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.
M3 Chip for Macs and A17 Chip for iPhone 15 Pro Will Reportedly Use TSMC's Second-Generation 3nm Process
- Thread starter MacRumors
- Start date
- Sort by reaction score
No. Even when Windows machines with Arm-based chips start rolling out, we’re still a good decade or so away from talking about X64-86’s death. I’ll be happy to be rid of Intel as a daily driver, my work and custom build PCs are AMD and my future Mac will be running an M3. There’s no more intelligence with Intel.
Apple's future M3 chip for Macs and A17 chip for iPhone 15 Pro models will be manufactured based on TSMC's enhanced 3nm process known as N3E next year, according to a new report from Nikkei Asia. The devices are expected to launch throughout 2023.
N3E will offer improved performance and power efficiency compared to TSMC's first-generation 3nm process known as N3, according to the report.
In the meantime, the report claims that Apple plans to use TSMC's first-generation 3nm process for some of its upcoming iPad chips. It's unclear which iPad models the report is referring to, as rumors suggest that Apple will update the iPad Pro next month with the M2 chip, which is manufactured based on TSMC's second-generation 5nm process. A new entry-level iPad with an older A14 chip is also expected later this year.
The report claims that 2023 could mark the second year in a row in which only the Pro models of the new iPhone lineup feature Apple's latest chip. Last week, Apple unveiled iPhone 14 Pro models with an A16 chip based on TSMC's 4nm process, while the standard iPhone 14 and iPhone 14 Plus models are equipped with the previous-generation A15 chip.
Article Link: M3 Chip for Macs and A17 Chip for iPhone 15 Pro Will Reportedly Use TSMC's Second-Generation 3nm Process
TSM Stock has been really knocked down this year.
Take a look at TSM - it deserves to be valued at a higher PE ratio for sure. PE 17? Market Cap $400B
It even puts out a dividend.
It should be higher than $140 per share right now, and it's almost half that
1. Where does NVDA get it's best silicon from?
2. Where does Intel get it's best silicon from?
3. Where does AAPL get I'd best silicon from?
4. All the "semiconductor" companies get their best silicon from TSM I believe - am I wrong?
5. How far ahead is TSMC in terms of nanometers jive? Way ahead - like 2 or 4 years ahead of all others.
6. Fear of Taiwan and some Annexment to China is knocking the stock down? Really?
TSM is a critical company for the whole world! Without them, we would not be progressing so fast in terms of efficiency of mobile electronics. The whole bitcoin world used to depend on TSM for mining hardware - what happened to that?
Larger M2? You are presuming that the overall package size is going to grow. That is based upon the M1 die occupying 100% of the current package. It probably does not. There is the die size and then the contain/package it goes into. There a lots of I/O pins/pads that come out the bottom of the package. Typically those are the larger constraint on how small you can make the package. If it is a 20% bigger die but the I/O pin/pad output is the same then it is possible to end up sharing the same package dimensions (or pretty close).
The downside of the larger die is more likely that it costs Apple more; higher costs to juggle against other components not a huge logic board change. Apple is likely trading off new economies of scale for the memory packages and other parts of the SoC for the die cost increase. But to get to better economies of scale they need to get to more M2s made, not lots less. (pretty decent chance the Pro and Max are not going to directly help as they may have moved on from N5P and use different SoC components. )
N5P has power savings over N5. So the M2 doesn't have to be clocked all the way back to the M1. 5-10% power saving means they can clock it up 1-3% and still save power on broad core loads and turbo higher in bursts (up to whatever the thermals allow). The corner case were likely to burn additional power is where there are just lots mores transistors "iit up and working". So where the core count increases, dial that back. That won't impact the clock speeds (or performance) of the still active cores. If you have 10B N5 transistors working vs 10B N5P transistors working then really should have not some power and/or heat issues unless pushing up the clocks way too hard instead of just taking the "head room" that the new process provides. In short, don't be greedy. In the tablet market space, a M2 is a huge competitive win. It is way better than the 3-4 year old iPad Pros, Android stuff , etc. Apple doesn't have to dislodge most M1 iPad owners from their device to do well. They just need incrementally better to be successful. (lots of iPadOS changes and non-SoC component changes to move folks to a new iPad Pro/Air ).
