There’s a discussion happening over here.
I think what we’ll see with Zen 4 is that there is such a thing as a cooler that’s too good and allows power consumption to keep ramping up well beyond the power efficiency sweet spot. This is why some reviewers are seeing e.g. the 7950X run 40% faster than the 5950X but consume twice as much power in a given benchmark. Pay particularly close attention to the cooler each reviewer is using. GN is using a Liquid Freezer II at full fans and clocking some of the worst efficiency numbers.
I think limiting PPT (and potentially TDC/EDC as well) is going to be one approach that works in some cases, like if there’s a known heat dissipation capacity in a case or room, or a specific section of the efficiency curve that you want to stay within. Heck, I do this today with some systems that I know to be thermally-constrained due to poor airflow or other factors: ECO mode!
The other option will be to right-size your cooler. I’m eager to see reviews of e.g. the 7900X with a decent tower cooler like the Noctua NH-U12A or Scythe Fuma 3. I bet the efficiency numbers will be through the roof (in a good way) without lowering thermal throttle or power limits. And performance will be close to what we’re seeing in these initial reviews.
Here’s the issue, though: if any meaningful load pushes Tctl/Tdie to ~95ºC, the CPU fan (and any other fans tied to the CPU temperature) will constantly ramp up to full speed. Sure, you can lower the maximum duty cycle to X%, but that just means the fans will constantly ramp up to X%. What are our options for sensible fan curves with this new paradigm? Can we key fan curves to power consumption metrics or something else?