MacBook Pro 16” M4 max vs. M2 max

[FalcHawk]

I currently have a MacBook Pro M2 Max and trying to decide if it is worth upgrading to the M4 max. Any opinions?

 

[Kotsos81]

It depends on your workload. If you feel that you need more speed in some CPU-depended applications, then it makes sense, since it is about 50% faster in single-core and 100%, i.e. 2x faster in multicore.

 

[FalcHawk]

It depends on your workload. If you feel that you need more speed in some CPU-depended applications, then it makes sense, since it is about 50% faster in single-core and 100%, i.e. 2x faster in multicore.

I have an application that can sometimes take 8 hours to run. Does that mean the M4 would run it in half the time?

 

[Kotsos81]

I have an application that can sometimes take 8 hours to run. Does that mean the M4 would run it in half the time?

If it is CPU-bounded, you will see great acceleration. It depends on the application type, but in most cases you will see anything from roughly 50% to 100% speed boost, in principle. Actual numbers may vary a little, of course.

 

[mr_roboto]

If it is CPU-bounded, you will see great acceleration. It depends on the application type, but in most cases you will see anything from roughly 50% to 100% speed boost, in principle. Actual numbers may vary a little, of course.

Actual numbers disagree. The best M4 Max is faster than the best M2 Max, of course, but not by that much.

Mac16,5 vs MacBook Pro (16-inch, 2023) - Geekbench

There are only two single-core subtests where M4 Max scores better than 50% faster than M2 Max. Not coincidentally, these two (Object Detection and Background Blur) are two of the three subtests documented as taking advantage of SME if it's available; SME is a new matrix-math instruction set first available in M4. If an application does not use SME, you can safely assume that +40% is the average boost and +50% boost the upper limit.

Multi-core scores are a bit closer to your optimism because M4 Max does have 50% more performance cores than M2 Max (12 vs 8), but despite the theoretical average being 1.4 (single-core boost) * 1.5 (core count boost) = 2.1x, the average is 1.72x. The downside of having lots more of the relatively power-hungry performance cores is that Apple has to downclock things more when all cores are running to keep power under control.

 

[Kotsos81]

Actual numbers disagree. The best M4 Max is faster than the best M2 Max, of course, but not by that much.

Mac16,5 vs MacBook Pro (16-inch, 2023) - Geekbench

There are only two single-core subtests where M4 Max scores better than 50% faster than M2 Max. Not coincidentally, these two (Object Detection and Background Blur) are two of the three subtests documented as taking advantage of SME if it's available; SME is a new matrix-math instruction set first available in M4. If an application does not use SME, you can safely assume that +40% is the average boost and +50% boost the upper limit.

Multi-core scores are a bit closer to your optimism because M4 Max does have 50% more performance cores than M2 Max (12 vs 8), but despite the theoretical average being 1.4 (single-core boost) * 1.5 (core count boost) = 2.1x, the average is 1.72x. The downside of having lots more of the relatively power-hungry performance cores is that Apple has to downclock things more when all cores are running to keep power under control.

Thanks for the details, very interesting.

Still 40% in single core and 70% in multicore is impressive boost, considering that we are talking for 2 generations apart, and can really make a difference if some application encounters a performance bottleneck in M2 Max.

 

[kitKAC]

I have an application that can sometimes take 8 hours to run. Does that mean the M4 would run it in half the time?

What's the app?