Over the past few weeks I’ve been having issues with my aging gaming PC just freezing up entirely. The only recovery was a one finger salute.
No, I didn’t flip my computer the bird, I used one of my fingers to perform a hard shut down by holding down the power button. Weirdos.
At first I thought it was a software issue. Maybe a corrupt driver somewhere. So I reinstalled Windows 7.
No change, still this phantom lock-up. So I knew it had to be hardware. Out of time for the day, I decided to pack it in and save the diagnostics for the next day.
The next day, the system won’t boot. Apparently the disk that I installed Windows 7 to took a dump. Can’t detect it. Awesome.
So I pull another drive I had on my desk and install Windows 7, again. I thought that maybe it was a bad disk so I went apeshit reinstalling all of my software. Played games on it all day. I left Steam to download my game library overnight.
The next morning my computer is locked up. Not the bad hard drive that I replaced. Awesome. My top two suspects are a bad RAM module or my old GPU a GeForce GTX 460. My money is on the GPU.
I have to test the memory first to remove my RAM chips from the running because a GPU stress test will cause a potentially faulty RAM chip to fail.
Testing the RAM with Memtest86+
Time for memtest86+. I like memtest. It has been the standard in memory testing for well over a decade. While it’s primary goal is to determine issues with system memory, it can also indicate issues with the motherboard or CPU. One pass with no errors indicated 99.9% certainty that it is not a memory issue with my gaming rig.
CPU Stress test with Prime95
On to Prime95. This software stress tests the CPU and underlying hardware by running heavy calculations on the CPU. The default settings are sufficient to determine if a CPU is faulty. I’m monitoring my hardware temperatures with Open Hardware Monitor to make sure nothing overheats.
There is a lot of debate on how long to run Prime95 to determine if the CPU is faulty. Some say 4 hours others say 24 hours. The thing is, run it until you feel comfortable. 4 hours is a decent goal to shoot for but no matter how much Prime95 is run, absolute 100% certainty that the CPU is not faulty will never be achieved due to the nature of most CPU faults. However, 99.9% (just like memtest86+ before) will be achieved.
Torturing the GPU with Unigine’s Valley and FurMark
Both of these pieces of software do the same thing. Unigine Valley just looks nicer.
FurMark puts 100% stress onto the GPU itself. In my case, it did not max out the GPU memory, which is something that I’d like to test. Regardless, I ran FurMark for 30 minutes, which should be enough time to determine if the GPU is bad.
I was able to push my GPU memory to about 75% utilization with Valley compared to about 25% with FurMark. Still not completely utilized like I had hoped. I let it run for 30 minutes with no problems.
More GPU Testing with EVGA’s OC Scanner
Designed more for the overclocker of graphic cards, EVGA’s OC Scanner comes with a variety of tests that allow one to benchmark their graphics cards for different type of rendering scenarios to include PhysX. I ran each test for 15 minutes with no faults found.
Where do I go from here?
Since all of my hardware checks out okay with the methods described above I now have to look into perhaps a software issue.
Diagnosing a system freeze-up like this that may be caused by software can be troubling mainly because this is such a random issue.
I run Overwolf so that will be my first suspect. I’ve had problems with it in the past but nothing this massive. I’m going to set it to start manually instead of when Windows starts and go for about a week or until the system freezes again. If it freezes without Overwolf, then I know it isn’t Overwolf and I’ll move onto the next suspect on my list… Battle.net…
Anywho, that’s it for now. At least I feel more comfortable with my hardware’s stability.