Taking games designed for desktop GPUs and running them on mobile GPUs with tile-based-deferred-rendering hardware will be a disaster. Mobile GPU designs will choke on modern games as they're designed around hardware features that mobile GPUs either don't have, or that run very slowly.

Peak theoretical throughput for the GPUs you find in ARM SoCs is quite good compared to the power draw, but you will not get peak throughput for workloads designed for Nvidia and AMD GPUs.

Isn't the GPU on Apple Silicon machines a tile-based "mobile" GPU design? Many of the hardware features that traditional GPU's have and mobile GPU's lack can be easily "faked" with GPU-side general compute.

But even the most powerful apple silicon GPU is terrible compared to an average Nvidia chip

While I agree with the general point, this statement is factually incorrect - apple's most powerful laptop GPU punches right about the same as the laptop SKU of the RTX 4070, and the desktop Ultra variant punches up with a 5070ti. I'd say on both fronts that is well above the average.

There is no world where Apple silicone is competing with a 5070ti on modern workloads. Not the hardware and certainly not the software where Nvidia DLSS is in it's own air with AMD just barely having gotten AI upscaling out and started approximating ray reconstruction.

Certainly, nobody would buy an Apple hoping to run triple-A PC games.

But among people running LLMs outside of the data centre, Apple's unified memory together with a good-enough GPU has attracted quite a bit of attention. If you've got the cash, you can get a Mac Studio with 512GB of unified memory. So there's one workload where apple silicon gives nvidia a run for their money.

Only in the size of model it can run, not speed of token generation.

Apple's MetalFX upscaler is pretty similar to DLSS (and I think well ahead of AMD's efforts on this front).

Ray tracing outside of Nvidia is a disaster all round, so yeah, nobody is competing on that front.

Ray tracing support on the newest AMD chips is getting good enough. They are still behind Nvidia but definitely not a disaster anymore

That simply isn't true. I have an RTX 4070 gaming PC and an M4 MacBook Pro w/ 36GB shared memory. When models fit in VRAM, the RTX 4070 still runs much faster. Maybe the next generation M5 chips are faster but they can't be 2-4x faster.

GP said laptop 4070. The laptop variants are typically much slower than the desktop variants of the same name.

It's not just power budget, the desktop part has more of everything, and in this case the 4070 mobile vs desktop turns out to be a 30-40% difference[1] in games.

Now I don't have a mac so if you meant "2-5x" when you said "much faster" well thdn yea, that 40% difference isn't enough to overcome that.

[1]: https://nanoreview.net/en/gpu-compare/geforce-rtx-4070-mobil...

Are there real world game benchmarks for this or are these synthetic tests?

Only a few, because it's not easy to find contemporary AAA games with native macOS ports. Notebookcheck has some comparisons for Assassins Creed: Shadows and Cyberpunk 2077[1]

[1]: https://www.notebookcheck.net/Cyberpunk-AC-Shadows-on-Apple-...

And it will consume almost as much power as the Nvidia GPU to do so.

a 4.5k$ M4 Max barely competes with an entry-level laptop with a 4060 which will be around ~1K in FPS in cyberpunk given the same settings. For AI it's even worse - on NVidia hardware you're getting double-digit speeds for FPS for real-time inference of e.g. stable diffusion, whereas on the M2 Max I have you get at best 0.5 FPS

Snapdragon doesn't do tile based deferred rendering the way Apple does (or did). Snapdragon does (or did) a form of tile-based rendering, but it is a completely different design, with completely different performance tradeoffs.

What about the Switch 2 (nVidia's Tegra) line? The one in the Swich 2 is using Ampere architecture.

That should be feasible, no?

Linux support is in a terrible state for Nvidia chips. Not going to happen.

You don't have to use tile-based rendering on these chips anymore. They can directly draw to the entire screen.

You can, but the immediate mode path is slower and uses significantly more power. Mobile GPUs are not good at modern desktop game workflows where significant portions of the frame are compute shaders. They're generally very memory bandwidth starved, and general compute sidesteps most of the optimizations the hardware has made to work around this.

I agree they won’t do a Steam Deck 2 that’s ARM. Maybe in the future?

