Preliminary M2 games benchmark results.

[Colstan]

Andrew Tsai has released a short video focusing on preliminary benchmarks comparing the M2 to the M1 in games. More specifically, he is comparing a standard 13-inch MacBook Pro featuring a 10-core GPU, vs. a base M1.



Quick summation of his early results:

GTA V: 40%-55% faster than M1
Shadow of the Tomb Raider: 40%-45% faster than M1
Metro Exodus: 45%-50% faster than M1

Again, he stresses that these are early results, and he plans to do extensive benchmarks with "AAA" titles in the near future. Keep in mind that GTA V is running under CrossOver's compatibility layer, and that Tomb Raider and Metro Exodus are running under Rosetta 2. So, none of these games are Apple Silicon native. Andrew concludes by stating that Apple undersold the M2's 39% increase in GPU performance, when it in fact appears to be substantially higher. Presumably, he will include Arm native titles, such as Baldur's Gate 3, in his full performance analysis, as he has done previously.

When the A15 was released, much was made of the Blizzard efficiency cores, whereas gains elsewhere appeared limited. With the release of M2, it appears that Apple also made changes to allow the Avalanche performance cores to clock about 10% higher, and substantially improved GPU performance, beyond what the simple addition of two cores would allow. Until the A15 design made its way into a Mac, it appeared to mostly focus on efficiency, but the improved thermals and form factor allowed the M2 to show the full capabilities of the new core. (Keeping in mind that this is still an entry level 13-inch MacBook Pro, not what will reflect a high-end desktop model with a more powerful M2 variant.)

 

[Colstan]

Andrew Tsai finished his first benchmark video comparing an 8-core GPU M1 vs. a 10-core GPU M2.



Quick summation of his results, M2 performance increase over M1:

Witcher 3 (running under CrossOver): 55%
Minecraft (native Apple Silicon): 50%
Metro Exodus (Rosetta 2): 100%
Warhammer III (Rosetta 2): 100%

This bodes well for future Apple Silicon gaming, as long as Apple is successful in courting developers to make a Mac version of their games. Metro and Warhammer are particularly striking, since they are using Rosetta 2, yet doubled performance.

He stated previously that Apple's estimate of a 35% increase in GPU performance is way underselling the actual results, at least in games. I think Apple might have learned a lesson from that oddball graph they showed comparing a 3090 to the M1 Ultra. I think Apple were comparing them using similar wattage, but the Ultra doesn't match the performance of a 3090 once the wattage takes off. It's better to underpromise and overdeliver.

 

[diamond.g]

Andrew Tsai finished his first benchmark video comparing an 8-core GPU M1 vs. a 10-core GPU M2.



Quick summation of his results, M2 performance increase over M1:

Witcher 3 (running under CrossOver): 55%
Minecraft (native Apple Silicon): 50%
Metro Exodus (Rosetta 2): 100%
Warhammer III (Rosetta 2): 100%

This bodes well for future Apple Silicon gaming, as long as Apple is successful in courting developers to make a Mac version of their games. Metro and Warhammer are particularly striking, since they are using Rosetta 2, yet doubled performance.

He stated previously that Apple's estimate of a 35% increase in GPU performance is way underselling the actual results, at least in games. I think Apple might have learned a lesson from that oddball graph they showed comparing a 3090 to the M1 Ultra. I think Apple were comparing them using similar wattage, but the Ultra doesn't match the performance of a 3090 once the wattage takes off. It's better to underpromise and overdeliver.

I wonder if he is going to do an Air vs Air comparison to see if the improvements still hold up (his video was M1 Air vs M2 MacBook Pro).

 

[Colstan]

I wonder if he is going to do an Air vs Air comparison to see if the improvements still hold up (his video was M1 Air vs M2 MacBook Pro).

As a desktop user, I admit that I am more curious about how the M2 Pro/Max within a Mac mini or Mac Studio enclosure will perform with games. I'd say I'm even more interested in the M3 generation, because I think Apple will have more substantial improvements, having two notches already under their belts. I've been burned by early adoption before, so I think it's generally best to wait until third-generation products, even though I think Apple Silicon has been quite successful with the M1. If nothing else, M3 will likely feature Armv9 and, if @Cmaier's hunch is correct, ray tracing support.

I assume that by the time M3+ generations become available, game studios will have finally switched from x86 to Arm binaries. I figured they would be the slowest software developers to make the transition, and as a Mac gamer, I don't mind waiting it out.

 

[Colstan]

While clearly succeeding in his attempt to cosplay as evil mirror universe Mr. Spock, Vadim over at Max Tech did some test runs of various game titles.



Where Max Tech generally does well is with these "bakeoffs", usually done by his less sensationalistic brother. Still, it is fun to watch Vadim randomly tossing out numbers while trying to play a game which he has no idea how to control. My only complaint is that, for the Windows games, he is using a version of CrossOver with a known stutter bug, which has been improved in the CrossOver 22 beta. Blaming the hardware for a software issue, particularly an x86 translation layer, isn't perhaps the best conclusion to make.

These sorts of videos are actually useful, whereas when they are making a big deal over some obscure patent, pretending to be semiconductor engineers, or voiding warranties with thermal pads...they are less than helpful.

Of course, somebody in the comments section had to chime in with the "well, you can always build a PC cheaper...", as is tradition. Those idiots get annoying, and they are certainly not helpful.

 

[Jimmyjames]

Not sure how reputable this site is but it's claiming a significant increase in game performance on a recent Ventura beta.

https://venezuela.detailzero.com/technology/44542/Apple-greatly-improves-the-performance-of-its-GPUs-using-Metal-with-the-latest-beta-of-macOS-Ventura.html

I'd love to see someone with a 'M' series machine and access to the beta (and some games!) test the before and after performance.

 

[DT]

I still need to run some games on the daughter's new machine (if you didn't see, it's an MBA 16GB with the 8/10 CPU/GPU setup).