Idk about all that for those games…
Maybe 720-810p at best on PS4, mostly 30 and 1080-1200p or so on PS4 Pro.
And really paired back.
The issue would be the CPUs I feel of those consoles.
I think it’s possible that they are aiming for the best perf per watt in portable mode and it scales fine for docked mode. I was trying to do the math the other day for it and it makes a lot of assumptions, I don’t think it actually works like this because these have curves not a linear relationship per se. Or a curve that can’t be ignored.
Ok, so the MX570 is an ampere based GPU on 8nm that has a max TGP of 25W @ 1155MHz, contains 2048 CUDA cores as well. if we apply it to Drake at the same frequency I’d assume that Drake would draw 18.75W, being 75% of the MX570 TGP because it has 75% of the cores.
Even if we halved that for portable mode, you’d get 9.375W just for the GPU, so I thought, “what if we moved from 10nm to the 7nm family?”
From TSMC 10nm to 7nm, it’s supposed to be a 40% power reduction while at the same clock frequency, no?
But here’s the tricky part: This isn’t easy as it’s 1) 8nm which is just an improved 10nm and 2) a different Foundries 7nm node.
Granted, it’s agreed upon that TSMC nodes are noticeably better than their Samsung equivalent, so for the sake of convenience I’ll say that a 8nm SEC is equal to a 10nm TsMC and if there was an 8nm TsMC (10nm+) it would be better than the 8nm SEC.
Anyway, “9.375W” at 8nm for the portable and “18.75” for the docked clocked profile.
If we reduce it by 40% or so in power which would align with the TSMC figure, you’d get “5.625W” in the lower clocked ~577.5MHz and “11.25W” in the docked configuration of 1155MHz.
If we drop it a little more in clock frequency to 460MHz and 921MHz respectively, we can reduce the consumption probably a little more.
Probably ~4.5W at 460MHz in portable mode and ~8.9W at 921MHz in docked mode. Docked mode is a less concern here though as that isn’t going to be so limited unlike portable mode.
In portable mode though…
Ok so, how would this compare to the TX1 though
, the 20nm version. Do you happen to know a way of figuring that out?
And yes yes, there’s a bunch of caveats here yada yada. Just working with what we got!