Originally Posted by The Pook
Don't use Furmark and stop suggesting people use it. It serves 0 purpose and most cards will cripple themselves to a lower P state if it detects it's running. Just run a benchmark if you want to test things out.
Furmark will max out power consumption at P0 on most recent modern GPUs without dropping down a state. This is the best way to test if there is an issue with power delivery, e.g. underpowered PSU. It's also very useful for VRAM artifact testing. Considering the OPs card failed 5 seconds into the test on a warm boot implies something
is wrong, and by reducing core offset and VRAM offset, they can eliminate those from the equation.
Caveat, it is less useful for general stability testing however - synthetic demos and games will be better for that.
, those green sections are the definitely the VRAM modules. Disassemble the backplate and card and ensure that the thermal pads are mating with the VRAM modules, when removing them they should have rectangular shaped indentations if they have been mounted. Unstable VRAM can cause kernel panic and driver lockup on the 980 Ti, especially Hynix memory, the fact that it occurs after a few hours means something is heating up to steady state, unstable core clock would result in a watchdog timeout, not a hard reboot.
CPU: i7-5820K @ 4.3Ghz (-0.08V/1.33V Adaptive = 1.25V); Cache: 3.9GHz (1.20V); VCCIN: 1.85V @ LLC 6 12hr x265 (AVX2) Stable
CPU: i7-6950X @ 4.0Ghz (-0.08V Offset = 1.22V); Cache: 3.2GHz (1.06V); VCCIN: 1.90V @ LLC 5 12hr x265 (AVX2) Stable
RAM: 64GB @ 3000MHz (14-14-12-31 1T, 1.35V) 12hr GSAT Stable
GPU: EVGA Hybrid GTX 1070 @ 2076MHz Core, 8.6GHz VRAM, Stock Voltage 24hr [email protected] Stable