Native Support for M1 processor and Metal in Mac

[Zeyphersan]

I know this doesn’t answer your question but I’m still curious: how does the game perform using Rosetta 2? Does it even run?

[Giginen]

I, too, would like to know if Square plans to offer an Apple Silicon version of the FFXIV game client. They've always had an excuse why they won't offer a true MacOS version, so they offered a poor emulation wrapper of their Windows client to the Mac users. For someone like me, who's played FFXIV for 1380 days, as of today, it would turn me off of Square, if they just kick the Mac users to the curb, because they don't want to put the effort in to making a proper MacOS port, that will support the great new Apple Silicon processors.

[Packetdancer]

I know this doesn’t answer your question but I’m still curious: how does the game perform using Rosetta 2? Does it even run?

It runs well. Honestly, I'd actually say surprisingly well.

I picked up the Mac license and did some tests on an M1 MacBook Pro, and while I didn't get 60fps or anything, performance stayed solidly between 28 and 45 fps. (The 28 end of that range was in Limsa. Unsurprisingly.) Fan didn't kick on and the laptop didn't get scalding hot, either, though I grant I didn't go try alliance raids or anything; I wandered around a bit to poke at stuff and make sure things weren't horribly broken, fixed up my UI for the MacBook Pro's screen resolution, and then crafted raid pots while I watched TV.

It actually ran better under Rosetta 2 on the M1 Mac unplugged from power than on the 2018 Intel MacBook Pro under Boot Campand connected to a power supply. Which was not the result I expected, let me assure you; ARM is a far more efficient architecture than the (slightly creaky) x86/x64 is, and Rosetta 2 is actually an astonishing bit of engineering, but I didn't expect the result to be even remotely as good as it is.

You could probably eke better performance out of it on the MacOS side with an Intel MacBook Pro with a Radeon eGPU connected via Thunderbolt 3, but honestly I'd bet only slightly better.

There were some initial issues with it, though they were on the Rosetta 2 side of things. Namely, Rosetta 2 is an AOT (Ahead-of-Time) compiler; it analyzes an Intel program the first time you run it, then actually generates an ARM executable. It's really, really good at what it does -- see my previous 'brilliant bit of engineering' remark -- but it was not super aware of how to handle Crossover bottles (like what the FFXIV client runs in). So the first time you downloaded a patch, the actual executable that ran FFXIV (i.e., the Crossover WINE-variant wrapped around FFXIV) did not change, but the data it loaded (i.e., FFXIV) did... which meant Rosetta 2 didn't see it as a changed executable to re-generate the ARM version.

The practical upshot of which was that it would work swimmingly when you first installed it, and then the first time you installed a patch it would just silently fail to run ever again. (Because the ARM binary was out of sync with the new version.) There were fairly simple ways to fix that—basically trickery to make Rosetta see the patched version as a new version—but they were not the sort of thing customer service teams would be likely to provide as advice. Or that the average user would probably feel entirely comfortable doing.

Sometime in the past few months, though, Apple seemingly fixed that issue in Rosetta 2; Crossover (and other WINE variants) seem to work fine now even when something in the bottle gets updated, which means FFXIV patching itself doesn't break the world any longer.

[Arrhin]

It runs well. Honestly, I'd actually say surprisingly well.

I picked up the Mac license and did some tests on an M1 MacBook Pro, and while I didn't get 60fps or anything, performance stayed solidly between 28 and 45 fps. (The 28 end of that range was in Limsa. Unsurprisingly.) Fan didn't kick on and the laptop didn't get scalding hot, either, though I grant I didn't go try alliance raids or anything; I wandered around a bit to poke at stuff and make sure things weren't horribly broken, fixed up my UI for the MacBook Pro's screen resolution, and then crafted raid pots while I watched TV.

It actually ran better under Rosetta 2 on the M1 Mac unplugged from power than on the 2018 Intel MacBook Pro under Boot Campand connected to a power supply. Which was not the result I expected, let me assure you; ARM is a far more efficient architecture than the (slightly creaky) x86/x64 is, and Rosetta 2 is actually an astonishing bit of engineering, but I didn't expect the result to be even remotely as good as it is.

You could probably eke better performance out of it on the MacOS side with an Intel MacBook Pro with a Radeon eGPU connected via Thunderbolt 3, but honestly I'd bet only slightly better.

There were some initial issues with it, though they were on the Rosetta 2 side of things. Namely, Rosetta 2 is an AOT (Ahead-of-Time) compiler; it analyzes an Intel program the first time you run it, then actually generates an ARM executable. It's really, really good at what it does -- see my previous 'brilliant bit of engineering' remark -- but it was not super aware of how to handle Crossover bottles (like what the FFXIV client runs in). So the first time you downloaded a patch, the actual executable that ran FFXIV (i.e., the Crossover WINE-variant wrapped around FFXIV) did not change, but the data it loaded (i.e., FFXIV) did... which meant Rosetta 2 didn't see it as a changed executable to re-generate the ARM version.

The practical upshot of which was that it would work swimmingly when you first installed it, and then the first time you installed a patch it would just silently fail to run ever again. (Because the ARM binary was out of sync with the new version.) There were fairly simple ways to fix that—basically trickery to make Rosetta see the patched version as a new version—but they were not the sort of thing customer service teams would be likely to provide as advice. Or that the average user would probably feel entirely comfortable doing.

Sometime in the past few months, though, Apple seemingly fixed that issue in Rosetta 2; Crossover (and other WINE variants) seem to work fine now even when something in the bottle gets updated, which means FFXIV patching itself doesn't break the world any longer.

I wouldn't necessarily credit all of that to the M1 ARM architecture, the GPU's in any Intel MacBook barely qualify for the term 'GPU'. If they can get their 12th gen power consumption under control then Xe might weigh in but that ship has sailed as far as Apple is concerned.

[Packetdancer]

I wouldn't necessarily credit all of that to the M1 ARM architecture, the GPU's in any Intel MacBook barely qualify for the term 'GPU'. If they can get their 12th gen power consumption under control then Xe might weigh in but that ship has sailed as far as Apple is concerned.

I mean, there's some truth to that; the Intel Iris Pro is a very (very very) sad excuse for a GPU, while the M1's eGPU has comparable performance to a Radeon RX 560 according to the various benchmark sites. But I actually got curious after that post and hooked the Intel MacBook Pro up to the eGPU that I have for TensorFlow work (which is a Radeon RX 560, coincidentally enough), then ran FFXIV... and honestly, the performance was not really that different.

Moreover, the Intel MacBook Pro couldn't run the eGPU without being connected to power and a monitor, while the M1 was able to run without needing to be plugged in; that's definitely a point on the M1's side in terms of performance along with portability.

(That said, running it under Boot Camp on the Intel MacBook Pro with that same eGPU connected, the performance was unsurprisingly better than either of the other two scenarios.)