If it failed at 1900 surely it would insta-fail at 1933? Or am I missing something? probably!
Not entirely. Dead straps do exist, we've seen this even on older AM3 platforms with the 2666 strap not POSTing and the 2133 strap getting out-performed by 1866. You may find that you can achieve an FCLK of 1800, 1866 and 2000, but every strap in between refuses to POST no matter what you do. These would be considered dead straps.
Is That the AMD CPU or the motherboard issue?
AFAIK, all b550 Asrock boards are still on agesa 1100D and there's a 1900 IF hardcap somewhere in the bootloader.
I've found it more to be a board/firmware issue. The memory controller might not like specific timing configurations at specific straps, but the inability to POST at specific straps is almost always related to the board and its specific firmware implementation.
Which BIOS version is atm on a board? I just saw one user reported ss with PRD ~500% in CB R23 multicore test, and clocks were stretched as hell, though it was on Asus 570 tuf.
Board came with 1.30 out of the box, first thing I did was flash 2.00 for AGESA version 1.2.0.3C. I have tried rolling back to a couple of older versions, but the memory support ends up being horrible and the issue still remains.
So did you experienced something like high Power Deviation reported by HWInfo or low multicore score in CB R20/23 accompanied by clock stretching, i.e. effective clock much lower than peak ones in a first frequency group?
The update to this is the board functions fine. No issues to report after that easy RMA. I did learn my 5900X has weak cores with very little headroom so I went for the push the memory as far as possible route. Tested to 900% in HCI. I'm still messing with the curve optimizer at stock power limits since my cores don't gain much with more power. Upside is that it saves me the heat at noise.
I see your cpu usage is at 100%, what kind of load were you using during that period?
Honestly can't remember. Used a mix of Prime95, LinpackXtreme, y-cruncher, and Cinebench R23 on loop.
