Every way that is Multicore.
M1 Pro 8P cores
M3 4 P cores
Geekbench
M1Pro 12575
M3 10450
Cinebench
M1 Pro 12390
M3 10430
GPU Gflops
M1 Pro 5300
M3 3500
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.
M2 vs. M3 Buyer's Guide: How Much Better Really Is M3?
- Thread starter MacRumors
- Start date
- Sort by reaction score
M1 Pro 8P cores
M3 4 P cores
Geekbench
M1Pro 12575
M3 10450
Cinebench
M1 Pro 12390
M3 10430
GPU Gflops
M1 Pro 5300
M3 3500
But, of course, that is the base M3
Something fishy going on here. You're not looking at the base spec M1 Pro but a more expensive BTO option. You copied the wrong geekbench results. https://browser.geekbench.com/mac-benchmarks
Geekbench multi core on 14" MacBooks:
M1 Pro 8 core: 10318
M3: 11608
I admit I was wrong about the GPU, though. I read that in a post here and didn't fact check it. M1 Pro takes that crown by about 20%.
Single core M3 blows the M1 Pro out of the water, though. So for music production the M3 should theoretically be better, right?
I guess, I am confused ... do we compare baseline M3 to M2 ... and then M3 Pro to M2 Pro ??
Exactly the question was M3 vs M1Pro.
My M1 Pro has 10 cores. It’s not built to order. It’s just not a binned chip, but an actual fully working M1 Pro.
The question was whether or not there is a use case for a chip between the M3 and the M3 Max.
As I described in my example there is, because the MPro chip is more powerful than the M chip.
Stop moving the goalposts. You asked your question specifically in relation to my M1 Pro. It’s a 10 core M1 Pro and like I said beats an M3 in all Multicore tests.
An M3 Pro beats an M3 in all Multicore tests as well.
Even then you seemingly inflated the numbers for the M1 Pro to help your cause, and lowered the numbers for the M3 too? Why? Anyone can look at the results like I did? Where did you get your numbers from if not the Geekbench website?
The difference between the fastest M1 Pro and the base M3 is minuscule in multi core! No one should be able to differentiate a ~3% multi core performance difference in real world music production...
In what way am I moving goalposts by noticing your numbers were wrong?
Obviously. Duh.
Last edited:
The numbers are straight from the CPU monkey website. Go plug in the M3 and M1 Pro 10 core and knock your self out.
Cinebench 2024- 802 vs 659. Now that’s not just 3% faster...I’ll leave you to do the math…but needless to say your number is wrong.
Geekbench5 12574 vs 10450. 20% faster, certainly not just 3%.
Cinebench R23 12390 vs 10437. 18% faster.
GPU 5300 vs 3550 50% faster.
Now the M3Pro stretches those Multicore margins even further! So I really don’t know what you are basing your 3% number on.
First you asked for someone to give a use case where the M3 wasn’t enough but the M3 Max was too much and I used my 10 core M1 Pro as an example for DAW use. An M3 Pro being more powerful than my laptop would be an even better example for DAW use that sits between the M3 and M3Max.
Then you moved the goalposts by trying to claim that maybe for enthusiasts it would be ok but not for professionals.
Then you moved the goalposts by talking about the Base model rather than the computer I said I used, to somehow claim the Pro chip is no faster than the M3!
So anyways, the MPro chip clearly fills the gap between the M and MMax chips for DAW use case and is a good 20% or more better in multicore performance.
Geekbench's website disagrees with your stats. After seeing how disparate your stats were with their own stats, I didn't bother to fact check the rest of your stats.
Let's give up on this conversation as it's going absolutely nowhere, we are simply arguing different points to each other, and we're just wasting each other's time. 👍🏼
Let's give up on this conversation as it's going absolutely nowhere, we are simply arguing different points to each other, and we're just wasting each other's time. 👍🏼
The numbers for Cinebench and Geekbench are right here.
Better you know the numbers than not.
Better you know the numbers than not.
A) Why continue this conversation?
B) Why do you/why would anyone trust a third party source for Geekbench results over Geekbench themselves?
C) Why even bother to mention Cinebench scores in an attempt to prove the M1 Pro is better at music production than the M3? Cinebench takes into account GPU performance, which isn't relevant. How embarrassing.
I won't read or respond to further messages from you.
I like my M3 base model ... Apple did nice work to improve the chip efficiency vs previous versions of chips. For me the low power usage is the main focus ... I am not a creative professional. I like to focus on the CPU per watt ... GPU per watt, neural engine efficiency ... if I can find them. I want to sip power.
