Here's a little story :)
The issues I described began after I flashed BIOS 2.09 so I tried downgrading to previous versions to see if the problem would go away.
Before I started tinkering with BIOS updates and having boot problems, I had a 6850K running on 2.03. The night before swapping out to my current 5960X, I ran Memtest86 7.5 (UEFI) for 50 passes, 31 hours to make sure that if any issue arose, it wouldn't be RAM related.
When I first installed the 5960X, it hung on post code 95 and sometimes 96. After a single CMOS reset, it finally booted on the same BIOS chip (2.03) as the 6850K. I got into Windows to do some testing and oddly enough, the computer would auto-restart within an hour of running Prime95 without any error or BSOD. The core temperatures were all around 50C, same thing for the motherboard VRMs. My PSU is a Evga Supernova 1200 P2 and just a few days earlier I was running my 4.3GHz 6850K for 24H+ in some custom AVX/FMA3 small FFT loops; that was pulling 200W vs the stock 140W 5960X.
Doubting that my PSU or motherboard had suddenly died on me, I decided to check if any BIOS updates had been released in case they fixed some kind of compatibility issue. For security reasons (spectre etc.), I wanted to flash the latest version anyway.
After I flashed 2.09 and every single BIOS down to 2.02, the computer would remain stuck at post code 95/96 regardless of CMOS resets and restarts. I had to flash 1.12 or 1.20 to get those BIOS chips bootable again, anything higher would bring back the problem. This whole thing started looking like some low-level software glitch more than a hardware problem, especially considering that 2.03 worked and then no longer worked when I reflashed it.
Not wanting to take apart my 2 water loops, I did the next best thing. I disconnected literally everything that connects to the motherboard except for 1 stick of RAM, my GPU and the case power/reset/LED wires. I also flipped off all but 1 PCI-E switch.
I then attempted a first boot on 2.09 which got stuck at the usual 95/96. I did a single CMOS reset and finally, it booted all the way into the BIOS. I repeated this procedure for the 2.03 BIOS chip and that worked too.
I then reconnected everything I disconnected and still booted successfully into both BIOS (2.09 and 2.03). I just finished running a 2-hour loop of Prime95 FMA3, 112K in-place and the computer did not auto-restart. This specific test auto-restarted after like 20 minutes earlier.
All in all, this leaves me with a few questions:
1- it seems like the CPU and motherboard struggled with their initial "handshake". Does anyone have an idea of what could have gone wrong that got magically fixed after stripping the computer to barebones? No component other than the CPU was changed during this upgrade.
2- although I haven't run Prime95 for more than 2 hours yet to completely verify that the computer no longer auto-restarts, if it is truly fixed then something weird was going on at a BIOS/CPU level. The computer would no longer boot even after reflashing the original bootable version (2.03). It's as if something at the BIOS level persisted from the 6850K to the 5960X which allowed the computer to boot but also made it unstable (auto-restart under heavy load). The moment I flashed a BIOS using the 5960X, that persistent information was gone and the computer would no longer boot until going barebone. When it did finally boot, the instability was gone. Anyone wants to speculate on this?
3- one thing I did add during the upgrade was the Classified IO cover. In one thread, someone recommended removing it without mentioning why. Does anyone know why someone would recommend removing it? Does it pose some kind of risk? (it is made of metal)
Thanks!
post edited by H-Emmanuel - 2019/11/28 14:47:13