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.
First M2 Macs Expected to Launch Later This Year Using TSMC's 4nm Chip Process
- Thread starter MacRumors
- Start date
- Sort by reaction score
Odds are it would have modestly improved single core speeds over M1.PCIe lanes will be a given on some iteration. The desktop "Pro" will require that, but M1 already has a number of PCIe lanes because that's how Thunderbolt works.
I wonder if we won't see some kind of retro-future solution to the limited unified memory problem like a PCIe RAM disk with bog-standard slots that peaks at ~100GB/s of throughput. I'm really impressed with how transparent the M1 & MacOS combo handle swap and that's "just" with a 3GB/s SSD.
on the pro side, better single core score, more power efficient, longer battery life.
The M1 came out 2 months after the A14.
There was no M2 last year after the A15.
The first M2 devices are not expected now until fall.
Today's rumor itself says M2 is based on 4 nm process.
Both the timing and today's rumor suggest M2 will be based on the A16.
marked.
Mark my words: M2 will be released in Q3/4 2022 ... and don't be surprised to see M2 in the next iPhone ...
HobeSoundDarryl
macrumors G5
I wonder if the (not) secret connector that facilitated ULTRA could be engineered onto the other end in M2 to then link ANY number of M2 MAXes together in a strip.
I wonder if that connector could be engineered onto the sides of M2 to add a second dimension of linked MAXes: top, bottom, left or right.
Base Mac Pro with M2 "whatever" could start with perhaps M2 ULTRA (duo connection or maybe trio) to hit Apple "starting at" price for the big reveal. If the connector could be added to the sides, imagine the grid possibilities: 4 MAXes, 5, 6, 7, 8, 9, 10, etc. Is this X-grid, version 2 INSIDE one Mac instead of linking several Macs together?
If such a rumor comes out, it may be time to invest in copper futures.
All wild speculation... and I did NOT stay at a Holiday Inn last night. ?
Last edited:
The next iPhone will have an A16 which would be the M2 with fewer cores.
Even the base M chips are (and will be) to big/power_hungry to fit in a phone.
I agree with a late 2022 release. M2 in iPhone pro would be hilarious.
HobeSoundDarryl
macrumors G5
What if instead of M1 being Macs A-Series brand, it's more like OS-X branding... meaning that while the hardware inself improves with each iteration the letter & number painted on the chip stays the same for many years (much like I5, I7, I9 in former Macs)?
Of course, A-series makes it easy to anticipate M-series following suit with new numbers. But until there is an M2 or an N1 or whatever Apple chooses to call the next ones, there is no actual pattern yet. Personally, I do expect the next one to be called M2 but then again, I expected Mac Mini PRO & MAX on TUE too... and Apple went somewhere almost entirely different with that heavily-rumored (for months) expectation.
Last edited:
but of course they had takeouts/samples of M2 that started last year, they are probably close to tape out the final version soon if they haven't already.
And think about it, with the M1 now in the iPad Air, it going to be a question of time that A chips will be for lower demand devices (eg HomePod, monitors) and M chips will go Ito iPhone, no need to run these volumes on 2 different chips ...
It'll be interesting to see with using their own chips if Apple gets the Macs onto an iPhone-like scheduled release cycle. Not yearly I imagine but I wonder if it will end up being a thing where every generation is on an 18 month cycle or something like that.
I just can't see a 2022 MacBook Air using M1.
My bet is that the 2022 MacBook Air will have M2 and it will be faster than M1 by:
- 10% from N4P improved manufacturing tech
- 10% from new core designs
- 25% GPU boost by adding two extra cores
Perhaps plus a few neat tricks from an updated neural engine / media engine.
That all seems like a solid upgrade for a MacBook Air, but still keeps the chip relatively small and cheap.
The big question I have is whether single core performance will match or beat best of the x86 processors that will be coming out during the M2's lifespan.
My bet is that the 2022 MacBook Air will have M2 and it will be faster than M1 by:
- 10% from N4P improved manufacturing tech
- 10% from new core designs
- 25% GPU boost by adding two extra cores
Perhaps plus a few neat tricks from an updated neural engine / media engine.
That all seems like a solid upgrade for a MacBook Air, but still keeps the chip relatively small and cheap.
