M5 Pro and Max unveiled

NotebookCheck's analysis of the M5 Max/Pro (also includes 15" M5 but not 13" for some reason):

www.notebookcheck.net

Apple M5 Pro & M5 Max CPU Analysis - M5 Max is not much faster than the M4 Max

Notebookcheck analysis/review of the Apple M5 Pro and M5 Max SoCs with the new 10-core CPU.
www.notebookcheck.net www.notebookcheck.net

M5 Max is doesn't do great here, but for some reason Apple shipped the 14" laptop with 128GB of RAM and it's clear the 14" model simply cannot cool the chip for sustained performance - so GB 6 shows really nice gains over the M4 Max, but CB R24 does not. Though efficiency improves a little. M5 Pro which has the same CPU is for some reason slower and less efficient than the M5 Max (though still an improvement over the top end M4 Pro as again now it has the same core count as the Max).
 
dada_dave said:
NotebookCheck's analysis of the M5 Max/Pro (also includes 15" M5 but not 13" for some reason):

www.notebookcheck.net

Apple M5 Pro & M5 Max CPU Analysis - M5 Max is not much faster than the M4 Max

Notebookcheck analysis/review of the Apple M5 Pro and M5 Max SoCs with the new 10-core CPU.
www.notebookcheck.net www.notebookcheck.net

M5 Max is doesn't do great here, but for some reason Apple shipped the 14" laptop with 128GB of RAM and it's clear the 14" model simply cannot cool the chip for sustained performance - so GB 6 shows really nice gains over the M4 Max, but CB R24 does notThough efficiency improves a little. M5 Pro which has the same CPU is for some reason slower and less efficient than the M5 Max (though still an improvement over the top end M4 Pro as again now it has the same core count as the Max).
Click to expand...
I skimmed the article and find their conclusion perplexing. Doesn’t seem accurate to say it’s not faster. Also the claim that the X Elite has caught up but they haven’t tested it yet is a weird thing to say.
 
Jimmyjames said:
I skimmed the article and find their conclusion perplexing. Doesn’t seem accurate to say it’s not faster. Also the claim that the X Elite has caught up but they haven’t tested it yet is a weird thing to say.
Click to expand...

The burst performance of the M5 Max is higher, but the M4 Max scores practically identically to the M5 Max in CB R24 indicating poor sustained performance. So for sustained loads, important for Max-level devices, it is not any faster. Hopefully this is just a 14” problem. However, again similarly the 18-core Pro which is the same CPU is not any faster in the 16”.

As for the X Elite, they may be going off of Qualcomm’s own benchmarks which naturally are suspect until actual devices are out.
 
I am noticing a lot of these reviews -- both written and video -- are poorly done, and they seem rushed. There are a slew of typos, mistakes, just weird shit. These are professionals, supposedly, and they've had time with the machines, yet can't manage to find errors in their own processes. For example, a popular YouTuber that benchmarks transformer models completely screwed up testing in the video because he didn't fully offload the model to the GPU....... which is like the entire point of the video. Then he literally asks "why is only part of the GPU performing?" Someone pointed this out in comments, but it's far from only this YouTuber, and a bunch of reviews I'm reading are just extremely poorly done. I put zero faith in most of these reviews, including ones with nonsensical conclusions. I'm very disappointed and irritated.
 
dada_dave said:
The burst performance of the M5 Max is higher, but the M4 Max scores practically identically to the M5 Max in CB R24 indicating poor sustained performance. So for sustained loads, important for Max-level devices, it is not any faster. Hopefully this is just a 14” problem. However, again similarly the 18-core Pro which is the same CPU is not any faster in the 16”.

As for the X Elite, they may be going off of Qualcomm’s own benchmarks which naturally are suspect until actual devices are out.
Click to expand...
It seems there is an improvement on Cinebench 26 so perhaps it’s a matter of software optimization? 11% iirc. Also I am a little reluctant to draw too many conclusions from one render focused benchmark.
 
Jimmyjames said:
It seems there is an improvement on Cinebench 26 so perhaps it’s a matter of software optimization? 11% iirc.
Click to expand...

Perhaps, if so it is more likely that Cinebench 26, which otherwise seems to favor older (and x86) CPUs, is responding to the 2 extra cores/threads better than Cinebench 24.

Jimmyjames said:
Also I am a little reluctant to draw too many conclusions from one render focused benchmark.
Click to expand...

Of course one should never draw sweeping conclusions from a single benchmark, but that's still not a great result on a benchmark which Apple Silicon has typically performed very well on.

dada_dave said:
The burst performance of the M5 Max is higher, but the M4 Max scores practically identically to the M5 Max in CB R24 indicating poor sustained performance. So for sustained loads, important for Max-level devices, it is not any faster. Hopefully this is just a 14” problem. However, again similarly the 18-core Pro which is the same CPU is not any faster in the 16”.

As for the X Elite, they may be going off of Qualcomm’s own benchmarks which naturally are suspect until actual devices are out.
Click to expand...

Actually it turns out Notebookcheck did test the Elite 2 themselves on a hands-on, however, it was a reference device, so again real world devices may behave differently.
 
