Matisse binning from Silicon Lottery
AMD Ryzen 7 3700X @ 4.05GHz 1.237V 100%
AMD Ryzen 7 3700X @ 4.10GHz 1.250V top 74%
AMD Ryzen 7 3700X @ 4.15GHz 1.262V top 21%
AMD Ryzen 7 3800X @ 4.20GHz 1.275V 100%
AMD Ryzen 7 3800X @ 4.25GHz 1.287V top 53%
AMD Ryzen 7 3800X @ 4.30GHz 1.300V top 20%
AMD Ryzen 9 3900X @ 4.00GHz 1.200V 100%
AMD Ryzen 9 3900X @ 4.05GHz 1.212V top 87%
AMD Ryzen 9 3900X @ 4.10GHz 1.225V top 68%
AMD Ryzen 9 3900X @ 4.15GHz 1.237V top 35%
AMD Ryzen 9 3900X @ 4.20GHz 1.250V top 6%
Any AIO CLC liquid cooler or custom water loop with at least 240mm of radiator space.
Gelid Solutions GC-Extreme
Thermal Grizzly Hydronaut
Thermal Grizzly Kryonaut
ASUS X570 Crosshair VIII Formula
ASUS X570 Crosshair VIII Hero
ASUS X570 Prime X570-Pro
ASUS X570 Pro WS X570-ACE
ASUS X570 STRIX X570-E
ASUS X570 STRIX X570-F
Gigabyte X570 Aorus Xtreme
Gigabyte X570 Aorus Master
Gigabyte X570 Aorus Pro
Gigabyte X570 Aorus Ultra
Dual Channel DDR4-3200 (Up to 4 dimms single or dual rank)
Dual Channel DDR4-3333 (Up to 4 dimms single rank or 2 dimms dual rank)
Dual Channel DDR4-3466 (Up to 2 dimms single rank)
Dual Channel DDR4-3600 (Up to 2 dimms single rank)
There is absolutely a difference in silicon quality between the 3700X and 3800X, to get that out of the way. On average, the 3800X will hit about 100MHz or more higher all core. I know a lot of people have been wanting some kind of confirmation on this. We won't have statistics up right at launch as our sample size is still a bit small (a bit of trouble getting these CPUs right now), but they should be posted within a couple weeks.
Overclocking on these CPUs in general is definitely not warranted, as the stock boost algorithm is great- even better than 1st and 2nd gen Ryzen. Full AVX2 support on these CPUs has put us in a pinch in regards to validating overclocks on these CPUs, as they run much warmer under AVX2 stress tests than previous generations.
At stock settings these CPUs will run an all core frequency 100-200MHz lower when running Prime95 small ffts compared to lighter stress tests, which is something we can't replicate as there is no AVX offset available on the platform. This in turn can cause some of the lower binned CPUs to have a lower all core frequency than the stock algorithm will provide in other workloads, not to mention workloads not using all available cores.