Bitcoin Forum
July 25, 2024, 10:54:33 AM *
News: Help 1Dq create 15th anniversary forum artwork.
 
   Home   Help Search Login Register More  
Pages: [1]
  Print  
Author Topic: Help with cgminer, 4x7950, Windows 7  (Read 3094 times)
rtgornik (OP)
Newbie
*
Offline Offline

Activity: 20
Merit: 0


View Profile
April 24, 2013, 04:18:20 AM
Last edit: April 24, 2013, 04:34:04 AM by rtgornik
 #1

Greetings Bitcointalk!

I'm having trouble getting cgminer running for more than 12 hours at a time.

My hardware setup is:
- GPU: 4 x Gigabyte 7950 GV-R795WF3-3GD
- Mobo: Gigabyte GA-H61MA-D3V
- PSU: Rosewill LIGHTNING-1300  / Seasonic SS-1250XM
- CPU: Cheap Celeron
- RAM: 4GB
- HDD: Whatever

I have run cgminer 2.11.4 for 12 hours straight on a few occassions with these settings at about 580kh/s each card (LTC mining):

Code:
cgminer --thread-concurrency 16158 -g 1 -I 18 

I was running 13.3 beta catalyst and corresponding APP SDK.

However, after about 12 hours, cgminer would simply freeze. To resolve the issue, I tried:
- replacing the GPUS
- replacing the mobo
- replacing the RAM
- replacing the PSU
- compiling cgminer myself

None of this worked. I finally decided that it might be a software/driver issue, so I reformatted and reinstalled Windows 7. I installed Catalyst 13.1 drivers, the latest APP SDK, the latest cgminer 3.0.0, and nothing else.

Now I cannot find any thread concurrency values that do not generate HW errors.

My questions:
- Is it normal to have completely different cgminer "tunes" for two identical hardware platforms? For the above, even the older cgminer 2.11.4 doesn't work with the settings that I was originally running 12 hours with.
- Is it possible to not be able to find a --thread-concurrency value that doesn't product HW errors?
- Is it normal to tune each card individually, or to have a single globale --thread-concurrency setting?

I've been struggling with this setup for weeks, and would truly appreciate any help the community can offer!

miaviator
Donator
Hero Member
*
Offline Offline

Activity: 686
Merit: 519


It's for the children!


View Profile WWW
April 28, 2013, 03:04:27 AM
 #2

What are your clocks set to?

Silicone chips are unique with up to a 20% variance depending on a lot of things this should make each card have a "sweet spot" or combo for TC/mem clock/gpu clock.

If your getting hardware errors, try reducing TC.

Even better grab a spread sheet, run your TC up in increments of 2048 and log the hashrate and hardware errors for 10 shares on each TC setting. You may also want to play with I 13 - 20 on each card to see where the HW errors stop at.  If your using risers try without them.

Pages: [1]
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!