Last edited:


This is an actually pretty well-produced video. The final test doing 3 things at once is enlightening. It's very amazing what M5 Max achieves. It's revolutionary
 
This score from the iphonedo review shows a 24% multicore increase in CB 26.
1773100258698.png
 
In the event you don't want to watch, which you absolutely should watch because it's one of the only coherent videos I've seen, he runs:

1. A 45 minute audio interview transcript with 5 tasks using 70 billion 6 bit transformer model
2. 1024 by 1024 pixel photo using a diffusion model
3. Rendering DaVinci Resolve 8K video with a 4K timeline

At the same time.

The highest end M3 using 80 cores and 512 GB of memory with 8TB of storage took 1-3 minutes longer across each task than the M5 Max did IN A NOTEBOOK on battery power, and it only drained 10% in that stress testing.

It's a true testament to Apple silicon! It's so amazing.
 
Jimmyjames said:
This score from the iphonedo review shows a 24% multicore increase in CB 26.
View attachment 38316
Click to expand...
Apple gave him the 16" model - a pity he didn't run Cinebench 2024 as well ... and a pity Notebookcheck didn't run 2026 ... sigh it's so hard to figure out what is going on with incomplete data. I really wish I had the time and wherewithal to generate my own ...
 
dada_dave said:
Apple gave him the 16" model - a pity he didn't run Cinebench 2024 as well ... and a pity Notebookcheck didn't run 2026 ... sigh it's so hard to figure out what is going on with incomplete data. I really wish I had the time and wherewithal to generate my own ...
Click to expand...
Yeah I guess I’m put off by the headline that the Max showed little improvement when what they mean is the 14” shows little improvement when compared to the 16” M4 Max
 
Jimmyjames said:
Yeah I guess I’m put off by the headline that the Max showed little improvement when what they mean is the 14” shows little improvement when compared to the 16” M4 Max
Click to expand...
Fair, the one note of concern though is that the 16" Pro has the same CPU but with cut down memory bandwidth as the M5 Max and for some reason it's slower than the 14" M5 Max and is less efficient as well! Every previous time I've gone with the "it's the memory bandwidth" hypothesis to explain odd CB 24 results, I've usually, eventually, come to the conclusion that it wasn't. But here ... again, I don't know how else to explain this even though, as CPU bandwidth goes, the M5 Pro has a ton! (Not sure how much it can access admittedly) I mean, workstation CPU level bandwidth. So I dunno what's going on.

So it can't just be the 14"'s lack of cooling.
 
RockRock8 said:
In the event you don't want to watch, which you absolutely should watch because it's one of the only coherent videos I've seen, he runs:

1. A 45 minute audio interview transcript with 5 tasks using 70 billion 6 bit transformer model
2. 1024 by 1024 pixel photo using a diffusion model
3. Rendering DaVinci Resolve 8K video with a 4K timeline

At the same time.

The highest end M3 using 80 cores and 512 GB of memory with 8TB of storage took 1-3 minutes longer across each task than the M5 Max did IN A NOTEBOOK on battery power, and it only drained 10% in that stress testing.

It's a true testament to Apple silicon! It's so amazing.
Click to expand...
This is basically the tell that any reviews mentioning little improvement were likely done incorrectly.

There is obviously major improvements with the GPU on M5, but also the Media Engine received updates too (I've seen testing of the base M5).

Combine this with the Fusion Architecture and brand new Performance core paired with the world's fastest core (Super core), you're getting a machine that can smoke a desktop workstation class computer on battery power with ease, despite 4X less memory, 200 GB per s less bandwidth, 14 less CPU cores, and 40 less GPU cores.

Might want to think about how amazing that is lol.
 


In addition to the amazing performance increases (2X the tracks to his notebook for example), this person noticed that Logic Pro only currently supports 12 cores. He noticed that 6 super cores and a group of 6 performance cores would be active, and that it would switch between 1 group and another group of 6 performance cores. So he would use 18 cores, but only 12 cores at single time.
 
More Cyberpunk benchmarks. The lack of increase in some scores is weird. Again from Hardware Canucks
1773104887129.png


This Warhammer 3: Total War shows a similar lack of improvement.
1773104996765.png
 
MerryCherry said:
Are the 12 P-cores on a single cluster with 16 MB shared L2?

Or is it two clusters of 6 P-cores each, with 16 MB sL2 per cluster?

The former is not impossible, but that means less cache per core, which might lead to cache starvation?
Click to expand...
From @JRLMustang: https://arstechnica.com/gadgets/202...ook-pro-m5-max-and-its-new-performance-cores/

dada_dave said:
Welcome to the forum! That is a very good question. The "E-"cores are currently 4-6 cores per cluster with 4MB of cache, so 12 "P-"cores with 16MB of cache isn't impossible - in terms of width those "P-"cores are closer to "E-"cores, but in terms of clock speed, closer to "S-"cores (if the leaks are accurate which we'll find out soon). So I could see either configuration.
Click to expand...

The M5 6-E-core cluster has 6MB of cache while there are a 2 clusters of 6 "P-"cores each with 8MB of cache and the 6-"S-"core cluster has 16MB.

JRLMustang said:
For those interested, Andrew Cunningham over at Ars is apparently putting the Max through its paces:

“Testing Apple's 2026 16-inch MacBook Pro, M5 Max, and its new "performance" cores”

arstechnica.com

Testing Apple's 2026 16-inch MacBook Pro, M5 Max, and its new "performance" cores

M5 Pro Max's "performance" CPU cores definitely aren't just rebranded E-cores.
arstechnica.com arstechnica.com

“M5 Pro Max's "performance" CPU cores definitely aren't just rebranded E-cores.”
Click to expand...
 
Back
Top