12% on its own is unnoticeable to basically everyone (even if you're cutting videos or playing really demanding games on it). Even intel hit that pretty frequently.
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.
Leaked Benchmarks Confirm M2 Chip is Up to 20% Faster Than M1
- Thread starter MacRumors
- Start date
- Sort by reaction score
12% on its own is unnoticeable to basically everyone (even if you're cutting videos or playing really demanding games on it). Even intel hit that pretty frequently.
(Probably) no. The M1 regular/Pro/Max/Ultra all run the cores at the same clock; the only difference is memory bandwidth, and it barely moves the needle at a single core. This will presumably be true of the M2* as well. They’ll be variants of the regular M2 with more cores and more high-end features — but at single-threaded performance, they’ll be identical.
Last edited:
Nah. It has 4% more space, so it barely seemed worth mentioning.
I'm also unsure what Apple means with their distinction between "LED-backlit display with IPS technology" and
"Liquid Retina display", and my guess is: not much of anything at all.
Colors, yes.
RAM: no — my comparison was chiefly about why you would buy the low-end M2 over the low-end M1. At the higher ends, the answer is indeed that you can spec it up higher.
Yeah it is.
This is true, and it's why even the low-end M1 has eight cores.
so multicore performance is much more important than single core. And any professional tasks always use multiple cores. For example doing XCode Builds an 8 core M1 Pro is about 20% faster than the original M1.
GitHub - devMEremenko/XcodeBenchmark: XcodeBenchmark measures the compilation time of a large codebase on iMac, MacBook, and Mac Pro
XcodeBenchmark measures the compilation time of a large codebase on iMac, MacBook, and Mac Pro - devMEremenko/XcodeBenchmarkgithub.com
The M1 Air's multicore performance is within 5% of the 8 core M1 Pro, meaning it should be roughly 15% faster at builds. If you do a hundred builds a day, even twenty seconds adds up.
The same is likely true for Photoshop, video editing, and every other professional tool you could use.
This is where you've lost me. You're bringing up Xcode, Photoshop, video editing, and "professional tools". Are we still talking about the MacBook Air? Because none of those are ideal on the Air. If you use those regularly as part of your job, get the 14-inch Pro.
If you casually use them, the Air may be good enough.
You really shouldn't.
Guess that's why the new MBA also gets the revised form factor and magsafe to sweeten the deal. The next MBA upgrade will likely reuse the same form factor with minimal changes to the design at best.
Maybe I'm looking at this from the wring direction but it says UP TO 20%. It needs to be qualified so we're doing Apples to Apples.
Single core for example goes up by 11.56%, and whilst it a little simplistic 3.2 increased by 9.1% gives you the 3.49GHz that the M2 is running at so how much is down to architectural changes and how much to just increasing the clock?
I mean the 12900k saw a leap of 80% in terms of performance compared to the 11900k in certain benchmarks but let’s not let facts gets in the way of a good story.
For Those of you expecting huge jumps every generation, it will never happen. I don’t believe they will ever let the new baby chip ever surpass the previous pro/max chip. It’s not good business to do so, tbh. Imagine if the base M2 was somehow faster in every way that the recently released, more expensive, Pro chip.
I’m making a baseless claim here, but what if M2 Pro/Max is actually based on TSMC N3/A16 cores - in order to really differentiate and create a much larger gap between the two classes of consumer and prosumer chips? So next year when baby chip M3 comes out also based on A16 cores, it will still be slower than M2 Pro/Max. Baseless speculation but I mean, they could.
I’m making a baseless claim here, but what if M2 Pro/Max is actually based on TSMC N3/A16 cores - in order to really differentiate and create a much larger gap between the two classes of consumer and prosumer chips? So next year when baby chip M3 comes out also based on A16 cores, it will still be slower than M2 Pro/Max. Baseless speculation but I mean, they could.
11900k is 8 cores, and 12900k is 16 cores. Of course it’s going to be 80% faster. Single thread for the performance cores is only around 18-22%. And there was a die shrink that happened also, 14nm to 10nm.
Single-threaded perf went up 9.3%, proving the point of the person you responded to.
Multi-threaded went up 60.1%, which is largely because they added 8 e-cores.
Lets not forget that Intel was asleep at the wheel before and getting absolutely hammered by AMD and Apple. So they decided maybe we should actually improve our processors 🤣
I think you described i well. See geekbench comparison bellow and you will see you get bigger the 10% gain only in few areas. And GPU increase relates to power draw increase by 33 %. But thanks to faster memory bus and other things we can be surprised by real life gains in demanding tasks and of course video editing.
TSMC 3nm won’t bring a lot by virtue of process alone. It will give Apple (and others) the capability to increase transistor counts and enhance the architecture, driving performance that way.
Apple can’t do magic, the modest improvements from M1 to M2 in just under two years show pretty well where we are at today compared with a couple of decades ago. But these devices are SoCs, and drive performance by adding and enhancing a number of subsystems. This time it was the enhanced neural engine and e-cores, video support and shift to LPDDR5. Next time it could be a further enhanced video block, spiffier NAND control, functional enhancements to the GPU, and more substantial I/O. For instance. None of which would show up on GeekBench single/multi, but would still enhance perceived performance. Lithography creeps forward at a snails pace these days, so it comes down to using the provided capabilities wisely.
To me, Apple seems to by and large do so, but then again I personally appreciate their modest power draw targets.
It is easier the other way around.
Geekbench when Apple Silicon loses = Geekbench is pointless for Apple Silicon and isn’t representative.
Geekbench when it’s good = Benchmarks confirm Apple is amazing
Geekbench when it’s good = Benchmarks confirm Apple is amazing
Last edited:
I’m excited to see the MBA in person.
I have the 14” Pro Max - I’m incredibly satisfied with the M1, but I would love a second machine as a lightweight utility and leave the 14” hooked up to my office setup.
I have the 14” Pro Max - I’m incredibly satisfied with the M1, but I would love a second machine as a lightweight utility and leave the 14” hooked up to my office setup.
Well, in this case, it's neither. M1 to M2 isn't amazing, nor is it terrible.
Geekbench, as far as I can tell, is quite representative across architectures and various workloads. What I find lacking are things like 1) what specific classes of cores does the CPU have (Apple and Intel now do two; Qualcomm even does three); 2) how do they behave, respectively (my understanding is the M2 e-cores got much more of a boost); 3) how does behavior differ between brief bursts vs. lengthy use (e.g. 20 seconds vs. 20 minutes).
The biggest surprise to the M2 is that they released it without transitioning the Mac Pro to AS. I thought they would have done that and then released the M2.
Yep. It's only another 4.2%, so we'll almost certainly break 2,000 with the M3.
Buuuuuut by that point, Geekbench 6 might be out
and will likely change the index again.