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.
Apple's First 3nm Chips for MacBook Pro Expected to Enter Production This Year
- Thread starter MacRumors
- Start date
- Sort by reaction score
Yeah, I agree. I kinda hope they push for performance since I honestly don't need more battery life on my M1 Pro MBP—it's already at a point where I just don't think about battery level anymore. It's plugged in when I'm working at home and the battery life is so good that I actually don't take my power adapter with me at all during the day anymore.
the only one I can find right now is this one which is in line with the report above
TSMC to Begin 3nm Chip Production Next Month
3nm chips from TSMC are getting closer.
But, risk production started months ago, design would have been final, volume is obviously lower and then potential yield impact, so we might see 3nm based products before the end of this year ...
Kuo? Really? Most of what he reports doesn’t come true. lol
Well, they should be comparing generation as well as the number.
We shouldn’t choose product names to appease the technologically ignorant. Lol
It still amazes me when I think of how small 3nm actually is.
Another rumor today says Apple won’t release 3nm chips until 2023. Perhaps Apple will announce the new Apple Silicon 3nm Mac Pro in December, and will ship them in early 2023… 🤷
I’m definitely holding out. My 2013 MPB is showing its age and I think 3nm will also be good for 10 years.
My question is: after the 3nm card is played, what's next?
I haven't heard of 2 or 1nm chips production yet, so my fear is that there will be stagnation for a while.
Not to mention after 1nm: I heard we hardly can go smaller.
And since I'm at it: what's gonna happen after we reach the miniaturization limit? To have more power the only option will be to make bigger chips?
Or we'll finally switch to quantum computers?
I haven't heard of 2 or 1nm chips production yet, so my fear is that there will be stagnation for a while.
Not to mention after 1nm: I heard we hardly can go smaller.
And since I'm at it: what's gonna happen after we reach the miniaturization limit? To have more power the only option will be to make bigger chips?
Or we'll finally switch to quantum computers?
Last edited:
I would like nothing more than for you to be right! However, with the chip constraints and shortages we still have I cannot see 3nm until probably mid 2023.the only one I can find right now is this one which is in line with the report above
TSMC to Begin 3nm Chip Production Next Month
3nm chips from TSMC are getting closer.www.tomshardware.com
But, risk production started months ago, design would have been final, volume is obviously lower and then potential yield impact, so we might see 3nm based products before the end of this year ...
Last edited:
Moore’s Law doesn’t exist anymore, it’s gone. Progress is slowing. As for: 7, 5, 3nm it’s all marketing at this point. Improvements are challenging, new materials and processes will help, but don’t expect great leaps and bounds over a 2 year cycle. Apple has an advantage as they control the chip design and the software that runs on it. They’ve done a stellar job with the M2 which will propagate to the M2 Pro, M2 Max, M2 Ultra and the M2 Extreme (or whatever name the high end chip they put into the next Apple Silicon Mac Pro). After those are out it will likely be decades before we ever see a 3X speed improvement for hardware. I’m not suggesting that innovation ends soon, just that what our devices do, will likely be more than what we expect from them today.
Last edited:
the chip "shortages" are for older nodes not for 3 or 5nm. Constraint yes as they can only produce a certain number of wafers, initially.
if you're so inclined, read this ...
Big Changes In Architectures, Transistors, Materials
Who's doing what in next-gen chips, and when they expect to do it.
semiengineering.com
Waiting to see the new Macs! The new Macs with 3nm chips may come out only in early 2023.
People tend to pay when they get product, not when you start to make it.
Making N3 dies is not like cooking some microwave popcorn or potato chips . TSMC is reported to be starting to put wafers into the production cycle in high volume manufacturing status in September. That process can be many weeks long until it comes out the other side as finished, tested, binned chip. If TSMC is doing some of the packaging ... even longer.
Even if the wafers finish in late November / early December then TSMC invoices folks to pay. Apple ships a payment weeks later... and ta-da you have slid into 2023. There is a decent chance some very small fraction of product might a full billing cycle done before get to
So TSMC is consistent here. We'll start making stuff in September and get paid for it in January. The Wall St. / investor folks really don't care about the first ( unless it has a short term financial impact). Don't really care about 3nm tech, just that 3nm tech generates more revenue.
