Failed Debugging steps taken: - Tested different video card as the vga driver apologized for the crash (Radeon settings have been restored due to unexpected power failure) - Tested different PSU (it was a power failure after all?) - Set RAM clock to 2133MHz (DDR4 spec) - Tested older BIOS version - 12 hour CPU stress test using CPU-Z stress (No crash during the run) - 12 hour GPU stress test using Furmark (No crash during the run) - 3 hour RAM error test (no crash) - 6 hour combined CPU/GPU stress test (no crash) - updated/reinstalled drivers for everything
Currently testing: - High performance Windows powerplan
System bahaviour: CPU idles arround 33°C (@ 20°C room temp) with temperatures fluctuating into low 40s GPU idles at 38°C During load, CPU reaches 55°C and GPU caps out at 65°C
Software that causes the most crashes: - Heroes of the Storm - Sony Vegas 13
Notes: I am a bit lost as to what to do now. Halp!
This is a case of crazy off ram timings. Not something you can fix with more voltage. A lot of boards have this issue especially since the Agesa 1006 update where you will encounter random machine check exceptions leading to an immediate reboot.
Updated RTC 1.01
EDIT: RTC Removed. It will eat your PC and your Kids.
Only use those values with the T behind them, these are the Ratios your board uses.
The other values are the actual parameters and time in ns.
TRCD is used twice once for the TRCDWR and TRCDRD. Pay particular attention to tFAW and tRFC
In addition to that you want to set GearDownMode Disabled and set ProcODT to 60ohms. Stock for Ryzen ProcODT is 53.3Ohms however 60Ohms helps a lot with getting an overclock stable.
Weirdest thing so far is Stock+XMP suddenly crashing RAM after toggling things like nope that one did not work, back to bios and toggle it back, aaaand now RAM just refuses to work like after Taichi has tried its best it has decides to set it default.
Then did reset default settings for bios and Stock+XMP was again just fine.
I'm confident that I did not toggle extra things by myself, like either 1. Something toggles another setting but doesnt dial it back the same way 2. Bug
Anyways, nice that RAM disables that XMP profile / goes default if its not working.
Ok. I'll start listing the parameters to change and update this post accordingly.
Ok so first thing:
Lets set your RAM to 2933Mhz. I know that's something your board should support. Currently I still see it sitting at 2133Mhz.
Then set the CPU Load Line Calibration to Level 3 (much better than Level5)
Also set the VDDCR_SOC Load Line Calibration to Level 1 and Set the VDDCR_SOC Voltage to the lowest offset 0.00625 or some such value. You can also try going one offset higher to help stability of the memory controller.
Next set the Fail Count to 1 (No point retrying several POSTS if the first fails before reverting to default settings