@Metroid
You continue to show that you don't know what you're talking about, even with the 50+ reviews that you have read.
First of all, PB is not some fixed table ... If this volt and this temp than that frequency.
No, far from it.
The clock depends on so many things it's just mind-boggling:
- Temperature at any given moment
- Incoming core voltage (which is not fixed btw, what is set in bios is just a "hint". It still varies around that value quite a lot !)
- Any power limits (watts, amps) that are being reached
- What parts of the instruction set is being run (AVX loads seem to clock a lot lower than Integer for example)
- Freekin' MEMORY efficiency ... I've changed memory to lower/higher and noticed the boost behavior has changed slightly (my guess waiting for the memory makes the PB logic attempt to save CPU power as well by lowering clocks)
- The actual instructions that are being executed !!
Btw, I've watched many of those reviews as well and saw a LOT of inconsistency.
After buying mine, started to understand why:
TESTING CONDITIONS AFFECT PERFORMANCE (sometimes significantly !)
This is completely different than typical Intel CPU, which basically runs at the exact same performance no matter what. You set 4.9 Ghz, it sticks at 4.9 Ghz and done. It only gets slower when throttling for reaching TJunction.
To answer the humble request:
No, and Yes, and "Depends"
Running
single core loads varied from ~4.2 all the way to 4.4 (with everything "default"), and if I added +100 to PBO limits, even 4.45 sustained for short periods of time.
- SuperPI hit 4.4 often
- Cinebench didn't (tops at 4.3, 4.325)
- Old Cinebench R10 (Motorcycle) hit 4.4 for very short times
- Tried some old 3DMark 2003, 2005 and other tests (which are limited to single threaded), seen 4.25, 4.35... 4.4 even, depending on which part of the test it was in. (With repeatable results)
Whatever AMD did, it's very complex and can't be measured by simply "Hitting RUN" and be done with it. That is the old way.
The new way is "per application maximum performance extraction"
How to make these go as fast as they can:
- Get the CPU as cool as possibly can and afford, because it will clock higher at lower temps (not by much but observable in benchmarks)
- Use under-volting in very tiny steps to find the sweet spot for your particular CPU/Motherboard/Case/Ambient, because sometimes it can get faster with lower voltage (up to a point). If the voltage gets a teeny-bitsy too low, it starts to cut off from the higher boost levels, probably to maintain stability ... somehow the CPU knows how much can it go before falling off the edge
- Use fast RAM with tight timings 3200-CL14 or 3600-CL16, because this CPU LOVES getting fed with data quickly !
- Use BPO+AutoOC (+100 to +200) if having high-end cooling and a good mobo with very stable voltages, because it DOES clock higher, it's not a fantasy, but it's not like with Intel... simply goes higher and crashes if unstable). No, Zen 2 doesn't seem to crash. Instead, it simply runs a tiny bit slower to maintain stability.
---
@Bones - Saw your post
You set everything manual, that's why, and cancelling
PB.
But by doing that, you made the CPU very inefficient, running higher voltage than needed for various tasks, and consuming a lot more power.