Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
So M1 Max with 32 Core GPU matches RTX 3080 Mobile.
- Thread starter iBug2
- Start date
- Sort by reaction score
Right I saw that too, but upping the resolution (to 1440p or higher) shifts the bottleneck back to the GPU which should make the CPU moot.
I don't think removing Rosetta's hit would make the M1 Max on par with a 3080 that's basically 2X faster in gaming.
It's not that they're not metal, it's that they're both x86. That's why Rosetta needs to translate. For the CPU limited scenarios that matters. When it isn't CPU limited, it's still not great but there the issue is how optimized the Metal ports are. Probably not very and definitely not for the M1-style GPU. Like all games and a lot of other intense software, optimized is a big step beyond simply being native. Having said that, this probably reflects the state of most games on the Mac, and probably for awhile. But it doesn't reflect *what can be*.
No, definitely not. Removing Rosetta AND optimising for Apple Silicon likely could come close. After all, the Aztec benchmark is very good. The Mac version of these games has been an afterthought, they're likely not very optimised.
That's not to say the benchmarks are useless, they aren't. They reflect the current performance of some of the most popular available Mac games. But not the potential performance of the hardware.
Also, even if the graphics code is using metal, it doesn't mean it targets TBDR:s like Apples GPUs. Anything that runs on the x86 Macs is fairly sure to target IMRs. (They could conceivably make a seperate code path for AS, but I haven't heard of anyone doing it.)
With the amount of bandwidth M1 Max has it shouldn't matter if it is treated like IMR or TBDR, right?
No, it's definitely not using TBDR. Porting to AS would come much sooner than switching the entire pipeline to TBDR.
You're potentially unnecesarily copying gbuffer textures back to main memory when they're already available in tile memory. It still matters.
Which is strange because these games are well optimised for Metal. It's not clear why the M1 does not perform better at very high resolution.
As for Borderlands 3, this is a awful port. Performance on macOS may be only half that on bootcamp on an intel Mac.
Anyone seen any reviews that discuss the GPU performance in Octane, Redshift, Cycles4D yet?
When we sit back and think about it, gaming performance is in line with what the M1 does and scaling is very good.
The difference is that the M1 has been compared to iGPUs, which all had lower specs on paper.
Now, the M1 Max is compared to GPUs that have much better specs and higher power consumption. That's why we're disappointed. But results are entirely expected.
The difference is that the M1 has been compared to iGPUs, which all had lower specs on paper.
Now, the M1 Max is compared to GPUs that have much better specs and higher power consumption. That's why we're disappointed. But results are entirely expected.
How much does it matter? With 400GB of bandwidth how much performance are you leaving on the table pretending you are on IMR GPU instead of TBDR?
EDIT: I grammar/spelling
The only way to know for sure is to run custom benchmarks and I don't have a M1 Pro MacBook (yet). Apple even provides sample code of modern rendering techniques on the same scene, from WWDC videos, so it wouldn't be too difficult to try.
But I'd expect that there's still a lot to be gained despite M1 Pro/Max's massive bandwidth. You have added latency on every operation since you'd be accessing system memory when writing to most of the intermediate buffers (some would be cached on-chip anyway even on IMR GPUs, some wouldn't), plus if you need to access neighbouring pixel data (think gaussian blurs) you'd already have it on cache due to improved locality since neighbouring pixels are likely to be in the tile.
And on IMR you'd be shading A LOT more triangles than needed since the visibility testing is done later in IMR pipelines (it's culled earlier in TBDR GPUs), wasting core resources like ALUs shading geometries that would never appear onscreen. If your scene is limited by ALU utilization (I don't know how common that is in modern 3D rendering, I've mainly made compute work on the GPU until very recently) that would still have a big impact in performance.
According to their FAQ it’s metal
“Q: Does Civ VI Mac support Metal?
A: Civilization VI Mac does support Metal.”
Supporting Metal and being compiled for
arm64 are two completely different things. You could make Metal games long before Apple Silicon was introduced.
Sigh, again: Metal is the GPU API, x86 vs arm64 is the CPU instruction set. The Mac graphics engine for Civ VI is written in Metal (GPU), but it is also running translated under Rosetta as it is compiled for the x86 architecture (CPU) and does not have an arm version (CPU). It is also like that's rendering pipeline (GPU) is not M1-optimized (GPU).