Difficulty diagnosing inconsistent freezes in Time Spy Stress Test

[JumbledThought]

I recently got some new hardware and I've been using Time Spy to diagnose some crashes and verify everything is on the up-and-up. I can run the benchmarks just fine with a decent score whether I'm at default GPU speeds, under-clocked or overclocked. However, I've been having very inconsistent freezes in the stress test. I've successfully run it a few times and I've had it freeze 15 minutes in, I've had it freeze 5 minutes in at low temperatures. I've had it crash and pass with two different PSUs. The freezes are always the same - frozen, unresponsive screen that requires a hard reboot using the reset button on my case. No indicator in Windows as to what failed, just that "unexpected shutdown" notice later on.

What more can I do to figure out why the stress test is crashing? I can run Furmark for ages (at least 25 minutes reliably) so I don't feel like my hardware is bad. Is this stress test reliable enough that anyone is able to run it successfully 10 / 10 times? If it's just DirectX 12 being buggy, I'm not so concerned since it's not my hardware crapping out on me and an RMA isn't going to change things.

Hardware: EVGA Z170 Stinger motherboard, Zotac 1070 Ti AMP!, Intel i5 6600K, Cooler Master V550 PSU, 16GB RAM

Thanks for any light you can shed on my problem.

[UL_Jarnis]

Furmark is not a very good stress test since it stresses only a very small portion of the GPU (shaders). Yes, it can find temperature-releated issues since it tends to peak the GPU temperature hard, but if portions of the GPU not stressed by Furmark have a problem, it would never see it.

Hard freeze is almost always a hardware issue. Sadly no easy way to determine what is the cause without spare components. Would need to test this GPU on another system, or another GPU of similar power use in this system. Freezes could be due to the GPU, due to RAM / RAM settings, motherboard, PSU (tho you said you tried two so that probably rules out PSU).

[JumbledThought]

Thanks for the info, I was wondering precisely that - whether Time Spy and DirectX testing would reveal different hardware flaws than OpenGL testing. Sounds like I should kick off an RMA since I was able to rule out the PSU as the freeze-testing. I've had the motherboard and RAM for two or three years and the crashes only started when I got the new PSU and GPU.

[TikTok-va02gaming]

digging up a old post here but i am having the exact issue you had i know it was a while ago but i have a x670e-f with a 7800x3d and a 4090, what your symptoms are is exactly what happens with mine, i have tried 2 different cpus both 7800x3d, 2 different ram kits, 2 different motherboards, 2 different 4090s and 2 different psus and literally timespy extreme stress test will crash every time! not the benchmark but the stress test, out of about 10 runs i have had it pass once, i haven't had it in windows or games "touch wood" did you ever solve the issue? like i say i have tried literally double of everything hardware wise the only thing i havent tried is installing windows on another nvme drive but i have tried running timespy from a diffrent drive and still the same result. like i say i guess i dont have faulty hardware as 2 sets of eveything surely cant be faulty lol the only other thing i havent tried is a older drive incase its that maybe? honestly i have no idea.

[nuggdoctor]

could be a bios issue with the gpu. when i would have failed runs or crashes it was usually because my drivers were out of date.

[UL_Jarnis]

Note that you should also try to eliminate all background programs. Some things that draw overlays or otherwise mess with the GPU driver can sometimes cause crashes. Before swapping hardware, it is a good idea to close absolutely everything nonessential.

Also it is a good idea to try to reset all motherboard BIOS settings to stock and ensure motherboard BIOS is up to date.

In your case also ensure your PSU is actually strong enough for all that hardware. Granted, usually too weak PSU results in a shutdown/reboot and not test crash. Test crash is generally either faulty GPU, unstable RAM (memory test would find this) or background program that is misbehaving and crashing the GPU driver.

[JumbledThought]

In the end, I returned that Zotac for an RMA, sold the fresh card on eBay, and never purchased a Zotac again. Got an MSI card that passed that test just fine. I think it was just a card that passed crappy QA. It often worked fine but that 3dmark test killed it and it would rarely, randomly do the same thing when playing games.

Sounds like you have the best possible testing setup with two of everything. Are the two cards the same make / manufacturer? Maybe you got two of a bad batch that came off the line next to each other.

Good luck!