I’m interested in a comparison between the M3 Max 30-core and M2 Max 38-core, especially since they reduced memory bandwidth from 400 to 300 GB/s in that M3 Max version. Has anyone tested it yet?
I’m interested in a comparison between the M3 Max 30-core and M2 Max 38-core, especially since they reduced memory bandwidth from 400 to 300 GB/s in that M3 Max version. Has anyone tested it yet?
Interesting. That really needs to be done.
I’ve recently watch an Artisright video on YouTube. First time seeing that channel and he claims to have done a bunch of testing and has created graphs that supposedly show there isn’t much difference between the two m3 max SOCs. I just don’t know how much to trust his claims since he seems to contradict some of the more proven reviewers.
To me, 1 100 GT/s drop is huge. I think it’s more of a possible bottleneck issue though and you won’t see anything except the core count difference in performance until you’re beating the thing to death with data using after effects or some scientific software.
I’m interested in a comparison between the M3 Max 30-core and M2 Max 38-core, especially since they reduced memory bandwidth from 400 to 300 GB/s in that M3 Max version. Has anyone tested it yet?
Interesting. That really needs to be done.
I’ve recently watch an Artisright video on YouTube. First time seeing that channel and he claims to have done a bunch of testing and has created graphs that supposedly show there isn’t much difference between the two m3 max SOCs. I just don’t know how much to trust his claims since he seems to contradict some of the more proven reviewers.
To me, 1 100 GT/s drop is huge. I think it’s more of a possible bottleneck issue though and you won’t see anything except the core count difference in performance until you’re beating the thing to death with data using after effects or some scientific software.
I'll try to convince the folks that helped me with this review to pick up a M2 Max 38-core. Anecdotally, other folks I've spoken to about it during the process of making this review are seeing gains in the M3 Max 30 versus the M2 38-core, even when hammering the computers, but I don't have specific data because they are outside the folks that use these packages.
I prefer the comparisons to the M1 MBP because there are far more of those users eyeing up this product (myself included) than anyone who just bought an M2 a few months ago.
Could AI do a story that identifies the software that can utilize multiple CPU cores to enhance performance? And the software that utilizes single cores/single threading that can’t benefit from multiple cores? All the hype about multi core performance is meaningless if you don’t run software that is written to support multiple threads.
"Multi-core" isn't just about splitting a single task to run across multiple cores there's also a need for smooth multitasking. The more cores you have the more a system can process at the same time; today's operating systems run hundreds of processes and thousands of threads in the background, so those extra cores come in handy.
And I would say that CPUs today are smart enough to know what task can be broken off into its own thread even if the app wasn't specifically designed for it. Even Apple's SoC's can take a task and determine which processor would be best to complete it; CPU, GPU, ANE, ISP, etc.
If you are a time is money person then you are not buying a laptop. You are probably buying a high PC desktop system with NVDA graphics cards that give top performance in maya, premiere pro, photoshop etc.
How many of those people drag that desktop PC on the road with them? Designers may work 100% in a studio, but photographers and videographers (and I'm sure many others) need something in the field.
Wait! In what world does a 16 core CPU, keep pace with a 24 core CPU on multi-core score? Even allowing for 16% faster single core score, this math doesn't work.
Multi-core CPU performance does not scale linearly.
Just go look at the GB scores for those high core CPUs...
Wait! In what world does a 16 core CPU, keep pace with a 24 core CPU on multi-core score? Even allowing for 16% faster single core score, this math doesn't work.
The most notable thing here is not the M3 performance but that Apple is capable of delivering 3 versions of one of the most complex SoC’s in the world on the world’s most advanced node in one step. Competitor design houses, take note. Also, remember (to everyone’s surprise and puzzlement) when they bought all the flash memory around the world - they knew what was coming next. They’ve bought all TSMC’s 3nm capacity, so what comes next…?
I'll try to convince the folks that helped me with this review to pick up a M2 Max 38-core. Anecdotally, other folks I've spoken to about it during the process of making this review are seeing gains in the M3 Max 30 versus the M2 38-core, even when hammering the computers, but I don't have specific data because they are outside the folks that use these packages.
I'm sure all the figures for the M2 38-core are out there, but probably few people have the new M3 30 GPU-core variant yet. This is the model I'm most interested in, because as much as the 40-core looks phenomenal, it's just a little outside my budget.
Comments
"Multi-core" isn't just about splitting a single task to run across multiple cores there's also a need for smooth multitasking. The more cores you have the more a system can process at the same time; today's operating systems run hundreds of processes and thousands of threads in the background, so those extra cores come in handy.
And I would say that CPUs today are smart enough to know what task can be broken off into its own thread even if the app wasn't specifically designed for it. Even Apple's SoC's can take a task and determine which processor would be best to complete it; CPU, GPU, ANE, ISP, etc.
How many of those people drag that desktop PC on the road with them? Designers may work 100% in a studio, but photographers and videographers (and I'm sure many others) need something in the field.
Multi-core CPU performance does not scale linearly.
Just go look at the GB scores for those high core CPUs...
96-core AMD -> 18,000
64-core AMD -> 20,000
56-core Intel -> 19,000
24-core Intel -> 21,000
16-core AMD -> 19,000 (120W TDP)
16-core M3 Max -> 21,000 (30W TDP)
M3 Max: 16 (12× high-performance + 4× high-efficiency)
75% of the M3 cores are high performance, and only 66% of the M2 Ultra cores are. Maybe that has something to do with it?