Apple Likely Already Designing Chips Based on TSMC's 2nm Process

[NT1440]

MR, can you guys even bother to do your homework?

If a new node is being developed, Apple (and others) have already been working for years on designing to the new spec. This is a tech forum, the bare minimum should be explaining how this process works…

 

[ChrisA]

If you plot these process numbers on a graph the line slopes down. Using this line I predict negative nanometers will be reached in the 2030's. At first, these will be modest like -.5 nm and, then over time we will have negative numbers with larger magnitudes such as -10nm. Power consumption seems to track feature size and so with these large negatives, we can expect our computers to actually generate power and no longer need batteries. With larger negative features later in the century we can power our homes from our Apple devices.

OK, either the above is correct and the line will cross zero. Or the line will bend horizontally and track just above zero and no longer go down. In the first case, we have a future power source. In the second case, we are living in special times where the line is still moving down. These times will not last because very soon there will be no more "down".

What will Apple do when they run out of room to decrease node size? It can't continue forever because atoms have a finite size and no matter what was said above, you really can't cross the zero line. The hard barrier is a fraction of a nanometer above zero and they are getting close.

What will they do? Improving performance after hitting the feature size limit will require adding hardware and increasing power use or using smarter designs that get better performance from fewer transistors. This could mean more kinds or cores and software designed to use extreme parallelism. If will be fun to watch because hitting the feature size likely will FORCE innovation. The current trend is very boring

 

[AgeOfSpiracles]

Actually, most people do not understand that for many, many years now, “nm” has just been a marketing scheme (clearly MacRumors doesn’t).

Clearly? Because this "controversy" is mentioned in the comments of every article that makes minor mention of this or that process node. Does anyone EVER reply to these comments with "Hmm, didn't know that, thanks" ?

 

[NT1440]

If you plot these process numbers on a graph the line slopes down. Using this line I predict negative nanometers will be reached in the 2030's. At first, these will be modest like -.5 nm and, then over time we will have negative numbers with larger magnitudes such as -10nm. Power consumption seems to track feature size and so with these large negatives, we can expect our computers to actually generate power and no longer need batteries. With larger negative features later in the century we can power our homes from our Apple devices.

OK, either the above is correct and the line will cross zero. Or the line will bend horizontally and track just above zero and no longer go down. In the first case, we have a future power source. In the second case, we are living in special times where the line is still moving down. These times will not last because very soon there will be no more "down".

What will Apple do when they run out of room to decrease node size? It can't continue forever because atoms have a finite size and no matter what was said above, you really can't cross the zero line. The hard barrier is a fraction of a nanometer above zero and they are getting close.

What will they do? Improving performance after hitting the feature size limit will require adding hardware and increasing power use or using smarter designs that get better performance from fewer transistors. This could mean more kinds or cores and software designed to use extreme parallelism. If will be fun to watch because hitting the feature size likely will FORCE innovation. The current trend is very boring

I think what you’re describing is exactly why Apple has been focused on performance per watt. It’s a long game, but they’re well positioned for power overhead while Intel and AMD are just now starting to take that seriously (but still just pump enormous power through their high end for spec-chasers).

 

[ScooterComputer]

Apple acquired all of TSMC's 3-nanometer chips in 2023 for iPhones, iPads, and Macs, for example.

Should be noted that Apple acquired all of TSMC's 3nm "N3" process chips, not 'all of [their] 3-nanometer chips". And reporting showed that Apple did so almost begrudgingly, and only after TSMC had shopped N3 capacity to everyone else, all of whom declined to buy, after Apple initially declined to buy N3 "M3" chips (for the Mac). This is evidenced by how early there had been rumors of M3 Pro MacBook Pros… but without Apple shipping any, and then how initial M3 Pro MBPs came with older versions of macOS. They'd been manufactured, sitting in warehouses, awaiting a "go, no go" decision! You can bet Apple muscled up in that deal. I'm guessing Apple not only got the remainder of the N3 capacity for super-duper-cheap—as TSMC needed to clear it in order to make way for N3e process production—and you can bet Apple made sure, as part of that leverage, to get a good chunk (if not all) of the N3e capacity as well.
Plus it had been hypothesized that Apple had come to a point, given the N3 process' not-so-great (compared to 5nm N5/N4 process) performance/watt gains, of having decided to just wait for TSMC to get N3e online, leapfrogging N3 to N3e for the MBP and desktops, but that yields and build-out delays forced their hand and they had to ship the N3 M3 Pro lineup that they'd been sitting on.
So I'll not be surprised if we see a stealth N3e 'bump' for the MBPs when refreshed M3 Pro/Max/Ultra Mac Studios, Mac minis, and Mac Pros ship sometime later this year. (If Apple doesn't upgrade those models this year, something is very wrong within Apple Hardware.) The N3e will only be incrementally faster, but will be better with power consumption; which means faster at higher power draws, which fits for the desktops. (Would LOVE to see a M3 Pro iMac too. But, who knows?)

2nm isn't anywhere near production ready, another year out. And 1.4nm is still at very early stages, at least 2 years out… at least. So anyone waiting for significant performance boosts from Apple Silicon will be waiting quite a while.

ALSO… quite interesting, the coincidence of this announcement/leak, given the rumblings of rumored Chinese aggression towards Taiwan within the next year. Puzzles me why Apple isn't being more forceful with TSMC to build-out fab capacity in the U.S. If China makes moves against Taiwan, every one in the tech space is going to feel the pain, but especially fab-less players like Apple and Nvidia. (And given that Apple spurned Intel so publicly, can't imagine even "new and improved" Intel would welcome them back with open arms.)

 

[UltimaKilo]

Clearly? Because this "controversy" is mentioned in the comments of every article that makes minor mention of this or that process node. Does anyone EVER reply to these comments with "Hmm, didn't know that, thanks" ?

I did not say the people who post on the MacRumors message board... None of us are a part of MacRumors. Read the article. The poster was not aware, and they work for MacRumors.

 

[svish]

Not surprising. Apple will begin work on future products years earlier than they are announced.

 

[AgeOfSpiracles]

Still Apple uses those terms in communication with end consumers. Or they at least spread it.

Wikipedia has a good article about the 3nm process. It ist shocking how far it is way from having a 3nm measurement anywhere:

3 nm process - Wikipedia

en.wikipedia.org

Quote:
The term "3 nanometer" has no direct relation to any actual physical feature (such as gate length, metal pitch or gate pitch) of the transistors. According to the projections contained in the 2021 update of the International Roadmap for Devices and Systems published by IEEE Standards Association Industry Connection, a "3 nm" node is expected to have a contacted gate pitch of 48 nanometers, and a tightest metal pitch of 24 nanometers.

Sure, Apple uses the terms, because those are the terms that the suppliers use. If they just made up their own terms, that would only be more confusing to the consumers who understand and care about the implications of using TSMC's "2nm" process or whatever node is being talked about. Whether armchair engineers think that the marketing names is specifically correct enough does not change the fact that the names refer to very specific production methods. And it still holds virtually no meaning on the vast majority of consumers, who don't even really know the difference between a cpu core and gpu core... you might as well measure them in mouse turds for all that matters.