MSI 7950 crossfire crash/shutdown in games

Toggle sidebar Toggle sidebar
T

Top-Dog

Honorable
Dec 24, 2012
19
0
10,510
Hi, I have a setup as follows:
950w Rosewill PSU
2x MSI 7950 Twin Frozr III in crossfire (one is stock at 880 MHz, one is stock 960 MHz)
i7 3770k (not OCed)
GA-Z77x-UD3H (latest BIOS)
1x HDD, 1x SSD
3x 24" monitors Eyefinity (HDMI, DP -> DVI, DVI)

I have this problem where if I play a game my computer shuts off after a few seconds to a few minutes of game play. There is no warning, no BSOD... nothing that I can tell. The problem only occurs if crossfire is enabled. The problem happens regardless if the graphics cards are overclocked or running at stock settings. The only way to turn the computer back on is to turn the PSU off for 20 or so seconds and then turn it back on.

By the looks of things I would say it's a heat or PSU issue (short, over current, load spike...) but I don't know for sure. I have bought a Corsair 850 W but it's still in the post. This is where things get interesting though, I can over clock both cards and bitcoin mine for hours and stress the CPU at the same time, run full screen 3D stress tests, and never have a crash. My temps also seem normal; 60-75 for CPU, 70 - 85 for GPUs (deg C, under full load).

So, what do you guys think the problem may be? The fact the PSU needs to be reset mean it's a power problem, unless the motherboard can send a shut down signal to the PSU, then it may be a heat or driver issue, or something else.
 
Sort by date Sort by votes
Th3pwn3r

Th3pwn3r

Distinguished
Aug 29, 2011
257
0
18,860
What I would do is stop running Crossfire. I had the same issue when I ran SLI, it was so buggy that I just pulled one card out and was done with SLI and crossfire. Never again will I run 2 cards due to my experience personally and all the issues others have had.
 
Upvote 0 Downvote
T

thewisdomp

Honorable
Jul 18, 2013
105
0
10,710
I used to have a similar problem once. It was my PSU over heating. I would suggest giving it a good cleaning with compressed/canned air. The miniscule amounts of dust/dirt on the fans have a big impact on their performance, not allowing it to cool properly. If that doesn't work, id suggest replacing that PSU. I've never had a Rosewill PSU, but i personally wouldn't trust them to perform perfectly.
 
Upvote 0 Downvote
T

Top-Dog

Honorable
Dec 24, 2012
19
0
10,510


The computer is pretty much new, I built it at the start of the year and only just added the second card. There is virtually no dust hanging around the computer chassis or PSU. What I think it may be is a sudden increase in load from the GPU that makes the PSU detect a fault (hence buying a new PSU) , because when I mine the load is constant and not dynamically changing, but I still don't get why it doesn't crash in a 3D stress test (I've used Heaven, Furmak and Kombuster).
 
Upvote 0 Downvote
T

Top-Dog

Honorable
Dec 24, 2012
19
0
10,510


I don't want to give up yet. This technology has been around for a few years now and the gfx cards and motherboard all should be fine with it. Interesting that you say SLI though, I thought nVidia was much better for drivers and compatibility in general? I'm not an AMD fan I just got the cards because they seemed cheap and well suited to running my three displays.
 
Upvote 0 Downvote
T

Top-Dog

Honorable
Dec 24, 2012
19
0
10,510


How do you completely uninstall the drivers? I've tried uninstalling CCC and deleting the drivers from device manager, but that didn't seem to work, for example, I would still be able to get output from the gfx cards even though I hadn't installed anything new yet. Also I am installing from CCC 13.4 (from AMDs website) and letting that install the drivers, is there another way?
 
Upvote 0 Downvote
T

Top-Dog

Honorable
Dec 24, 2012
19
0
10,510
Well, I followed the steps here: http://www.brightsideofnews.com/news/2012/4/4/guide-how-to-completely-uninstall-amd-graphics-drivers.aspx?pageid=0 and once done it seemed to work, BUT when I looked in CCC overdrive the clocks were 440/625 and 480/625 (MHz), when I tried setting at their defaults i.e. 880/1250 and 960/1250 (I think?) the games start crashing the whole system again. It's obviously not worth playing with the clocks set that low as the performance is even worse than having one card.

So something must be up when the GPU enters high performance mode, like for example after a few seconds of game play.
 
Upvote 0 Downvote
T

Top-Dog

Honorable
Dec 24, 2012
19
0
10,510


I agree with you, but I'm ruling out other causes as well. I just tried starting both cards mining (at stock clocks) within a few seconds of each other and the system crashed... so there must be an instant where the load current peaks and trips the PSU, because I can OC both cards and mine as long as the first card has already reached full load before I start the second one. Having said that I'm not sure why 3D bench mark programs don't kill my system, maybe they only run off one core (and don't use crossfire even though it's enabled and running full screen)?
 
Upvote 0 Downvote
T

Top-Dog

Honorable
Dec 24, 2012
19
0
10,510
Well, I subbed out the PSU and the problem has gone away. So, it's not that the Rosewill can't handle two gfx cards it just can't respond quick enough as two GPUs that throttle up simultaneously (or it detects this as a fault). Maybe, having the 12 V lines split onto 4 separate circuits also limits any headroom for burst currents. I also got rid of MSI afterburner because it was causing some strange/bad things to happen once the gfx cards were being used for an hour or more (screen blanking, disabling the DP output, changing screen refresh rates), even though temps stayed around 65-80 deg C.
 
Upvote 0 Downvote
L

lor037

Distinguished
Apr 18, 2014
4
0
18,510


Did you figure it out? It sounded like your power supply...

Never mind, I saw that you did. I'm about to setup this up here soon myself.
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom