Apple forgot about Hardware Raytracing Cores again... 😩
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.
Benchmark Results Reveal Graphics Performance of M2 Pro and M2 Max Chips
- Thread starter MacRumors
- Start date
- Sort by reaction score
I believe (if I remember correctly) Geek Bench GPU testing was proven to be flawed when it comes to Apple's M-series as it cannot push "tests" fast enough to completely flood all the cores, which leaves a certain number cores sitting idle while waiting for more tasks. (Basically the GB tests do not max out Mx GPU cores, especially on the higher end... the cores were finishing the tasks so fast that they went idle before the test could send another task.)
Side note: I believe it was MaxTech (don't hate) that was running some tests on an M1 Ultra and it showed that the GPU was only reaching a little over 70% usage. He couldn't understand why it wasn't being maxed out. But after running some other real world tests, it did finally max out and had the performance that was expected.
Last edited:
I will never buy an OLED computer monitor. Way too many static screens I leave up for literally days at a time without changing them, depending on what I need to be doing.
Same here. I put off getting a Mac Studio (didn’t really need to upgrade my iMac), but with a 3nm M3Pro or M3Max, the mini/Studio lineup will push me to upgrade (assuming the performance and efficiency scale as expected).
Reading is fundamental. The article never says Apple used the words “game changer.” That’s a queue to us, the readers.
Further, if you look at the linked article, Apple’s GPU vs. PC GPU performance boasts are right there. But that’s using integrated graphics. And not even lower-end PCs use integrated graphics, they use a separate dedicated graphics GPU/card. And even very cheap (~$250) PC dedicated GPUs destroy the graphical performance of M2.
I have the base model Mac Studio with M1 Max, 32GB ram. I use a 42 inch 4k monitor on my desk as my monitor (I primarily do video editing and like the large canvas of a 42 inch monitor to move windows around).
My gaming experience so far:
- World of Warcraft at 4k max no prob
- Dota 2 at 4k I have to turn off some things for it to still look good. Since it doesn't support metal it's very inefficient graphics wise but I figured out the right settings to have smooth game play and still look good at 4k
- Civ VI at 4k - it looks great and runs smooth even on a HUGE map.
My guess is that at some point (M3 and beyond maybe) Apple silicon Macs will be able to play pretty much any game out there if they were to make a Metal version of it and eventually there will be so many Macs with this spec that game studios will start porting their games over. Though I suspect it will be several years from now when enough people have M3 or later Macs before AAA titles are ported over at launch.
Last edited:
Here is a long list of ARM native Mac games. There may be others missing there, like SimCity 4 Deluxe, Grid Legends and No Man's Sky.
Last edited:
The 32 core m1max gets 67k but the 24 core gets 60k metal.
Definitely something wrong there.
That's not completely true. Under normal circumstances scaling percentages do diminish as core counts get higher, but that's usually due to (1) memory bandwidth and (2) voltage drop across each core.
(1) GPU require an enormous amount of bandwidth in order to keep all cores chugging along. Arguably, Apple has taken care of memory bandwidth so that shouldn't be an issue.
(2) Normally, the more cores you introduce in a chip the less power each gets to consume. Otherwise you'll end up with major thermal issues. That's why you see a 64-core AMD running at 2.7GHz and the 32-core variant running at 3.6GHz, with basically the same multi-core scores.
So, as long as your scheduler is efficient enough, and you can keep the power draw consistent and have the memory bandwidth to feed all the cores, there's no reason that it won't scale linear with each new core.
From what we've seen in the M1 generation, Apple was able to keep the speed consistent through all variants, because of how little power each core required. So the M1 GPU drew about 10W and the Ultra can draw close to 120W. (Would think it would be a simple doubling M1@10W, Pro@20W, Max@40, Ultra@80. However, the more cores, the more voltage is required to keep them running at the same speed. So it's more like, M1@10W, Pro@30W, Max@60W, and Ultra@120W)
Apple's GPU designs are extremely efficient and have room to add more and more power to allow the cores to run at the same speed across all their variants without having to worry about a thermal ceiling. (The entire M1 Ultra can draw well over 200W. This includes, CPU, GPU, Neural Engine, and RAM, among other things.) Technically this means we should see a linear increase in performance with each new core. And this has been shown in real world performance.
Last edited:
You mean speedy high-resolution graphics for certain kinds of modern games and certain resolutions (are we talking 4K at 144+ Hz?). I play Civ VI on a M1 MacBook Air with 8 GB of RAM without problems (> 30 FPS). It bogs down in late-game but I haven't had a computer not bog down; my main gaming computer is a Ryzen 7 5900X with a RTX 3070 and it can get noticeably slower late-game.
What if a gaming studio made a game that was optimized for macOS (Metal, etc)? Couldn't the M2 produce high-resolution graphics for the game with high framerates? Many of the limitations are due to poorly optimized games or translations (e.g., DirectX-to-Vulkan translation).
Also, the M2 in a MacBook Air peaks about maybe 9 W of power under heavy load. I don't know what the draw is in the new Mini but it's not likely much more. The M1 and M2 are impressive for how efficient they are.
Folks, these are the FASTEST, MOST POWERFUL, NEWEST Macs EVER in the history of Macs!
The first graphics-focused benchmark results have surfaced for Apple's M2 Pro and M2 Max chips, offering a closer look at GPU performance improvements.
Metal scores on Geekbench reveal that the M2 Pro with a 19-core GPU and M2 Max with a 38-core GPU in the new MacBook Pros offer around 30% faster graphics performance over the M1 Pro and M1 Max, in line with Apple's advertised claims.
The high-end M1 Ultra chip released for the Mac Studio last year is still about 9% faster than the M2 Max based on Metal scores:
However, OpenCL scores for the M2 Max and M1 Ultra are roughly on par.
Geekbench results also reveal that the M2 Pro and M2 Max in the new MacBook Pros both have single-core and multi-core scores of around 1,900 and 15,000, respectively, meaning they offer up to 20% faster CPU performance compared to the M1 Pro and M1 Max, which is also in line with Apple's advertised claims.
The new 14-inch and 16-inch MacBook Pros are available to pre-order now. The first deliveries to customers and in-store availability will begin Tuesday.
Article Link: Benchmark Results Reveal Graphics Performance of M2 Pro and M2 Max Chips
Indeed. These are pretty darn good increases given that the chips are on a very similar process as M1.
All signs have been pointing to the M3 as the one to look forward to for a while now. It'll be on the 3nm process and, if we believe the rumors, should have a good chance of having the new greatly upgraded GPU cores, maybe with ray tracing support.
Still, these M2 machines should be great for anyone who gets them. If I hadn't bought an M1 24" iMac a couple of years ago I'd be all over the M2 Pro Mini (of course what I really want is an Mx Pro 24" iMac... looks like that's not coming any time soon, although maybe with the M3?)
Anyway: exciting times for the Mac these days.
This is where all the geek bench related reviews and comments are wrong. They just focus on CPU speed which does reflect some apps but the apps that really need the boost tend to use GPU speed or a combination of CPU and GPU speed.
Its when you factor in both the CPU and GPU bump together used in unison that I think we see more value to the M2 over the M1.
Its still an incremental update but maybe not as incremental depending on the use case.
Its when you factor in both the CPU and GPU bump together used in unison that I think we see more value to the M2 over the M1.
Its still an incremental update but maybe not as incremental depending on the use case.
Where did you hear that? I can’t find anything on that statement
But that would be neat!
The OLED rumors say that Apple is planning on using a dual-layer OLED screen with two stacked layers of OLED panels . That prevents warping that can happen in panels at this size. It is also intended to reduce the chance of burn-in, probably because each layer doesn’t need to be as bright and higher output increases the risk of burn-in.