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.
3nm M2 Pro Chip for MacBook Pro Reportedly Entering Production Later This Year
- Thread starter MacRumors
- Start date
- Sort by reaction score
Almost certainly not.
Apple effectively only makes two Mac SoCs, and for good volume production design bandwidth reasons. The last thing they need to do is add MORE chips.
Rev'ing the M2 to 3nm isn't worth the time, resources, money.
I suspect the M2 Pro/Max are being made because TSMC could only do a low volume of 3nm when Apple needed them?
Names are just that and can be anything marketing wants them to be.
But....
The M2 is based on A15 cores designed in 5nm, so anything 3nm would need a redesign.
If the next "Pro/Max/Ultra" chips are based on the upcoming A16 cores it would be fair to label them "M3" (but Apple may choose to stick with M2 for marketing reasons). Problem is A16 is most likely still 5nm.
The might do some A16+ cores in 3nm exclusive for Mac use but I doubt that (and the naming options will still be the same).
Now with TSMC fully starting 3nm in early 2023 (at which volume?) I wouldn't be surprised if the 1st chip of that kind will be found in a proper MacPro which will not be able to share many things with M/A series if it is to support PCIe and RAM expansion.
HobeSoundDarryl
macrumors G5
How does Apple make ultra profits if they peak the chips at MAX? 😉
Last edited:
Not necessarily. Multiple alternative tracks are possible. For example
Track 1: The M class designs are for low-end portable (iPad, MBA, lowest MBP). Desktop and high end portable will get a new design with new features and a new branding (think Xeon vs i3/i5/i7). So the next chips could be the Q1 series?
Track 2: We keep M branding, but M2 was always a quick and dirty chip meant purely for the largest market. The real M1 successor, based on the A16 cores and requiring N3 because it is somewhat larger, will be called the M3 series, so M3 Pro/Max/Ultra/..., and there will never be an M2 Max/Pro. The M3 basic for iPads etc may appear middle of next year (or may never appear, as Apple manage to get all the schedules on sync with a unified M4 release in a year or so).
Meanwhile, apart from the branding issues, truth is, we have NO CLUE as to the actual constraints determining Apple's timelines.
Are the SoC designs ready, but waiting on TSMC?
Is TSMC ready, but waiting on a design issue within Apple?
Are both Apple and TSMC ready, but the computers for which these are targeted are all supposed to get something else that makes them really special (a boring version of this might be something like faster LPDDR5; a fancy version of this might be something like an Optane equivalent – MRAM based? secret company based?)
The "heat" of the M2 appears to be grossly exaggerated by exactly the sort of click-chasing bottom-feeders you'd expect.
Do you have an A15 based iPhone? Anyone who does will tell you that thing is freaking astonishing in how little energy it uses, even compared to previous iPhones. And yet, same stuff essentially as in the M2...
Same thing when I read reviews in places like the Blender forums, reviews by people whose primary concern is actually using the devices, not stirring up fauxtrage.
When I read these "heat" claims, written by people who clearly have no idea of the difference between power, energy, and energy-delay product, I do not find myself especially impressed...
If mass production on 3nm is truly starting next month doesn't that debunk the earlier claim by Trendforce that Intel's Meteor Lake issues have caused TSMC to delay 3nm?
An earlier article on this website (https://www.macrumors.com/2022/08/04/apple-chip-supplier-hit-by-intel-delay/?Bibblio_source=true) stated:
If 3nm mass production is full steam ahead then that debunks the Trendforce rumor that Intel's purported delays have hurt TSMC to the point of delaying the next advanced node.
Perhaps Intel Meteor Lake is not delayed after all (we'll know more next week at the 2022 Hot Chips conference), or even if it is delayed, TSMC 3nm is full steam ahead!
An earlier article on this website (https://www.macrumors.com/2022/08/04/apple-chip-supplier-hit-by-intel-delay/?Bibblio_source=true) stated:
If 3nm mass production is full steam ahead then that debunks the Trendforce rumor that Intel's purported delays have hurt TSMC to the point of delaying the next advanced node.
Perhaps Intel Meteor Lake is not delayed after all (we'll know more next week at the 2022 Hot Chips conference), or even if it is delayed, TSMC 3nm is full steam ahead!
Last edited:
Why in blazes would you whiff on using the “M3“ branding to describe the “3nm” process?
I mean great, & the sooner the better, but seems like a missed opportunity for the marketing dept.
I mean great, & the sooner the better, but seems like a missed opportunity for the marketing dept.
N3 is volume 2H2022. Theres been no update to this schedule in a while.
