Well, settings are settings and usually they should "work" in an absolute sense, no matter the BIOS version. Esp. memory timings. And that's not the case here, I think this might have started from versions 17xy onwards, having read lots of threads about those newer versions.
But I think I've now stabilized it, actually 1901 seems to be _better_ than 1302 in regards to memory oc. The issue for me was not vcore/CPU-related as I initially thought (though there's a slight difference in LLC levels, I'm pretty sure), but memory. Esp. RTL/IO needs to be different and in general, slightly higher freqs (100-200MHz) might be reachable on 1901 for me. But that's beside the point, first prio was to stabilize my previous 1302 timings:
1. Set RTT_PARK to 240 for both channels. (as suggested in some Maximus BIOS threads --> also seems to fix long POST times)
2. Set XMP1 BIOS profile (before I always used Manuel. But I set all timings manually anyway, so I'm not sure this does anything)
3. RTL/IO all Auto in order to POST my previous timings, then proceed as usual to determine the best values. (resulted in tighter settings than what I had before! Don't leave those on Auto if close to the edge!)
4. All other primary/secondary/tertiary timings remained the same, no single regression, bandwith/latency about the same as on 1302 and appears to be stable.
For reference, this is with 4 sticks of good B-Die 3200 CL14, running at 3913 16-17-17-35 with very tight subtimings overall. Currently testing 4013, which didn't POST on 1302, now it does... 🙂
