Updated to Windows Insider Fast Build 18995.1 and it appears to have improved threading performance. For single core, the load is staying on Core 0 now rather than bouncing between core 0 and core 2. Single core score now matches my manual score
To qualify, I did that testing only to prove that a single core could run at 4.6 GHz using the FIT voltage. You should not set your processor to 1.465V for everyday use. Those voltages are what AMD uses your lightly threaded workloads. Applying that voltage manually will cause that voltage to be applied even during all core workloads. You should enable precision boost overdrive and auto overclocking to remove any TDC, EDC, PPT constraints and then run a multicore workload. If you cooling is adequate, your processor will stop boosting at a specifici voltage even though no other limit has been met. This will be the voltage limit for your specific processor based off AMD's FIT limit. Running over that voltage will cause degradation in the lifespan of the silica. How much, is unclear. If you want to manually overclock, this is the voltage you should use.
As for the discrepancy observed when your "fastest" core isn't used in a single core workload. There is actually a difference between how fast cores are determined between Ryzen Master and Windows.
I can not get my 3900X to Boost to 4,6 GHz. BIOS & Ryzen Master on Default.
Cooler: H150i Pro RGB
MB: X570 Aorus Elite
BIOS: F12f (Agesa 220.127.116.11 B)
When enabling just 2 cores, they boost to around 4,55 GHz.
A little late, but unfortunantly These processors don't properly clock if you have Cores within a CCX with different frequency values. If you have 1 core at 4.6 and the rest at say 4.55 then instead of getting 4.6 and 4.55 you will end up with something strange, usually clocking below 3800mhz in loads. You're going to want to make the whole CCX with your fastest core, and more importantly the CCX with the core(s) that windows choses for its single core loads, at 4.6ghz.