This 2H2022 is surely too late for iPhone, but NOT for the next gen high end mac chips (both because they have flexible schedules, and because the volumes are much lower).
There doesn't seem any good reason for Apple not to switch to having the highest end chips first. The low volumes mean they're perhaps a better target for getting the process optimized over the first few months, and while they are more complicated in some ways, they're also less absolutely demanding of super-low energy all the time, so minor flaws in the design (as long as they just impact energy or performance in a small way) can be tolerated, while the fix is added to the phone chip to be manufactured a few months later.
I'd bet big money that not a single person here would see a substantial boost in productivity with the M2 Pro or Max vs the M1 versions. These machines are so fast at most mundane tasks that a 15-20% increase will mean nothing. There are some heavy lifters out there and they know who they are but even video editors doing multiple 4k streams and even 8k streams are easily handled by today's M1 machines. That said, faster is better and maybe in a few years everyone will be shooting stereoscopic 3D 8K with their iphones so...
TSMC will begin production of 3nm chips for Apple by the end of 2022, according to a report this week from Taiwan's Commercial Times. A separate report from the publication claimed that TSMC will begin mass production of 3nm chips in September.
The report, citing industry insiders, claims that the M2 Pro chip may be the first to use TSMC's advanced 3nm process. Bloomberg's Mark Gurman previously claimed that Apple plans to use the M2 Pro chip in the next 14-inch and 16-inch MacBook Pro models and a high-end Mac mini, which could be announced later this year or in the first half of next year.
Apple's A17 Bionic chip for next year's iPhone 15 Pro models and the M3 chip, suitable for future MacBook Air and 13-inch MacBook Pro models, will also be manufactured based on TSMC's 3nm process, according to the Commercial Times.
DigiTimes also previously reported that TSMC would begin volume production of 3nm chips for Apple in the second half of 2022, including the M2 Pro chip.
Apple products transitioning from TSMC's 5nm process to 3nm process would naturally result in faster performance and improved power efficiency for future Macs and iPhones, which could contribute to longer battery life. Apple is nearly finished its two-year transition away from Intel processors in the Mac, with only the high-end Mac mini and the Mac Pro tower yet to switch over to Apple silicon chips.
Article Link: 3nm M2 Pro Chip for MacBook Pro Reportedly Entering Production Later This Year
Missed opportunity. They should be called the M2 Pro Max and the M2 Ultra Max Pro.
Less confusing that way.
Quite impossible? Not at all.
Apple has done this on multiple occasions.
The A10 was on TSMC N16, the A10X was on TSMC N10.
The A9 was on both TSMC and Samsung's 16nm class processes.
Going way back, the A5 was on Samsung 45nm and 32nm.
It really boils down to the abstraction level of the design. It's like writing code in a high level language vs assembly.
If you write in assembly (ie design at the lowest level) you can get the fastest possible code (high GHz design) but the design is crazy fragile – takes a long time to design, only works on one process. This is the Intel model.
But if you write in a high level language (ie design to abstractions) then you can "recompile" the design for a different process without too much pain. This is what Apple has done for years, and it seems unlikely that they would stop now.
The more interesting thing, however, is that N3 gives you more transistors to work with. If they are available, perhaps you will tune various aspects of the design to take advantage of that? The dumbest version of this is that you simply add more cores (CPU, GPU, NPU) vs what you might expect from the M1 Pro/Max. But you could use those transistors elsewhere. Perhaps instead you give each cluster a substantially larger L2, or a redesigned AMX block? Depending on how well parameterized the design is, you may even be able to start fiddling with the insides of each core (give each core more physical registers, for example) and still be able to fit within timing margin.
At the end of this is it still "just" an M2 derivative? It's kinda a meaningless question.
Perhaps we should start to think of these chips as iOS vs macOS? Both have common SW owned by Apple and run on a common underlying Darwin; but each team tweaks the details of Darwin and the higher level SW to exactly their needs.
Well, Max is three letters, Maximum is seven. Thus we can assume Max is only around 43% maximum power. Rounding we get 50%. The Max is half of the Ultra. Problem solved.
...makes all others look like chunks.So many chips. So little time.
Yes, it's totally strange they would stick with the same exact naming scheme of their products, huh?
Yep it’s supply and demand. Although the M1 and M2 are/will be used for more devices, Apple can effectively update its MBA and iPads by introducing new hardware and using that as the primary selling point.
3nm on M2 Pro and above makes sense because the performance gains are more important to that customer segment than the opposing direction; just power efficiency.
How long are you willing to wait?
So far, the most likely route might be an officially licensed W11 Arm running on a VM. There doesn't seem to be much interest on either side for direct booting via boot camp.