So fellow Zengardians, Is it good to have a 3700x that boosts to 4.475GHz on two cores, and 6 of the 8 cores going above the rated 4.4 boost?
No BCLK shenanigans here, stock settings other than slight PBO tweaks
Why is your PPT 3W below stock?
My score with the trick is up from 5001 without PBO and 5013 with EDC 83.
My chip boosts about 50-100MHz higher ST and up to 4.2GHz MT, but that also means it pulls 1.4V briefly, dropping to 1.36V for the duration of the test. Without PBO, Vcore is usually 1.3V in the test, up to 1.337V with EDC 83 PBO. Knowing my chip and how badly it's binned, this seems proof as any that the EDC trick probably breaks PBO and maybe some other safeguards as well, though FIT and PROCHOT are clearly working as normal.
The performance in tests is often very erratic and inconsistent, wildly more so than Ryzen is even at stock. Clocks don't seem to change but ST CPU-Z varies 45pts and MT varies up to 150pts in back-to-back runs. Cinebench ST straight up refuses to finish. Snipping Tool opens only half the time. Sometimes, the CPU will visibly throttle and instantly lose 40pts in the middle of the ST test, although it's clearly not for thermal reasons, as can be seen here with these two done back-to-back:
Needless to say, I just went back to stock. My chip is batch 1924 (mid June 2019), and knowing full well that it can't even manually sustain 4.25-4.3GHz at any remotely reasonable Vcore, I don't like what I'm seeing here with the EDC trick, especially with the errors, instability and freak readings. The fact that my board grossly underreports (~25-30%) doesn't help either, maybe the new B550 board I've got on the way will change things, maybe not.
If your 3700X is one of the newly manufactured ones that are capable of outstanding performance like 4.3GHz @ 1.1V, then I'm sure it's a much safer setting for your chip. Though at that point, it might be worth it to just manually overclock 4.4GHz at a sub-1.25V Vcore, since it's a pretty safe voltage.
