Active core tuning is a turbo bin setting. It means you’ll see the 1 active core multiplier when the system is using 1 core, which almost never happens. The 4 core bin is the most sane, because that’s how windows usually works shuffling threads arounds.
Your xtu shows you can set a per core multiplier. So set your best core to the multiplier you want (61), and set the rest lower if you want a safe 1 core boost.
If you go into the bios and only enable 1 e core per group it should be the same in theory, problem is how to set the affinity in windows without APO. Maybe turning off hyperthreading will force it to work if a game needs more than 8 threads.
This has nothing to do with apo.
You need multiple clusters of E cores for it to even do anything. Basically APO uses one e-core per bank of 4 to maximize L2 cache for a game thread (l2 is shared by 4 e core clusters). So the only 12th gen cpu that will get meaningful performance is the 12900k. That means it only makes sense on processors with 8 or more e cores.
This is the first generation of UE that drastically overshoots console spec by a wide margin. UE2 and 3 were basically built around OG xbox and 360 hardware, which is why nearly every UE3 game ran at comfortable fps on the 360 at native 720p. UE4 was fairly easy to run on PS4 (though some games had horrible shader compilation stutter).
But suddenly we need 720p and upscaling to get variable fps between 40 and 60 fps on modern consoles. Using Lumen and nanite just because they’re available is probably over doing it. UE always seemed like a console engine first, but now it feels experimental and unoptimized, which isn’t what I think of when I think of games like Arkham Knight and Bioshock.
5700xt has some hardware for primitive shaders which should have launched on vega but was broken, so likely that is being used, which Pascal has none.
Remedy has been upsampling since Alan Wake 1. Quantum Break required it and more recently Control had softened native image quality as well, which is why dlss 1.9 and later 2.0 were better than native there. So this is a case where the developers are actually authoring the materials with upscaling in mind, vs just making an non optimized pos and using dlss to call it a day.
Not sure who we’d blame here. MCM should have yielded much better performance. The cpu group makes the changes neccessary to stay on top (chiplets, 3d cache), while the gpu division tries a few things, doesn’t get optimal results, then continues on as if that’s all there was to be accomplished.
Nothing at radeon ever gets perfected or finished.
60fps is pretty pedestrian these days. Once you start gaming with higher framerates in mind, the newer cpus make a lot more sense.