I've had this PC for half a year, and the specs are:
CPU: Intel i5-4670
GPU: Sapphire R9 290 (Came with stock cooler, but I installed the Gelid Icy Vision Rev.2 custom cooler on it)
PSU: Corsair CX750 (Non Modular)
Motherboard: Gigabyte B85M-HD3
HDD: WD Caviar Blue 1TB
OS: Windows 8.1, with the latest software updates.
My case has acceptable ventilation, Got 1 front intake, 2 top intakes and 1 back exhaust, and I am getting somewhat acceptable temps when gaming/stressing - My GPU core doesn't go above 70C, CPU cores usually max at 60C and GPU VRM1 is maxing at 80C and VRM2 at 65C. (One is notoriously hotter).
The problem is that my PC never actually manages to reach those max temps, because it immediately restarts when the GPU Core hits ~63-64C and the VRMs hit ~70/55. It can happen in ANY game, even on less demanding ones like Dota 2. (Although more common on more demanding ones like BF4 or Crysis 3).
I've never had this problem before, although I did have the famous black screens on my R9 290 (I should have RMA-ed before I installed the cooler, I didn't give one black screen much attention), but I've never had then since this problem began (those were caused my the faulty Elpidia memory overheating, I assume), probably because my PC never actually survives long enough for the VRAM to overheat.
I've ran Furmark and the PC restarted within a minute, as soon as the core temp hit 64C. I have no idea why on that specific temp - Maybe it's just a coincidence, and something different malfunctions.
I suspect that it's 99% the GPU, but I have no idea what else can cause this issue.
Any help and advise on the topic would be appreciated!
Run something really stressful on the onboard graphics,and you can rule the graphics card in or out.
After playing around for 2 days by changing voltages, power limits, clocks and bioses (to no succes) I decided to give up and put my stock cooler back on for the RMA. I removed the card, and was able to run 10 runs of Valley with my onboard HD4600 maxing at 56C - a perfect indication it was the card's fault. While it was running, I was assembling back the stock cooler. I tried one last time with the stock cooler - and holy crap, it worked!
Was able to survive through Unigine Valley 10 runs with Extreme HD and 5 mins of Furmark. The VRM1 temps are significatly lower (maxing at ~60C, VRM 2 is 80C), and although my core is hitting 95C, the card actually fucking works.
My main explanation to the restarts is the memory overheating and crashing because of the TINY heatsinks on the VRM and VRAM chips that came with the Gelid cooler. Add in the fact that the cooler kept pushing the hot air around the card and case, and you have a recipe for disaster.