - Posted: 30 Aug 2017 | 12:01:43 UTC

Sounds like a power supply issue.

You might try stress-testing your system, using tools like:

[for CPU]:
- Prime95
- Intel LINPACK dlls put into LinX

[for GPU]:
- Heaven
- Valley
- PrimeGrid

Might need to test CPU stress alongside GPU stress, ie: run multiple programs at once.

ID: 47835 | Rating: 0 | rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2346
Credit: 16,290,365,968
RAC: 6,912,525
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 47837 - Posted: 30 Aug 2017 | 18:57:30 UTC - in response to Message 47835.
Last modified: 30 Aug 2017 | 18:58:58 UTC

Sounds like a power supply issue.

You might try stress-testing your system, using tools like:

I would like to add furmark for GPU testing.
Before you do any stress test: power off your PC, unplug the power cord, then disconnect all power connectors inside your PC (especially the CPU, 24-pin MB, PCI-e GPU connectors) and check for any burn or melting marks (especially on the 12V (yellow cable) pins of the 24-pin MB connector). If there's none, reconnect them and the power cord, and then do the stress test.

ID: 47837 | Rating: 0 | rate: Rate + / Rate - Report as offensive    Reply Quote
nedmanjo
Avatar
Send message
Joined: 29 Jan 16
Posts: 3
Credit: 153,854,015
RAC: 0
Level
Ile
Scientific publications
watwat
Message 49244 - Posted: 7 Apr 2018 | 3:17:59 UTC - in response to Message 47837.

Checked every cable and connection. Suspected PSU even though the 1100 watt system ran Furmark without issue. Added supplementary PSU, no joy. Finally picked up a used Supermicro box w/1280 watt PSU, moved my major components over and problem solved. New problem after, major increase in electricity but happily crunching away. Thanks for the advice.
____________

ID: 49244 | Rating: 0 | rate: Rate + / Rate - Report as offensive    Reply Quote
Post to thread

Message boards : Number crunching : GPU Computing Issue - Need advice

//