Results of my own. Upgrading from an i7-11700 to a 7700X, then again to a 7800X3D didn't change my gaming experience one single bit. They're all very different CPUs, and I love them for different reasons, but that's besides the point.
I probably havent been GPU bound (at least without extreme settings) for a few years at least now. (not including VRAM issues).
However several times I have had game related issues because of CPU. So when I upgraded there was some quite significant improvements, mostly on stutters (min frame rates). I also upgraded multiple generations in one go.
The way I would describe the importance of CPU on gaming is it allows more complex games, and also is important for consistency in game experience. With GPU's if they cant handle what you doing on their compute its so easy to fix, turn down the FPS cap, lower the resolution, turn down fog density and so on. They only become a hassle if the issue is VRAM related in which case usually the only option that has an effect is texture quality. However if CPU bound you usually screwed, as its rare for a game setting to affect CPU utilisation enough to alleviate issues.
You are clearly someone who upgrades frequent given the CPU history you posted so perhaps explains why you have suggested people should just swap hardware instead of tuning.
To me the gains that have been identified for efficiency are.
Tuning the windows power schema, some of the defaults are shocking, perhaps the biggest culprit is the setting that forces e-cores to max turbo clocks whenever at least one p-core is unparked, this will peg the CPU at high voltages for low-moderate loads. This doesnt affect older Intel or AMD CPUs because the bad default is only a used setting if the chip in question has multiple performance class in the chip design.
This may also have contributed to people reporting power savings with e-cores disabled.
HTT identified in this thread, although historically HTT has always been known as power inefficient. I still want to try this on a more modern game as imperium galactica is quite old single threaded game. In theory HTT shouldnt really be impacting a single threaded game, but windows was still trying to share that thread over two logical cores on the same physical CPU which almost tripled my package power. I had a similar issue on my old pfSense CPU.
Voltage and PL limits. W1zzard has done his own testing on reducing PL and assessing impact on games, but of course there is likely a window on many chips depending on chip quality for reducing voltage as well.
In the same way some people will always OC their chips, I tend to be obsessed with efficiency tuning. Have had varying degrees of success with it across different hardware.
I have previously posted on TPU in another thread the impact on idle/low load power consumption with the amount of p-cores unparked. I had posted there was no measurable difference between having them all parked and having just the 2 preferred cores unparked. But noticed was an increase with them all unparked, well it turns out with the 50% setting, that unparks them all (but not the extra logical core per physical core so HT off), its very similar to having just the 2 main cores unparked,
so the culprit for my original observation was HT.
If it was pure need, I would probably just about to start thinking about ditching my i7-7700 non-K for something better.
