Fatal CPU Error

[Weric]

Hey folk, new 5900x installed, temps are good, everything good. Recently I set a curve in PBO to a negative offset of 30(the max). The next couple days, playing games and being idle were fine. Today I booted up my PC and when I went to launch web browser, I had a restart. The error was fatal hardware error, cpu core. I lowered the offset to 25 instead.

 

Think this is from my undervolting? Fresh windows with the basics installed, no overclock or nothin, just had the PBO undervolt curve set to 30 which was fine for a couple days playing some games. I didn't do any long benchmarks either

[Skiiwee29]

Possible. You should do some benchmarking on the CPU for like 30m to an hour to ensure stability. It likely was unstable at the negative 30 offset. I know my 5900x is unstable at the 30 offset and I have to run at 25. 

[YoungBlade]

4 minutes ago, Weric said:

Hey folk, new 5900x installed, temps are good, everything good. Recently I set a curve in PBO to a negative offset of 30(the max). The next couple days, playing games and being idle were fine. Today I booted up my PC and when I went to launch web browser, I had a restart. The error was fatal hardware error, cpu core. I lowered the offset to 25 instead.

 

Think this is from my undervolting? Fresh windows with the basics installed, no overclock or nothin, just had the PBO undervolt curve set to 30 which was fine for a couple days playing some games. I didn't do any long benchmarks either

I, too, have a 5900X, and I, too, have been experimenting with PBO undervolting. I decided to test negative 30 and everything seemed fine until I tested using a Windows Defender custom scan on one of my NVMe drives. I like that as a hardware test because it's a real-world workload that brings the CPU up to 90-100% usage on all cores, but doesn't have a lot of AVX stuff, so the boost clocks can go much higher than when running something like Cinebench. I was monitoring things with HWInfo64 and it looked pretty good. Temps were lower, clocks were higher. But then something odd happened: the temps on CCD1 (Tdie) jumped to over 86C out of nowhere. A few seconds after that, the computer straight up shut off.

 

So yeah, just using the Curve Optimizer can cause system instability. It's one of the only risks of undervolting. I've now lowered it to 28 and it's seemed to pass all of my tests so far.

[Weric]

2 minutes ago, Skiiwee29 said:

Possible. You should do some benchmarking on the CPU for like 30m to an hour to ensure stability. It likely was unstable at the negative 30 offset. I know my 5900x is unstable at the 30 offset and I have to run at 25. 

Thanks. After doing some digging I'm seeing a lot of folk have this issue with their 5950x and 5900x, the majority of them were using the curve optimizer, so I'm leaning towards -30 not working out very well despite it being fine playing some games and running like half a cinebench test lol.

 

Also a good bunch of folk on reddit with 5900x were getting these crashes on fresh installs with all STOCK settings and actually had to put a positive offset of +4 in curve optimizer and it got their systems stable. Weird, hopefully setting it to -25 solves it though

[Weric]

1 minute ago, YoungBlade said:

I, too, have a 5900X, and I, too, have been experimenting with PBO undervolting. I decided to test negative 30 and everything seemed fine until I tested using a Windows Defender custom scan on one of my NVMe drives. I like that as a hardware test because it's a real-world workload that brings the CPU up to 90-100% usage on all cores, but doesn't have a lot of AVX stuff, so the boost clocks can go much higher than when running something like Cinebench. I was monitoring things with HWInfo64 and it looked pretty good. Temps were lower, clocks were higher. But then something odd happened: the temps on CCD1 (Tdie) jumped to over 86C out of nowhere. A few seconds after that, the computer straight up shut off.

 

So yeah, just using the Curve Optimizer can cause system instability. It's one of the only risks of undervolting. I've now lowered it to 28 and it's seemed to pass all of my tests so far.

Thats really weird. Yea my PC didn't BSOD either it just straight up rebooted and I saw the event viewer. -25 seems to be working for now, hopefully its sorted out. I really hope it does though, I'd rather not go below -25 cuz I don't see much reduction in Temps when I'm at -20.