Ive been struggling to figure out why my system kept failing hci memtest, but would always throw the same amount of errors (8). Sometimes it would get these errors under 100% coverage, sometimes it would take up to 250%. (and if I let it keep going, it would accumulate to 16 errors, 24, and so on)
I fiddled with every voltage and resistance I could, but to no avail. Then, something odd caught my eye. The default setting for cldo vddg with xmp enabled was being displayed as 1.1v as per ryzen master. I knew from research that this voltage should always be ~40-50mv lower than my soc voltage, and never higher. Since my soc voltage is displayed as 1.08v, I immediately realized that my cldo vddg was much too high, and I went into bios and lowered cldo vddg to 1.04v.
So far hci memtest has passed 1500% with no errors.
Make sure you get check that cldo voltage in ryzen master and confirm it is 40-50mv lower than displayed vsoc. If not, you can manually lower it in bios by changing vddg ccd and vddg iod. (both are component voltages of cldo vddg)
My guess as to why this is happening, the voltage mismatch resulted in a brief erroneous condition in the cpu, resulting in all 8 CCDs spitting an error each, hence 8 errors.
I took weeks to figure this out, hoping this can help somebody out there!
AMD Threadripper 3990x || 256GB G.Skill TridentZ Neo @ 3600mhz 18-22-22-42 || Asus Zenith 2 Extreme Alpha || 2080TI FE || Samsung 960 Pro 1TB || Samsung 970 EVO 4TB || Evga 1200 P2 || Meshify S2 || Custom Loop
Last edited by happyluckbox; 07-21-2020 at 03:03 PM.