Originally Posted by neurotix
1st guy: Yes, 8d is Fclk related. I saw this a lot when originally getting this setup, trying to overclock to 3800/1900, and not using Thaiphoon Burner properly (and trying to do a V1 profile in Ryzen DRAM Calc with the built-in profile). This resulted in way too aggressive timings for my kit, which is only a 3200c14 kit. Consequently, I would fail to post and usually see 07, 15, 22, and eventually a hang on an 8d code no beeps. Rebooting would give me 8d~F9 and beeps and then the board would reboot and proceed with F9 recovery procedures, resulting in the board posting in safe mode @ stock settings.
I have also given up and gone back to 1001 via usb flashback. Maybe I'll rma this board and get an MSI. This bos and whole experience has been a dumpster fire, hot trash. I've used nothing but ASUS boards since 2008 and ROG since 2011 and never had such a horrible, buggy experience, the bios layout is bad (put all the relevant AMD CBS settings from Ryzen DRAM Calculator in Tweakers Paradise!, still can't find Boot DRAM Voltage anywhere, DRAM VRef Channel A/B is broken in Tweakers Paradise and wont increase above 0.600v, many more issues). I paid for a premium product that is $50-100 above competitors boards yet the bios and support and being behind other vendors bios releases makes me feel like I purchased a $50 Biostar Micro ATX board or something. Even my still running Crosshair V Hero from 2011 has and always had a functional bios with a much better layout- and that was FX, when no one cared about AMD and the boards and bios had about 1/8th the effort put in as Intel boards got. This reminds me of that. The other vendors are doing it right from what I can tell...
My rig will. not. post. at the exact same settings for 3800/1900 c16 and just the other day I was running 3800 cas14 and I've also done 4066/1866 @ cas16. With this new bios, when rebooting with all settings identical I get an instant F9 post code (starting recovery module), I don't see 00-09 even, it doesn't freeze on 07, 15 or 22 codes before showing F9, the board is literally going to recovery options right away. I tried the normal manual oc fast profile and the slow junk "safe" profile and both immediately show F9. I think it might have something to do with 1) cLDO_VDDG being seperated into 2 settings, 2) PCI-E 4.0 nvme drives (Is the 970 Evo compatible even? Dont think so) and VDD_SoC above 1.1v (which I set, and noticed if you turn on D.O.C.P. it warns you about increasing SoC voltage above that corrupting 4.0 NVMe drives but again, I dont think mine supports 4.0 anyway) or 3) Overvolting DRAM (at 1.4+ my memory requires, the voltage control turns purple.) Ive used ROG boards for a long time and generally safe voltage is white, high but safe is yellow, possibly dangerous over time or on air is purple, and LN2/you could destroy your hardware is red. That color highlighting for voltage controls makes its appearance in this new bios, yet every other ROG board I ever bought had it when I got the board, even on low bios versions. Just goes to prove quality control is slipping and the C8H series was rushed!
Either way, yes, any redditors or people saying the memory overclocking is broken are most likely correct- my board isn't even going through normal dram training at all, even for just 3-4 seconds before failing- it is literally instantly spitting out F9 and not even trying on boot. This bios is broken. Take it from someone who had 40 gold cups on hwbot at one point despite never using LN2
2nd guy: thanks for your advice. No, I went from 3200/1600 directly to 3800/1900 and did not try 3600/1800 first. Because of my aggressive activities/benching this is not an acceptable solution to me, as I have to reset cmos often while testing different memory speeds. I am not willing to gradually increase fclk every time, through 3 steps. It is still good information though, rep+
You didn't get scared of the purple color and back off did you? I left my vdimm at the 1.49v it was before, purple or not. CCD VDDG and IOD VDDG turn yellow at 1v, but thats what I use (they are both at 0.950 right now, to test). Not too much has changed for me. Maybe .4ns lower ram latency and .3ns lower L3 cache latency (corrected after looking at a before picture!)