This production cycle is long. That is one reason mega fabs have large power backup facilities. If there is a complete power loss many weeks worth of product goes completely down the crapper. A chip like a N3 could have something like 20-28 layers to be done until complete. So have to go through that many passes to complete. So if power goes out not only mess up dies with 1-10 layers done , but also dies with 1-15 , and 1-19 done. All of those would have be flushed and start back as 'blank' wafer.
Wafer starts for July-August production iPhone SoCs could start as far back as April (depending upon how early in July ramp to high volume). iPhone's show up in September because a ton of work is done in the previous Spring. Apple doesn't start collecting money for those iPhones until mid-late September. Few end users are paying in June for September iPhones.
Decent chance it will be in October and still March 2023. They have done June (WWDC) and December a couple of times. That 5+ months.
If they are going to do a significant shift in the Mac Pro form factor then they have a "Mac Pro is dead, Long live the Mac Pro" transition period where folks who are impacted by the form factor change have time to get purchase orders out or lined up and the 'wrong track' speculation can stop. (Apple did something similar at end of XServe phase which had a long off-ramp transition period before ending. )
If Apple rolls out a M2 Pro Mini and a "sneak peak" Mac Pro then they can claim "2 year transition" victory in October be done with it for 2022 dog-and-pony shows. Effectively closer to "well, you saw it in 2022" than "you could buy in 2022" but close enough to get on stage and do the victory dance.
They likely would have had made 1,000's of N3 chips during the 'at risk' phase months ago, so would have systems to do the 'look but don't touch' show with. But no volume to ship with for customers.
Pretty good chance these new Mac Pro's will cost at least as much as the current ones. 5 months lead time also gives folks time to find funds to pay for them. And for folks looking for some end of the year capital equipment write-off to go looking at something else.
Back in March when Apple said "one more mac to go" tagline tease would be a huge stretch to extend all the way out to December. In March, Apple probably did not now when N3 would going into high volume production. TSMC didn't know either. ( TSMC had a 2H '22 timeline which is bit 'hit the broad side of barn' limit. Could be July or December. Could be 2022 or 2023 ) . By October Apple would/should have ordered up 1-3K wafer starts. They would know the flow was going to come and arrive within in a +/- couple weeks target date range. It would be time to pass along some time expectation updates. They could throw out something vague like "Spring 2023" to not lock themselves down as tight as the data they had.
If Apple waited until December to say something there is likely something 'wrong' with the Mac Pro. That they need more time to come up with some "dog ate my homework" story.
What happens to chip production if China "Hong Kong's" Taiwan?
The UK handed over Hong Kong well over a decade ago. Nobody handed over Taiwan at all.
One of Hong Kong's issues is that they were/are in denial that they got handed over. They keep pointing that the "two systems" and China keeps pointing at the "One China". That "two systems" doesn't have a 'forever' clause attached to it. At the start, a 50 year window seems like a long time. At the start, UK's 99 year lease seemed a long time too. However, 20 years into a 50 year window is substantially shorter.
China has already 'taken' TSMC's N7. It will only take a several more years to take N5 (probably going to take some ASML stuff too to get that done). China doesn't need to 'blow up' chip production to actually get what matters over the long term.
No. Intel has already put things on a path. Instead of saying "2 nm" they are saying 20A (angstroms). Just shift the decimal point to the right one digit with a change of units and you back to double digit numbers you can count down. The bonus is now something that would have been 0.2 , 0.3 nm increments is whole digits. 20A -> 18A went two whole digits on the other side of the decimal point. 20A -> 17A . Smaller hurdle to jump over to show whole digit changes. With two digits to work with it will be the same as the 16 -> 3 path they took with 'nm' as the units.
That is going to perhaps have 'field day' effect upon on the clueless tech spec porn consumer market, but it is't going to work at all on engineers who actually have to hire the fabs to do work. They aren't going to get bamboozled by a change in units.
The core problem at play here is that increasingly trying to measure a 3D issue with a single number that historically used to represent 2D.
Moore's Law had a technologic part and a pragmatic economic impact part. One was the transistor count doubled around every 18-24 months. The economic corollary to that was that electronics got cheaper as they got smaller. If could do the same amount of work with a smaller chip (or went from multiple chips to less chips ) then the chip/package price went down.
The part that is more 'broken' than the density part is the prices go down part. Apple isn't primarily looking for a cheaper SoC. They are looking for an SoC that will help them hold the price points they charge. If the wafer costs go up a bit... they'll just pass those along (as much as they can ... they have large margin to work with) and keep going.