I've encountered some isues with my setup lately and run the FurMark GPU Stress Test to test my graphics card. When running the FurMark in Fullscreen with 1920x1080 resolution the temperature of the GPU is about 70°C (where the cards max temperature is 98°C according to nvidia) and the water cooled cpu's about 50°C. Now here's the thing: When the furry donut is rendered, the stress test runs stable, but if i hit space to "toggle furry object rendering", the gpu makes some squealing noises, the temperature drops slightly and next the system shuts down completely. No blue screen, just a power outage. Even though the system logs are enabled, no log is written out. That doesn't make any sense to me, can somebody please tell me what is going on in the FurMark algorithm, when the fur rendering is toggled? As far as I understand the situation the fur rendering is quite an expensive technique, so by disabling it the card should have actually *less* to compute, why would it crash then, and not under maximum load? As the temperature should not be an issue, I first suspected my PSU but it is an Corsair VS650 (650W power supply), the gtx only requires 500W and the wattmeter I installed never shows a power usage > 350W for the whole system. I am absolutely clueless about what is going on here. If anyone has any idea what else I could try to diagnose the problem or has any insight about what FurMark is doing when the fur rendering is disabled, please let me know.
(Windows 10 64Bit, GTX 970 with newest whql driver, I7 4790K, MSI 7917-001R Mainboard GAMING 5, PSU Corsair VS650)