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.
Are we sure M3 is based on A17 Pro?
- Thread starter senttoschool
- Start date
- Sort by reaction score
BTW, regarding performance. Looking at GB6 single-core results and the chip frequencies, we get the following (I only consider highest scores I can find):
A14: ~ 740 points/GHz
M1: ~ 750 points/GHz
A15: ~ 750 points/GHz
M2: ~ 760 points/GHz
A16: ~ 760 points/GHz
A17: ~ 780 points/Ghz
M3: ~ 790 points/Ghz
Notice an interesting pattern?
A14: ~ 740 points/GHz
M1: ~ 750 points/GHz
A15: ~ 750 points/GHz
M2: ~ 760 points/GHz
A16: ~ 760 points/GHz
A17: ~ 780 points/Ghz
M3: ~ 790 points/Ghz
Notice an interesting pattern?
I did,but they didnt work. Mightve been my browser. Or he was too thorough for subtitling, too
The fact that m3 is on 3nm would suggest they share the same CPU design as the A17 Pro. Apple does not take old designs and apply them to new technology that's a waste of money and effort.
That is what Intel did for over a decade, and while Intel is down the toilet right now, it is not because of Tic-Toc strategy.
Not that this fact suggests m3 is a16, Im merely saying it’s a very proven strategy.
Yes, but the pattern relies on that A16 was always intended for 4nm, which we do not know for a fact.
The use of ~ is also handy to prove your point. It could just as well be that M3 is derived on A16 core, but 3nm vs 4nm and higher clocks allowed them to improve more than usual.
It is really down to whether high res die shots tells whether the core is A17 Pro or A16.
The most obvious pattern is this:
M1 on A14
M2 on A15
M3 on… A17?
Intel failed when they said Mobile CPU wouldnt be as big as desktop.. How is that working out for intel
What i think most fail to understand is M3 was actually made and should have been released Before A17 pro.. due to bad production Apple has been holding these devices since Summer 2023. So its actually the A17 pro thats based upon the M3, not the other way around.
Why does it matter what it was intended for? We have a concrete product with concrete performance.
Higher clocks don’t change IPC in the same design (at least not with such small differences). And yes, these are all approximate values because benchmarking is not an exact science, especially not if the results are taken from crowd-sourced data.
A better approach would be to look at multiple results and their empirical distributions, but I’m not doing all that work for a few internet points.
I find this very unlikely. A more reasonable explanation is that Apple was stockpiling chips to meet the demands. M2 prosumer hardware was only released in January 22, it would have made no sense updating it in summer.
Then again, Samsung did make the chips for the first few iPhone generations, but neither them nor ARM became truly dominant in the new era. It's now Apple Silicon versus Qualcomm. Superficially it's all about performance per watt, but deeply you're in trouble when your customer makes much more profit from your ideas than you do. Eventually he will find a way to eat your lunch too. Poor ARM! 😢 The true reason for Apple's success with A- and M-series is vertical integration down from the silicon up to the tv+ service running on top of it all. The Intel tick-tock model and TSMC's fabless manufacturing service are just tiny aspects of the whole competitive business.
The 2024 iPad Pro will get the T8132. Looks like this will be the "Real" M3 perhaps with the A17 NPU?
Interesting observation though I think the more likely case is that they're simply incrementing the IDs and not matching architectures. We already know that the M3 CPU and GPU are based on A17 Pro.
This is what I do not like about the M2 and M3 chips.
My 16” M1 Max MBP could be just as fast as the M2 Max no problem if Apple allowed us to overclock the M1 Max to the same clock frequencies.
You are in the end paying a lot of $$$$ for extra performance which we could have gotten for free with a simple overclock.
At for the M3 Max, you get 4 more P cores and that’s something you can’t surpass with an overclock.
look here:
Post in thread 'M3 vs M4'
https://forums.macrumors.com/threads/m3-vs-m4.2425630/post-33120194
Yes seem so 😆 I understood in this way the M4 got the “correct” CPU from A17 proDid you just link me to my own post?
No, the A17 Pro CPU behaves identically to that of M3 and it lacks the new instructions that seem to be part of M4. I did test both M3 and A17 so I’m quite confident about it.