[SOLVED] 5900x + ASUS ROG Strix X570-e = BSODs and reboots in IDLE after exiting games/stopping burn-ins

[tim716]

Hey everyone!
Just made a new system and having problems with it.
The specs are:
AMD Ryzen 5900x
ASUS ROG Strix X570-e, BIOS 2816 (tried 2802 and 2812), default settings
Samsung EVO 970 Plus, tried 1st and 2nd M.2 slot
PSU Cooler Master 1000W, then changed it for Corsair MX1200i, no help, so the PSU is out of suspicion.
2x16 Cruical Ballistix "Red" 3200 U4 CL16 - 2nd and 4th slots , default 2666/1.2V, tried XMP 3200/1.35V
EVGA 3080 XC3 Ultra, 1st slot
CPU Cooler Thermalright True Spirit 140,
Case Fractal Design Define R5,two 140 blow out case fans on the back wall and on the back of the case ceiling, 140 and 120 blow in fans on the side wall and on the bottom of the front wall.
Chipset drivers from AMD.

I'm experiencing BSODs and self reboots 10-30 seconds I leave a game. It NEVER happens in a game itself. It can be almost 100% reproduced by restarting a game again within 30 seconds after leaving it.
In synthetic tests, it can be reproduced by burning in the CPU AND the 3080 simultaneously (not by burning in the CPU OR the 3080 separately). The BSOD/reboot happens either immediately or within 30 seconds after stopping the burn ins. If it doesn't, starting a game within 30 seconds after the burn ins leads to a BSOD in 99% cases.

Setting Core Performance Boost to OFF eliminates the problem, but I'm obviously not happy with this solution as it cripples the CPU performance.
The temperatures in burn-ins are about 80C CPU, 89C CPU cores, 79C GPU. The chipset temperature is about 70C idle and in burn-in.
The BSOD is WHEA uncorrectable error, 124-machine check exception, authenticAMD.sys, 0x124_AuthenticAMD_PROCESSOR__UNKNOWN , no futher details in the dump.

I have no computer workshop or spare components to check by exchange, so I'm asking for your advice to resolve the problem.

 

[tim716]

Ok, I probably solved my problem. I'll describe the solution here as it's probably a common issue and this may help other unlucky Ryzen 5600x/5800x/5900x & x570 chipset owners.
Just to remind you, my problem specifics were crashes in transitions from loads to idle, in idle within 30 seconds after a transition, or when applying a load again after these 30 seconds. NEVER under a load. Turning the Core Performance Boost off eliminated the issue together with the CPU performance. If you're in the same boat, try this, it should help.
The following is for the ASUS bios, for other vendors the same parameters may be hidden in another place.
The system is stable so far with the following BIOS settings:
Go to AMD overclocking, set the Presicion Boost Overdrive to Manual. Some additional parameters will appear. In there:

1. (The main thing) Set the EDC current limit to 200A.
2. (Just in case) Set the power limit to 130W.
3. (Just in case) Set the temperature limit to 83C.

1 is an increase, 2 and 3 is a decrease. Leave at zeros all the rest there.
Also, just in case, set Idle Voltage to Typical, Global C-states control to Disable, check that ECO mode is Off. Then you can set Core Precision Boost back to On, everything should work.
Looks like the MB and its BIOS wasn't tested with a 5000 CPU at all (or, if it was, it was like "Ok, it boots, that means it works, great, the job's done), and the BIOS just doesn't know about the larger peak currents of Rysen 5000s, and the BIOS' "digital fuse" is just too small for a new CPU. When changing its clocks the CPU, probably, tries to draw more current, the "fuse" (EDC current limit) kicks in and the CPU malfunctions and produces a BSOD.
These currents (or how the "fuse" works) also definitely depend on the MB and/or the CPU heating (I didn't have any BSODs when cooling the open case with a cold hair fan), that explains why not everyone with the config like mine has the same problem, people with better cooling (or a colder GPU) might be ok at defaults.

 

[drivinfast247]

Ok, I probably solved my problem. I'll describe the solution here as it's probably a common issue and this may help other unlucky Ryzen 5600x/5800x/5900x & x570 chipset owners.
Just to remind you, my problem specifics were crashes in transitions from loads to idle, in idle within 30 seconds after a transition, or when applying a load again after these 30 seconds. NEVER under a load. Turning the Core Performance Boost off eliminated the issue together with the CPU performance. If you're in the same boat, try this, it should help.
The following is for the ASUS bios, for other vendors the same parameters may be hidden in another place.
The system is stable so far with the following BIOS settings:
Go to AMD overclocking, set the Presicion Boost Overdrive to Manual. Some additional parameters will appear. In there:

1. (The main thing) Set the EDC current limit to 200A.
2. (Just in case) Set the power limit to 130W.
3. (Just in case) Set the temperature limit to 83C.

1 is an increase, 2 and 3 is a decrease. Leave at zeros all the rest there.
Also, just in case, set Idle Voltage to Typical, Global C-states control to Disable, check that ECO mode is Off. Then you can set Core Precision Boost back to On, everything should work.
Looks like the MB and its BIOS wasn't tested with a 5000 CPU at all (or, if it was, it was like "Ok, it boots, that means it works, great, the job's done), and the BIOS just doesn't know about the larger peak currents of Rysen 5000s, and the BIOS' "digital fuse" is just too small for a new CPU. When changing its clocks the CPU, probably, tries to draw more current, the "fuse" (EDC current limit) kicks in and the CPU malfunctions and produces a BSOD.
These currents (or how the "fuse" works) also definitely depend on the MB and/or the CPU heating (I didn't have any BSODs when cooling the open case with a cold hair fan), that explains why not everyone with the config like mine has the same problem, people with better cooling (or a colder GPU) might be ok at defaults.

Thanks for providing a possible work-around for people having issues such as yours.