Video card crashes when gaming

[occational_gamer]

Hello,

My monitor enters stand-by when gaming. It happens with a variety of games, but I'll use Dead Space as an example.

Specs: Windows 7, Intel E8400 @ 3GHz (stock), Asus P5K-E, 4x1GB OCZ ram, Sapphire 5770 1GB graphics card, 4 SATA disk drives, Antec NeoHE 550W PSU.

When playing Dead Space with V-Sync on, the load on my GPU is about 30% and I get low frame rates (this is a known general issue). However, I can play as long as I want. I can also play older games (e.g., Doom 3) without instability.

When playing Dead Space with V-Sync off, I get high frame rates. But after about 3 minutes of game play, my monitor enters stand-by. Audio continues for about 10 seconds more, before the system becomes unresponsive and requires a hard reboot. Logs from monitoring programs indicate (just before the crash) my CPU load is below 70% and temperature is below 50C; my GPU load is almost 100% and temperature is below 65C.

This problem has been occurring for quite a while. But because I don’t game often, I’ve put off trying to fix it. Heat does not seem to be an issue. My power supply provides 18A on the 12V rail connected to my video card and disk drives (it provides a total of 42A over all three 12V rails). Several MemTest passes yield no memory errors. I've downgraded and upgraded drivers, but with no change. I tried an older video card (Radeon 2600xt) and got similar results. However, that time, I was able to game for about 15-20 before crashing.

I’m not sure what the problem is or how else to troubleshoot it. I’d welcome any advice. Thanks in advance.

 

[vTL123]

Welcome to Toms Hardware!

This sounds quite similar to my case before I upgraded the powersupply. The power from the HD5970 was being loaded from just 1 of the 5 12v Rails and each rail only provided 18 amps. So I used an 8 pin power connector from one rail and a 6 pin connector from another rail to solve the problem.

But seeing that the HD5770 only has 1x 6 pin power connector, it is possible that the load level is being focused on less rails so when the computer has a high load on both video card and CPU, this may overload the PSU 12v Rail.

If you have any other 6 pin connectors, try swapping them over with the HD5770 and test using OCCT as it loads both CPU and Video Card.

 

[occational_gamer]

Thanks, vTL123.

My PSU has 3 12V rails: one for the CPU, one for the motherboard, and one for the PCIe and peripherals. So unfortunately, even if I use a different PCIe power connector, it'll be on the same rail.

Thank you for recommending OCCT. A friend also recommended Furmark (for OpenGL testing). Neither of the stress tests lasts longer than 3 minutes before the screen goes black. I've also used Prime95 and Memtest to eliminate the CPU and memory as culprits.

If my PSU is to blame, how can I test it?

 

[vTL123]

I think you may have overloaded the 12v rail powering up your Graphics card/HDDs/peripherals. See if you can use as little peripherals/HDDs and then run OCCT/Furmark again. After that you try swapping out the HD5770 with the 2600XT and then test it again to confirm.

If the overloading of the rail was the problem, then you'd have no other choice but to limit the amount of stuff connected to that rail or change the power supply to a different unit.

 

[occational_gamer]

I've run the test with only the video card and a hard drive, but the video card still crashes within three minutes.

Some PSU reviews mention that their performance can be affected by temperature. It's been rather cold these past few days (room temperature of about 17C). Could this contribute to the PSU not supplying sufficient power to the video card?

 

[vTL123]

Is the Video Card at stock settings and haven't been overclocked/altered?

What you experienced is quite similar when my overclocked settings for the HD5970 aren't stable. Back before I solved my video card power, the system would shut off instantaneously without warning and reboot.

I'm still very sure it's your power supply since it shuts off with both HD5770 and 2600XT but bugs me that it freezes up with the extra peripherals removed. If it is possible, run Furmark at the lowest resolution/settings in a windowed screen.

I hope you mean 17C Celsius and not 17C Fahrenheit. It should always work at 17C Celsius and not a problem should occure at those temperatures. 17C Fahrenheit? Well...

 

[occational_gamer]

The video card is at stock (as is everything else in the system).

I ran Furmark at 640x480 and the video card crashed again within three minutes. However, I think I've made some progress. I swapped the PCIe power cable from my (modular) PSU. I can now complete OCCT and Furmark (I stopped the tests after 10 minutes). I'm going to continue testing/gaming, but I think I've found the culprit.

vTL123, thank you for all your help and advice.

 

[vTL123]

Your welcome!

I never really took into account that the power cable would be the culprit, anyway, happy to hear that your PC is now up and running!

 

[occational_gamer]

Best answer selected by occational_gamer.

 

[Mousemonkey]

This topic has been closed by Mousemonkey