Well if there are no changes made to the game whatsoever, then it has to be using Rosetta. Fantastic would be DXVK approximate levels of performance hit. This is far short of that.
Rosetta is a known quantity of approximately 20% of a drop. DXVK can do about a 20% performance drop in certain situations, and perform better than that in others.
This is at about 50% performance drop translating DX12 to Metal, on top of the drop from Rosetta.
50% drop of what? Where are you getting your baseline from? Because, as far as I know, there is no native macOS port of Cyberpunk 2077.
At 1440p, and if we take this [0] at face value, it would be 50% of the performance of a RX 6700 XT paired with a 5950X, both desktop parts, which I think is pretty good.
The m1 gpu is broadly equivalent to a gtx 1650 in a host of benchmarks. This is getting less than half the fps a gtx 1650 does at these settings, and I am being charitable.
I don't know why you are looking at 1440p (2560 x 1440) as the m1 here is running at 1440x900. While I'm there, that 6700 XT posts 50fps, half of which is 25fps, which would indeed be alright. However this is putting out less than 15fps most of the time.
> I don't know why you are looking at 1440p (2560 x 1440) as the m1 here is running at 1440x900.
Ah, you are right, I messed up with the resolution.
> The m1 gpu is broadly equivalent to a gtx 1650 in a host of benchmarks.
Which ones?
> This is getting less than half the fps a gtx 1650 does at these settings, and I am being charitable.
... without having to emulate both CPU architecture and graphics layer.
I mean, this is not really a debate. DXVK is not comparable, Wine does not do the same either. We are talking about translating both CPU instructions and graphic API calls in real time, good enough that a triple A game runs without any modification on a laptop with 16GB of shared memory.
I'm wondering if this is also translating the binary with Rosetta.