F1 2017 - performance comparison

Look in the "advanced" pane of the Hitman Options window (that is, the one that opens before the game actually launches).

 

(Already answered, as I see…)

 

I didn't have the absolutely latest drivers, but far more recent ones than the Bootcamp drivers. The drivers I used definitely have specific optimisations for Hitman.

 

FYI, I'm getting 63.8 fps average with the same settings as used by imacken. This is on my i5 iMac with radeon 580. The performance difference is small, but at this resolution, the radeon 580 is under-utilised. Usage stands mostly between 60% and 70%. At 4K, GPU usage is >95%.

 

Oh, right, I didn't mention that: I have the 2016 15" rMBP with 2.9 GHz i7, 16 GB RAM, and the Radeon Pro 460 (4 GB VRAM).

It's actually quite consistent with most other performance comparisons I've made myself or seen elsewhere: no matter if Unreal Tournament, War Thunder, World of Warcraft, or Total Warhammer, the Windows version always shows usually a between 10 and 20% better framerate.

The F1 2016 benchmarks jeanlain quoted are still the only ones I've seen that reported a better performance for the Mac version.

 

I'm getting 62 fps with your settings.
FWIW, I just re-tested Hitman with the latest crimson drivers and the difference between Metal and DX11 has not noticeably changed.

Are there F1 2016 Metal-DX11 comparison tests posted online beside the ones we're discussing?
Also, DX11 War Thunder is way faster than the Metal version on my Mac. Like 50% faster.

 

Ah, I bet Codies/Feral are syncing to the display V-Blank to get better frame-pacing - by going through CoreAnimation there's no option to disable V-Sync in Metal but it is a triple-buffered display pipeline so the API won't block you until you hit 120Hz - you've got to do that yourself.

For racing games like F1 though the actual pacing of the frames at regular (33.3ms or 16.6ms) matters or it'll feel terribly hitchy and hard to control as the frame pacing wanders around. The game is going to be a much better experience to play with this feature forced on.

 

Rasterisation (pixel generation) is entirely done by the GPU. So as long as GPU usage is moderate (i.e., GPU is waiting on the CPU or V-Sync in on), changing resolution should not change performance much. It's a common pattern (and there's no reason to distrust the F1 2016 results that show this pattern). This is why GPU hardware performance should be compared at very high resolutions. At low resolutions, you compare different CPUs or APIs (driver efficiency, since drivers "live" on the CPU).

Note that in 4K, Metal and DX11 perform equally in Hitman. Difference is insignificant and ranking may depend on the particular run. Hence, the difference in performance between OSes pertains to differences in efficiency of the CPU code (engine and API/driver code), not the GPU code since performance is limited by the GPU throughput at high resolution and not by the CPU. Efficiency of the CPU code has always been the main cause of differences in perf between macOS and Windows, so it's common to see fps differences between OSes decrease as resolution increase. It's the case for the F1 2016 mac4ever test, but this time Metal has the lead in CPU efficiency.

Windows performance in Hitman is uncapped. DX12 is faster, but not reliable through the entire test (severe drops). I may retry.
Boot camp comes with drivers v. 16.2 or something.