Odd instability issues on new Ryzen machine

[StergiosTh]

Hi everyone, I have recently built a new PC, and I have been getting some odd instabilities when overclocking. Basically what happens is that from time to time my computer will crash and restart itself when it's just sitting on idle, or it will be unable to boot (CPU&VRAM leds light up), but doesn't show any instability in gaming or stress testing. I will go into more detail later in the post. What makes troubleshooting harder is that the problems don't occur often, so after tweaking something, there might not be any problem for days at a time, until it crashes or is unable to boot again. I put the computer together something like 3 weeks ago, and I have been gaming quite heavily on it. I have never encountered a crash or error while gaming.

Specs

• Ryzen 5 3600xt with Noctua NH-U12S

• Asrock B550 Phantom Gaming 4

• Corsair Vengeance LPX 16GB (2x8) 3200mhz

• Corsair RM850, 80 plus gold, 850W

• Samsung 960 EVO NVME SSD

• RX vega 64 Sapphire Nitro+

 

Troubleshooting so far

• While having my CPU OCed+XMP, I have run the following stress tests and benchmarks multiple times: prime95 for 2 hours, 1 hour long occt CPU and RAM stress tests, cinebenchR23 and 3dMark, as well as memtest86. Also tried AIDA64. These all always come up error free, with the exception of prime 95, where I am sometimes getting errors on, but I just assume I haven't found the perfectly stable all core OC yet. It has never crashed during any test or gaming.

• Reverted my CPU back to stock clocks, ran the same tests, no issues (including prime95), but the XMP alone seemed to (rarely) cause the aforementioned instabilities on boot and idle.

• Raised the ram voltage to 1.4. Had the same issue

• BIOS is updated to the latest version

• Have also tried using CTR 1.1 to see if would provide me a more stable OC, but it can't go through the process without crashing my system

• Have been checking the event viewer for system errors, there are occasionally WHEA errors (CPU related, event ID 19), but again, never while gaming or stress testing, only on idle

• Have been monitoring thermals, clocks, voltage and everything seems to always be normal

• Have now reverted my RAM back to the default settings too and haven't encountered a problem, but even with the OC active it might take 4 or 5 days before anything goes wrong

 

I have been trying to troubleshoot this on my own for weeks now and I have yet to encounter someone with a similar problem. I just simply cannot explain how I haven't encountered any issues under heavy load, either synthetic or gaming, and then sometimes I have problems on idle or during booting. At the end of the day I don't care about the small performance increase achieved by the overclocking, I just want to make sure all my hardware is working properly. Oh, and in case you are wondering about the vega 64 and the 960 EVO, they are just some parts I got from a family member for free.

Thank you for taking the time to read this and for any help.

[Master Disaster]

Check what voltage the SoC is getting, its generally in the same area of the UEFI as CPU Vcore. It will probably be at 0.9v to 1.0v on Auto, try bumping it up to 1.1v (you can go up to 1.2v safely).

 

Also standard Ryzen new owner PSA, don't bother trying to OC your Zen 2. You won't be able to go any further with a manual OC than Boost can already achieve automatically, all you are doing by setting an all core OC is removing the CPUs ability to do single core boosting. (You can watch LTTs "Don't build a Hackintosh" video to see this in action). The only thing you should be enabling is D.O.C.P, leave everything else on Auto.

[StergiosTh]

1 hour ago, Master Disaster said:

Check what voltage the SoC is getting, its generally in the same area of the UEFI as CPU Vcore. It will probably be at 0.9v to 1.0v on Auto, try bumping it up to 1.1v (you can go up to 1.2v safely).

 

Also standard Ryzen new owner PSA, don't bother trying to OC your Zen 2. You won't be able to go any further with a manual OC than Boost can already achieve automatically, all you are doing by setting an all core OC is removing the CPUs ability to do single core boosting. (You can watch LTTs "Don't build a Hackintosh" video to see this in action). The only thing you should be enabling is D.O.C.P, leave everything else on Auto.

Thanks for the answer! I will try raising the SoC voltage and reactivate XMP and see if I get any issues.