I know the title is weird, but that's exactly what this is.
It seems like different Ryzen 1700 "versions" behave inconsitently on the C6H and don't show or show bugs that happen on one CPU but not the other.
Let me explain.
I've had a week 21 1721PGT Ryzen 1700 (manufactured in Malaysia), which ran fine on 3.8 GHz. It had the multiplier bug, though, which is known on the C6H, where manual voltage override doesn't work and results in a low multiplier in Windows. No biggie, just use offset voltage + pstate and off you go, right? Right.
Sadly that CPU suffered from the
pre-week-25 Linux compilation bug, so I got in contact with AMD and got a replacement within a week.
Now I'm running a week 30 1730SUS (manufactured in China). This one has absolutely crazy binning and runs at 4.1 GHz (prime/avx stable), 1.3375 V. But it didn't do that from the get-go. When I applied BIOS settings similar to those of the older CPU, with the OC in offset voltage mode, I experienced the stuck low multiplier issue I previously had with manual voltage override. In stark contrast to the old CPU, it works perfectly fine in manual voltage override.
All tested on BIOS 1501 and 1602.
To be honest, the current state of this motherboard and the AMD architecture is some of the worst I've ever seen, even over half a year after the initial launch. I love this PC, don't get me wrong, but the inconsistencies are horrifying. I hope most of the issues will get resolved with the newer AGESA/EC updates.
tl;dr: The old 1700 needed offset voltage + pstate, the new 1700 overclocks like crazy but needs manual voltage override instead of offset + pstate (those don't work at all).