BUT, what about a “Steam Deck Mini”? Something at/above the current Steam Deck, maybe a little closer to Switch 2, but smaller/thinner/maybe a little cheaper?

Yeah you’re not going to run Cyberpunk 2087: Johnny’s Rent Is Due. But older games, less demanding indie games, and many emulators would still work great. Plus remote play of your big desktop if you have one.

I’m not saying they will, but I could see it as a possibility.

Apple not sharing their chips extends to Apple keeping their grip on the higher density nodes.

I wonder if it's still the case, but for a while Apple was buying the totality of TSMC's capacity for 3nm nodes, leaving the rest of the world with only 4nm+ chips to grab.

You don’t need to wonder. Top of the lines Snapdragon, Dimensity and Exinos SoC all use 3nm.

Amusingly, it’s the second time in two days I have this discussion here and I have noticed that a lot of people, who I think are American and using Apple phones by default, are completely unaware of what the mobile SoC landscape looks like nowadays. Apple lead doesn’t exist anymore as of this generation.

Apple still leads in raw performance. Their M5 is far ahead of basically everything in single-core performance. AFAIK it's because their architecture prioritizes IPC over frequency, and they can spend the entire silicon budget on a very large monolithic chip.

Bitmain is quite often first on new nodes, ahead of Apple, even.

Current gen ARM is pretty strong - for an example, the switch 2 runs on ARM and it's decently more powerful than the current steam deck.

Yes, the Nvidia GPU in the Switch 2 is more powerful. But not the ARM CPU.

The existence of Nvidia DLSS (upscaling and frame generation) alone is a huge advantage over the Steam Deck, too. The Deck can't use DLSS because it's Nvidia only, AMD FSR isn't as good, and the latest FSR isn't even supported (officially) on the SoC.

The bit about FEX is interesting. Taking x86 code and running it on ARM. The most important thing for Valve to do is pick what instruction set to use, one you can run natively or native hardware, or efficiently and reliably through translation on alien hardware. ARM might be a great choice, as hardware exists at scale on mobile devices, and emulated on other devices even if the CPU happens to be Intel or AMD. Valve is then in control, rather than Intel or Apple or Microsoft.

Why couldn’t they have an AMD GPU and an ARM CPU?

Does something like that exist on the AMD side?

You're basically describing nVidia's Tegra line. The latest is in the Switch 2 I believe.

Samsungs newer Exynos are somewhat like this, their GPUs are based on AMDs RNA2.

What would that do? Power consumption would mainly come from the GPU because these devices are for gaming.

> There are no ARM chips with enough power.

Disagree.

Both Qualcomm and Mediatek have mobile SoC which are more performant than the M2 and the X2 Elite is in the ballpark of Apple top SoC.

Considering how I currently use my Steam Deck, there is nothing my current phone couldn’t do. Sure, you won’t get PS5 performance but I’m personally completely happy with Switch 2 level performance.

> Both Qualcomm and Mediatek have mobile SoC which are more performant than the M2 and the X2 Elite is in the ballpark of Apple top SoC.

At what power consumption? And is that both CPU and GPU, or just GPU?

Do you mean there's no ARM chips that they can buy? Surely the ARM chips in Apple's devices are powerful enough aren't they?

You can’t buy an M3 chip on its own

Why can’t someone else make one?

That’s very much the trick. Apple is actually exceptionally good at making CPUs. Look at these single-thread benchmarks: https://www.cpubenchmark.net/single-thread/ Similarly, if you look for the M4 in this list and then look for other ARM chips, you’ll have to look quite a ways down the list: https://www.cpubenchmark.net/multithread/mobile

It’s baffling to me that no one else (Qualcomm) has not been able to come close.

My guess is that CPU design is existential for Apple, and no one else cares enough to be dedicated enough to do what Apple has done.

This is exactly what happens when you invest billions and hire the best industry specialists for decades. M-series processors did not magically appear out of nowhere. Apple perfected them for years in iPhones, but people didn't have the ability to compare since Apple doesn't share their processors with anyone.

Because it’s a proprietary design? You’d have to reverse-engineer the whole chip, which is really hard to do on that process node

Giving them some credit, I think they're asking "why isn't there a close competitor" and that takes a much more involved answer.