As for a new board design. Other stuff is going to change besides the Mx SoC. Probably better radios. System/Power management. NAND chips. etc. A new board is coming anyway. So even if the M2 package is 1-2mm bigger on two dimensions the iPad isn't that small (relative to an iPhone.)
It is a mix. That should be more clear when AMD's Ryzen 7000 rolls out. The gap isn't going to be erased but it is going to be closer than it has been in years. Later in 2023 Apple will move lots of stuff into the N3 or N3-family in a substantial gap before AMD does so.
It isn't instruction set architecture as much as product focus. AMD and Intel have a huge interest in server and high margin desktop chips. They sell lots of laptop SoC, but not "mobile only" focused. So they have cores that are more so desktop cores blended with some features to make them more mobile friendly. Gated power management zones but not to the obsessive compulsive level to interfere with maximum overclocking. Intel's 'E cores' are more so small server cores than mobile phone/tablet focused cores. That family will be placed in relatively high core count server SoCs later.
Apple is somewhat the opposite. They really don't care about Server ( zero products there and really zero interest in going there except for desktop Macs repurposed into that role). Don't particularly care a whole lot about very customized high end desktop. If can cover the desktop space with a laptop optimized chip then just do that ( e.g. Ultra ... two MBP 16" chips 'glued' together).
So far AMD hasn't really made laptop SoCs their top priority. They tend to roll out last in the new architecture roll outs for Ryzen. That is not where they are putting maximum pressure on Intel. As AMD isn't as poor as they once were, and can afford to fight across a broader front, they are incrementally stepping up the pace. That is going to over time force more attention to detail in the mobile market on both of them. ( also not just Apple nipping indirectly at their Windows SoC market. Qualcomm is coming after them also in a more direct way. ) Perf/Watt is becoming a bigger issue in the server space also. AMD/Intel still spend substantive time worrying about the overclocking , hot rod fringe , but that is increasing not where the long term future for the bulk of their revenue is going to come from.
M2 Ultra using N5P likely has problems. The Max is actually relatively extremely large for a "chiplet". If they make it an even bigger die then using the TSMC packaging technology used for the M1 Ultra probably doesn't work. The M1 Ultra just squeezes in under the size limitation for that tech. TSMC has another tech to can build larger packages with but it is more expensive.
It is not likely that Apple is going to drop a massive pile of M2 variations before the end of 2022. If there is a M2 Ultra (or bigger) there is a pretty good chance it is sliding into 2023. Apple could put "M2 Ultra" SoC on top of TSMC N3 if they wanted to. The high level architecture and the fab process are two different things.
Right.
Next time you could put an /s at the end when it's anything but obvious it's sarcasm
Sounds like TSMC N3 has been canned.
https://fuse.wikichip.org/news/7048/n3e-replaces-n3-comes-in-many-flavors/
"The N3 node is weird. It is TSMC’s first 3-nanometer class process technology, but it won’t be the mainstream node everyone will utilize. In fact, it appears to be a one-off node that the company abandoned sometime prior to rollout. It looks as if TSMC engineers encounter some roadblocks along the way and decided to change things mid-way. We say all of this because TSMC moved on to a whole different node this year called “N3E” which we discuss below. Most of the messaging at the Technology Symposium was about N3E with the original N3 node getting just a brief pass-by mention. The big takeaway here is that N3E is a very different node from N3. While the nitty-gritty node details have not been disclosed, at a high-level it has different PPAs along with what the company claims are “very different” design rules intended to improve yield. The result of this is that N3E will not offer any direct migration path from N3, making N3 sort of a dead-end node for designers which is why TSMC expects most customers to utilize N3E instead. So why N3 exist at all? The easiest explanation is to satisfy customers’ commitment to those early technology adopters. Long term, the initial N3 node will likely fade into obscurity."
