M1 / Apple Silicon compatibility

Since this thread has been dusted off again…

@Dave_ZPCMR : I can’t find it now but you linked to a table you created eons ago that showed the differences in what each device/CPU got you vis-a-vis graphics. X-axis had stuff like “ATV - iPad - M1/M2 - various Nvidia/AMD GPUs etc” and Y-axis was features like “Profile, rider shadows, tree shadows etc”

Can you link to that again?


There was a comment elsewhere recently (Reddit?) where a user commented that people overestimate the power of SoC’s and underestimate how poorly optimized the Zwift codebase is, which leads to discussions like the one we’re in right now.

I suspect Zwift is working hard, quietly, behind the scenes, to extricate themselves from technical debt to better optimize the game but it’s like changing an aircraft engine while in flight. You want to keep everyone riding and not have the plane fall out of the sky.

For every complaint about graphics, there are probably a thousand happy users running Zwift on a potato they pulled out of sock drawer.

This is horseplop. Benchmarks don’t count. You simply can’t compare different architectures in this way. Metal is a proprietary Apple API, it’s little wonder Apple silicon does it really well versus third party GPUs. In benchmarks.

3 Likes

That was me. It’s not helped by exaggeration like the one above.

1 Like

@Dave_ZPCMR I think you are beyond doubt the guy with the most knowledge on this subject.

Could you explain to me and everyone else for that matter, how let’s say A14 chips out perform PS3 when looking into game’s. I mean from own experience I know that COD mobile on iPad Air 4 performs better than Metal Gear Solid 5 on a PS3.
And looking at what COD mobile performs on a A8X chip vs Zwift on A10 and better is just cringing.

I know that benchmarks aren’t real world, but they must stand for something, right?

A couple of things I’d say. Firstly I’m not knowledgeable on games development or coding, not at all. I’m only ever going off my experience of what I’ve seen, and what’s actually out there in the real world (i.e. the games studios not releasing PS3/4-tier games for Apple products). Generally you can’t compare benchmarks across vastly different platforms for obvious reasons, but here’s one: Power Board | Compare performances of smartphones, tablets and PCs

Have a look at a GTX 970 versus A15 on that.

Secondly by no means am I excusing Zwift’s rudimentary graphics profile system, nor am I saying these Apple devices aren’t capable of better results in Zwift. I simply don’t know what they’re truly capable of when considering how poorly optimised Zwift is. It’s poorly optimised using the hardware it’s created and developed on (OpenGL, x86, Windows), let alone when ported onto other platforms.

COD Mobile is an interesting one. Personally having looked at that video I’m not sure I’d agree that it looks better than a good PS3 game, and I don’t know what the respective frame rates and resolutions are. But that aside, it’s an example of a publisher (one of the largest in the world) seeing the incredible revenue opportunity that comes with putting an extremely popular franchise within a free to play game on millions of devices. It’s in their interests to make it look great, to get more people playing and generate more revenue. This couldn’t really be further from Zwift if it tried - a niche game where the overwhelming majority don’t give a toss whether it’s 1080p30 Basic or 2160p60 Ultra, and everyone pays the same irrespective of how much they use it.

The fact that Apple TV is already so popular for running Zwift is all the proof needed, really. There is little or no incentive for Zwift to improve the graphics on Apple TV.

1 Like
1 Like

I’m not saying it looks better, but it performs better from my experience.
The PS3 MGS5 hangs at about 30fps at 1080p as the iPad Air 4 (A14) COD mobile reaches 60 fps on high graphics retina resolution.

Indeed. This horse has been beaten to death. A phrase I’m fond of in light of “benchmarks”:

“The difference between reality and theory is: in theory, there’s no difference between the two.”

Cheers mate!

Benchmarks are useful as a way of directly comparing products within the same controlled environment, given known performance of a baseline. Zwift typically responds to raw processing power, so it’s worthwhile knowing how a GPU or CPU compares to another when running the same benchmark.

However it only really works if else everything is the same; the architecture, the API, the operating system, the other hardware used and so on. So even comparing different generations of GPU/CPU on the same system otherwise can be fallible. There isn’t much linear scaling when it comes to Zwift performance, which backs this up.

1 Like

Coincidentally I was reading the following this morning:

which also includes discussion of the Apple TV 4K 2022 in comparison to the Xbox One, PS4, and PS5.

1 Like

We are starting to get to a point where Xbox One and PS4 games should, in theory, be playable on Apple TV without too many compromises.

Righto.

Hey look: Apple TV's path to becoming a real game console - FlatpanelsHD

Apple TV is now powerful enough to run many games from previous consoles generation

Show me, Rasmus. Show me.

well, not necessarily…
The fact that Apple TV is so popular for running Zwift could be — or should be! — a valid reason to improve the software on this platform!

3 Likes

That’s one way of looking at it, but far the most popular response to any note that Apple TV runs the lowest graphics profile at a low resolution and frame rate (compared to other options) is that users find it good enough, or simply don’t care at all - both of which are totally fine viewpoints to hold. Personally I want the best of what I’m paying to stare at, but that doesn’t make others wrong. They’re typically coming from an old/weak/slow/unreliable laptop, so by default it feels like a big upgrade in many respects.

2 Likes

Obviously, the majority in this discussion thinks differently about your “it’s good enough” statement.

Indeed! I too am part of that camp :sweat_smile:

And while I contributed to going off-topic w/a response to an Apple TV post, this thread is about Apple Silicon (M1/M2) of which the Apple TV does not have! So, let’s continue moaning about the lack of the lack of M1/M2 native support (no Rosetta) :grin:

1 Like

That’s partially incorrect.
Apple TV uses an A15 Bionic SoC, which is a 64-bit ARM architecture just like Apple M1/M2. Since the assembly code for the two CPUs is basically the same, the two topics are closely related.
What’s more, iPhones and (more importantly) iPads belong to the aforementioned 64-bit ARM family too.
These are all Apple platforms, so when you develop for a device you don’t have to reinvent the wheel if you want to port the source code to another platform: you usually just need minor fixes.

PS: I’m referring to the latest models in all cases.

They might both be ARM-based, but there are big differences in the architectures of A15 Bionic and M1/M2. You’re hugely over-simplifying.

2 Likes

No: “tvOS. Leverage many of the same frameworks, technologies, and concepts as iOS.” Source: Apple
Obviously, nobody forces you to design/write software by leveraging the same frameworks (to use Apple’s words), but it is advisable if you want to produce portable software.

This thread isn’t about portable code though.

This thread is about extracting high(er) performance from software. And that involves understanding the architecture to make the most of it.