7950 Crashing My System

[Gronnie]

Hello all. I purchased a HIS 7950 IceQ Boost Clock 3GB GDDR5 PCI-E Graphics Card H795QC3G2M less than 3 months ago (mid-October). Ran great at first. Cool, quiet, no crashes. Card has never been overclocked and never used for mining.



Starting yesterday this thing is crashing my system. I had a couple of friends over and they are just getting into PC building so I showed them my system and ran Unigine Heaven, and much to my surprise about 15 seconds in my screen went black. Very embarassing.


I tried uninstalling and reinstalling drivers, setting bios to optimized defaults, reseating the card, etc. Everything I could think of. I have tried running Unigine Heaven, 3DMark, and Furmark. All three crash the system after varying degrees of time. I am also getting coil whine when the card is at load, which I have never noticed before. I ran Intel Burn Test 50x on Very High to try to rule out CPU and to a lesser extent PSU and it completed just fine.


Recent Newegg reviews seem to indicate this isn't a problem that only I have had. Is there anything else I can try or is it time to RMA? I am dreading RMA as HIS does it through a third party and I have read that the experience is terrible.


System specs:

Corsair 500R
ASRock Z87 Extreme4
Intel i7 4770k
Zalman CNPS9900ALED Cooler
Samsung 840 EVO 250GB SSD
2x8GB Crucial Ballistix Low Profile DDR1600 CL8 RAM
3x1TB WD Caviar Green HDD
HIS IceQ 7950
3x LG BD drive, 1x Samsung DVD drive
XFX P1-650X-XXB9 650W 80 PLUS BRONZE Certified Modular Active PFC Power Supply
2x Acer P235Hbmid monitors
Windows 7 Ultimate x64

 

[Geforce man]

What kind of temperatures are you getting? Run MSI afterburner or GPU-z log to find out what it is hitting. Pay specific attention to the VRM temps. Also with GPU-z you can see the voltage that it is running at load. Please note this as well.

Fire up OCCT for 5 minutes (don't do it for longer, no need), and report back.

Thanks!

 

[KingFatty]

Get a utility (GPU-Z here http://www.techpowerup.com/gpuz/ ) to monitor the parameters of the video card, then duplicate the error, and see if there is a consistent pattern with the parameters.

For example, let's assume the problem is the heatsink is not making good contact with the video card, or the fan has hair stuck in it and isn't spinning fast enough to cool the card. You would run GPU-Z, and look at the GPU temperature and see that it's a bit warm. Then, you'd run Heaven and watch as the temperature went beyond specs, and the card had an error and reset or something. You could also watch all the other sensors and look for weird behavior. After confirming the temperatures were too high, you could remove the heatsink and clearn the fan and re-apply a new coat of thermal compound/paste, and the card would be fine.

Or, another scenario, is there are no visible anomalies in the parameters, everything is running smoothly, and temps are within range. Yet the problem persists. So this could be a software issue, or it could be the card is broken. Let's rule out the software by running the card using an entirely different software setup. You can do this by putting the card in another computer, or maybe you can find another hard drive for your current computer, or maybe you can find a "demo" or boot disc for Linux that has a video stress test on it. Then, boot the computer with that demo disc so you are not running off your hard drive at all. If you can duplicate the problem with a completely different software setup, then it's probably something wrong with the video card. You could also reinstall windows to get a fresh install, but that's a pain.

Now if you've confirmed the problem is the video card, and it's not software or a heat-related issue you can fix, your choice is to RMA the card or perhaps if you are adventurous you can try the oven trick to cook the card in the oven and reflow the solder. Or, try to sell the card on craigslist and describe the problem in the advertisement, and let an experienced miner buy the card and try to fix it himself (this is an actual market/thing, where people are willing to buy broken cards for quite a bit of money and try to fix them, just because there is a huge demand for 7950 cards). I've recently cooked a broken 5850 in the oven and it fixed the problem and has been happily mining away for 2 days now at 99% load after being baked in the oven, so I think the procedure could also work on your card. But it's way better to RMA if that's still an option, you don't get your hands dirty or spend a lot of time fiddling around with the card.

 

[Gronnie]

I will try all the suggestions this evening and report back. Thanks.

 

[Gronnie]

System was able to run OCCT GPU test and OCCT PSU test for 5+ minutes without crashing.

Tried uninstalling GPU drivers and trying older ones, still no dice.

Temps don't seem too bad, they max at 76C. Voltage never seems to go above 1.25. Still crashing to black screen in Unigine Heaven and 3DMark.

I am doing a fresh Windows 7 install now and will try again.

 

[Gronnie]

Fresh install of Windows didn't help. Still crashing in the same situations.

Going to try it in my other system to try and rule out PSU, but probably won't get around to that until tomorrow.

Oh, and I tried all the same tests with onboard video before reformatting and no crashes.

 

[Techhog]

Do you use sleep mode? I've found that my card's overclock seems to have issues after waking up from sleep mode that force me to have to reset the overclock before i can play a game.

 

[Gronnie]

So I never ended up RMAing as I have been very busy and not been gaming, and it never crashed my computer when not under load.

Decided to try Heaven and Valley benchmarks today again, and it runs them fine w/ absolutely no issues.

Any idea what the heck happened between then and now? Is there such thing as "breaking in" a GPU?

 

[Silverforce11]

I once had a faulty RAM stick, it would randomly crash during gaming but not normal usage, as games sometimes loaded enough into the ram to trigger the crashing.

Changed one stick out, fixed the issue permanently.

 

[Techhog]

What kind of temperatures are you getting? Run MSI afterburner or GPU-z log to find out what it is hitting. Pay specific attention to the VRM temps. Also with GPU-z you can see the voltage that it is running at load. Please note this as well.

Fire up OCCT for 5 minutes (don't do it for longer, no need), and report back.

Thanks!

Don't look at the maximum voltage unless the reading actially makes sense though. At least for me, that's bugged.