https://fuse.wikichip.org/news/7048/n3e-replaces-n3-comes-in-many-flavors/
"The N3 node is weird. It is TSMC’s first 3-nanometer class process technology, but it won’t be the mainstream node everyone will utilize. In fact, it appears to be a one-off node that the company abandoned sometime prior to rollout. It looks as if TSMC engineers encounter some roadblocks along the way and decided to change things mid-way. We say all of this because TSMC moved on to a whole different node this year called “N3E” which we discuss below. Most of the messaging at the Technology Symposium was about N3E with the original N3 node getting just a brief pass-by mention. The big takeaway here is that N3E is a very different node from N3. While the nitty-gritty node details have not been disclosed, at a high-level it has different PPAs along with what the company claims are “very different” design rules intended to improve yield. The result of this is that N3E will not offer any direct migration path from N3, making N3 sort of a dead-end node for designers which is why TSMC expects most customers to utilize N3E instead. So why N3 exist at all? The easiest explanation is to satisfy customers’ commitment to those early technology adopters. Long term, the initial N3 node will likely fade into obscurity."
Wow, so in essence, people rambling on about the new 3nm process being XX faster than the current 5nm.
They could be built exactly the same, however significantly faster that would warrant calling it an 'architectural upgrade'
That's actually wild.
Thanks dad.Right.
Next time you could put an /s at the end when it's anything but obvious it's sarcasm
TSMC Outlines 3nm Roadmap: Five FinFlex Nodes
TSMC's N3 node looks to have a big and bright future with FinFlex.
Sounds like primo ASi Mac Pro SoC material...
Pump up the volume...! ;^p
And here we are... Apple pretty much competing with itself as people get mixed information as to what flavor the next "M" will be and when it will be released.
The main issue that I see right now, is that Intel and AMD are now reporting processors that exceed 5ghz, which both companies have not been able to break for a very, very long time. Even I am excited about this news and Intel reporting a 6ghz processor on the horizon.
We don't get to see Apple's ghz equation - we just know that M2 must be faster than M1 and M3 must be faster than M2 and M1 and so on. I am sure M5 will be way faster than M1, M2, M3 and M4 and we know that Ultra is faster than Max which is faster than Pro. But we get confused when we begin to think is the M2 faster than the M1 Pro? It's not all that clear!
I think the best thing Apple should do is be upfront and provide a reliable road map and I would even suggest that now that all their processors is... fast and for most consumers... fast enough to do the following:
MacBook Air - Stay with M2 processor for three years
MacBook Pro - For God's sake, not but a straight processor in it - it's a pro - just have the M2 Pro and make it the fastest - forget the "Max" - match the processor to the name of the device it's in. Upgrade yearly.
Mac - It's a desktop - here is where you should put the M2 Max in - faster than the MacBook Pro - update every three years
Mac Mini - M2 or M2 Max - update every three years.
Those are my thoughts!
The main issue that I see right now, is that Intel and AMD are now reporting processors that exceed 5ghz, which both companies have not been able to break for a very, very long time. Even I am excited about this news and Intel reporting a 6ghz processor on the horizon.
We don't get to see Apple's ghz equation - we just know that M2 must be faster than M1 and M3 must be faster than M2 and M1 and so on. I am sure M5 will be way faster than M1, M2, M3 and M4 and we know that Ultra is faster than Max which is faster than Pro. But we get confused when we begin to think is the M2 faster than the M1 Pro? It's not all that clear!
I think the best thing Apple should do is be upfront and provide a reliable road map and I would even suggest that now that all their processors is... fast and for most consumers... fast enough to do the following:
MacBook Air - Stay with M2 processor for three years
MacBook Pro - For God's sake, not but a straight processor in it - it's a pro - just have the M2 Pro and make it the fastest - forget the "Max" - match the processor to the name of the device it's in. Upgrade yearly.
Mac - It's a desktop - here is where you should put the M2 Max in - faster than the MacBook Pro - update every three years
Mac Mini - M2 or M2 Max - update every three years.
Those are my thoughts!
It's actually quite dumb. At some point, we'll get into stacking, and then a new naming convention will arise.