The big question I have is whether single core performance will match or beat best of the x86 processors that will be coming out during the M2's lifespan.
true, but you could artificially (eg fusing) "downgrade" M chips to have power efficiency like A chips, would be much more economical for Apple to only have 1 chip ...
Apple’s emphasis on the chip also limits its ability to choose when to release new Macs. It wouldn’t surprise me if they eventually go for an upgrade for all Macs around the same time every 2 years. Just like with iPhones both Apple and customers will know what to expect and will be able to plan ahead accordingly.
For example, I currently wouldn’t buy the M1 iMac while not knowing whether the M2 version is around the corner, but I would buy it if I knew it still takes a year. I would be able to make an informed decision and avoid buyer’s remorse.
For example, I currently wouldn’t buy the M1 iMac while not knowing whether the M2 version is around the corner, but I would buy it if I knew it still takes a year. I would be able to make an informed decision and avoid buyer’s remorse.
If it's really 4nm, I think it's probably based on a new chip design. The A15 is a 5nm chip, so unless M2 is "just" a die shrink of the A15, it might be based on the A16.
HobeSoundDarryl
macrumors G5
If we go by rumors, substantially slower & less powerful. But M2 Ultra will probably be at least an interesting percentage faster and a bit more powerful. The thinking is that M2 will be base M1 modestly faster and with a few more cores. M2 PRO > M1 PRO. M2 MAX > M1 MAX. M2 ULTRA > M1 ULTRA. And so on.
Wild guesses (emphasizing power) for 3 generations:
- M3 ULTRA > M3 MAX and M2 ULTRA
- M3 MAX > M2 MAX but maybe < M1 ULTRA
- M3 PRO > M3 & M2 PRO but probably/maybe < M2 MAX and probably/maybe M1 ULTRA
- M3 > M2 but probably < M2 PRO and probably/maybe <M1 MAX
- M2 ULTRA > M2 MAX and M1 ULTRA
- M2 MAX > M2 PRO but probably < M1 MAX
- M2 PRO > M2 but probably/maybe < M1 MAX
- M2 > M1 but probably < M1 PRO
- M1 ULTRA > M1 MAX
- M1 MAX > M1 PRO
- M1 PRO > M1
- M1
If the improvements are incremental like A-series, it's reasonable that M1 ULTRA > M3 and probably M4 and maybe M5. Base chips are power sippers tuned for light-to-medium usage in consumer Macs... PRO/MAX/ULTRA variants are the "power chips" for the prosumer/pro Macs.
Of course, whether it plays out that way is to be determined. We can't even be sure the next series will be called M2 until Apple actually establishes that by releasing a new number painted on the chips. It could just as easily be N1 or M1.2 (borrowing from the long-term Apple precedent with OS-X branding).
It seems most plausible the next one will be M2 but until Apple does it, that's mostly logical guessing because that's how A-Series chip advances are marked. Apple could opt to do anything with the next generation branding.
Last edited:
It's getting very confusing for me - once M2 chipset is out then all M1 model become outdate right?
| measuring unit | length (in meters) | abbreviation |
| planck length | 0, 000 000 000 000 000 000 000 000 000 000 000 016 162 412 meter | none |
| yoctometer | 0, 000 000 000 000 000 000 000 001 meter | ym |
| zeptometer | 0, 000 000 000 000 000 000 001 meter | zm |
| attometer | 0, 000 000 000 000 000 001 meter | am |
| femtometer | 0, 000 000 000 000 001 meter | fm |
| picometer | 0, 000 000 000 001 meter | pm |
| ångström | 0, 000 000 000 1 (10−10) meter | Å |
| nanometer | 0, 000 000 001 meter | nm |
| micrometer | 0, 000 001 meter | µm |
| millimeter | 0, 001 meter | mm |
| centimeter | 0, 01 meter | cm |
| inch | 0,025 4 meter | in or " |
| decimeter | 0, 1 meter | dm |
| meter | 1 meter | m |
| decameter | 10 meter | dam |
| hectometer | 100 meter | hm |
There's plenty left to shrink to.
We'll go into the 0.xxx measurements.
Let's remember that we can always make a number (or dimension) smaller. There is no theoretical limit to it.
Now, as far as physical limit is concerned, I have no clue as to how far we can keep miniaturizing.