Last edited:
A- because other people reading might base their purchasing decisions on your incorrect information.
B- because they did the tests
C- because Cinebench tests CPU as well.
Go to the Logic benchmark thread here on MR to see how the M3, M3 Pro 10 core and M3 Pro 12 core do against each other.
M3 Pro 12 Core 16" = 180 Tracks,
M3 Pro 10 Core 14" = 147 Tracks,
M3 14" = 123 Tracks.
The 12 core M3 Pro is 45% better.
The 10 core M3 Pro is 20% better.
For reference my 12 core M1 Pro is 181 tracks (8 P cores vs the M3 Pros 6)
If according to your numbers the M1Pro was only 3% faster than the M3 for multicore use then the difference would be about 4 tracks. But instead it’s 58.
You are wrong by an order of magnitude.
In the real world of DAW use, 58 tracks is quite a huge difference and would let people finish songs that couldn’t run on the M3 without freezing or bouncing down audio.
You are basing everything off one Geekbench 6 score and disregarding every other multicore benchmark to try to stick to your 3% difference claim. It’s ridiculous.
Anyone thinking about whether they should buy the M3 or the M3 pro for audio use needs to disregard your poor opinion and instead base their decision on actual benchmarks. The M3 Pro is a significant boost in power over the M3.
Last edited:
Now do a benchmark comparing the M2 Ultra and the M3 Ultra.
LOL 😂😂
The big difference between these Apple Silicon models is memory bandwidth. The M3 chips are all considerably slower than the M2 versions at memory transfers. This is important. The low end Macbook Airs etc are not designed for high performance. Its SSD and CPUs just can't cut it for high performance programming. But the M3 Ultra has the bandwidth and performance.
M3 - 100GB/s
M3 Pro - 150GB/s
M2 - 100GB/s
M2 Pro - 200GB/s
M2 Max - 400 GB/s
M2 Ultra - 800 GB/s
LOL 😂😂
The big difference between these Apple Silicon models is memory bandwidth. The M3 chips are all considerably slower than the M2 versions at memory transfers. This is important. The low end Macbook Airs etc are not designed for high performance. Its SSD and CPUs just can't cut it for high performance programming. But the M3 Ultra has the bandwidth and performance.
M3 - 100GB/s
M3 Pro - 150GB/s
M2 - 100GB/s
M2 Pro - 200GB/s
M2 Max - 400 GB/s
M2 Ultra - 800 GB/s
The theoretical M3 Ultra will presumably have the bandwidth and performance, you mean?
Out of interest, which specific tasks are likely to utilise all that lovely Ultra level bandwidth, where the 400GB/s does not cut it?
There is, thus far, no M3 Ultra.
Also, bandwidth above around 200 GiB/s is irrelevant for the CPU. Only the GPU can take advantage of that much bandwidth. (It might be better for the Ultra. But even then, those cores aren’t going to use 800.)
And the M1's memory bandwidth was around 67GB/s. It's obvious that you shouldn't buy any base MacBook Air if you have high performance needs unless you are willing to accept its compromises to get the Air's size and weight benefits.
Coding involves a lot more than "just creating text files." One usually has multiple environments and one needs to test things.
I agree however that the idea of dedicated chip is 100% wrong-headed thinking. One wants chips to be as generic as possible to optimize chip costs. Then different parts can be enabled/disbled in software to differentiate when necessary.
Anyone "expecting at least 4 or 5 years run with what they just bought" in 2023 with 8 GB RAM better be intending to keep minimalist usage for the full 5 years. Choosing 8 GB RAM in late 2023 (almost 2024) is flat bad decision making for any but lowest usages.
Note the emphasis on the word choosing. It ain't Apple that puts lame 8 GB in a box for the next 5+ years, it is buyer choice.
IMO folks worry too much about Apple support over time. I buy top boxes and choose long life cycles, and Apple support has never been an issue. The last OS upgrade comes at ~7 years and security upgrades continue for ~8-9 years. Hardware (especially RAM) always realistically ages out at ~4-6 years.
Many folks will run a box for much longer, but that involves choosing age-lame computing. Some folks like that apparently.
Last edited:
That’s exactly what I did. 64GB of RAM and a 1TB SSD. I’m sure it will be more than enough for a number of years.
Well, if you want to try local AI, I think the improved Neural Engine is very important in the new M3 chips. I know we are talking about the higher end with M3 Pro, Max, and, hopefully, M3 Ultra ... their GPUs + NE. Maybe, the Mac Pro box with Ultra + extra 'AI' cards is the 'ultimate' machine to study AI?