Hardware Failure: 13700k CPU errors started out of the blue, especially while using the GPU

[Yagma]

I run a 3080-ti on Gigabyte z690 Gaming X mobo, 13700k with Deepcool LS520 - 240mm liquid CPU cooler. Corsair RM 850x 80+ Gold 2019 PSU. Been fine for 16 months. Using latest bios/intel/nvidia drivers.

Had no problems before. No overclock... the crashing began with Cyberpunk black screening, sound would continue stuttering with black screen; keyboard unresponsive; forced to do a reset. No error messages or dumps would take place; nothing in event viewer.

1. Ran memtest86, no errors.
2. Ran kombustor all night on the videocard, no errors...
3. Ran OCCT linpack multiple times, no errors...
4. Ran linpack with kombustor / furmark, CPU would error within minutes, this happened time and again. Average VRM MOS only around 58 degrees, PCH 46, CPU Package 69-73

After a bit of research I started to think it could be the PSU.

5. Ran OCCT "CPU" stability test instead of linpack in OCCT, this time without furmark/kombustor and errors began happening regularly, each time on different physical cores with power draw only up to around 170 watts. PCH 34 degrees, VRM MOS 57 degrees, Cpu 68-71. Errors generally take around 20 minutes to occur in this specific test. But, if I run furmark/kombustor at the same time that is cut down to around 2 minutes on average. No errors in furmark/kombustor, just cpu.

Other than for RAM, I'm not running an overclock. No motherboard boosts or enhancements. The only thing I did to the CPU in bios was disable C-States for individual cores, but left it on for the package... this was to reduce coil whine. This I ran for months without issue. And I also under-volted the CPU to reduce heat. But this undervolt I tested with and without during stability testing, both resulting in errors.

Evidence suggests to me is leaning towards a dying cpu. But, the fact that running a GPU bench simultaneously increases the crash frequency significantly makes me wonder if it could be a motherboard power delivery issue, or PSU issue... what are your thoughts?

Are there any VRM BIOS settings you would suggest I use for testing that could possibly help further narrow this down? Or in Throttlestop?

 

[ir_cow]

You said its undervolted and C-State disabled. Clear CMOS / Set everything back to defaults and see what happens. That is where I would start.

 

[Yagma]

You said its undervolted and C-State disabled. Clear CMOS / Set everything back to defaults and see what happens. That is where I would start.

Looks like it is the ram OC, or more likely its voltage. I decided to default that.... I passed the first 30 minute OCCT test (CPU+Kombustor) that usually errors after 2-4 minutes. It was only a small bump up to 3800 from 3600; but the latest motherboard bios' reduced DRAM power efficiency significantly, requiring a bump up from 1.38 to 1.42 volts to remain stable in memtest, which is pushing it a little extreme for 1.35 rating; Unfortunate because the previous bios' were able to accomplish this OC at 1.38v. I contacted Gigabyte technical support about this and they just sluffed it off.

I had to RMA these same sticks last year because of failure and that was running them at a lower voltage.

Anyway, I re-enabled Cstates and disabled the ram OC, and dropped its voltages to default. I will test again shortly with ram as is, and C-States disabled*.