Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MacRumors

macrumors bot
Original poster
Apr 12, 2001
68,231
39,037


Next year's iPhone 18 will use TSMC's next-generation 2-nanometer fabrication process in combination with an advanced new packaging method, and the world's leading pure-play foundry has reportedly already established a dedicated production line for Apple in anticipation of mass production in 2026.

Apple-Silicon-Teal-Feature.jpg

According to previous reports, Apple's A20 chip in iPhone 18 models will switch from the previous InFo (Integrated Fan-Out) packaging to WMCM (Wafer-Level Multi-Chip Module) packaging. Technically, the differences between the two packaging methods are quite striking.

InFo allows integration of components, including memory, within the package but focuses more on single-die packaging where memory is typically attached to the main SoC (such as DRAM placed on top or near the CPU and GPU cores). It's optimized for reducing the size and improving the performance of individual chips.

WMCM, on the other hand, excels at integrating multiple chips within the same package (hence the "Multi-Chip Module" part). This method allows more complex systems, such as CPUs, GPUs, DRAM, and other custom accelerators (e.g., AI/ML chips) to be tightly integrated in one package. It provides greater flexibility in arranging different types of chips, stacking them vertically or placing them side by side, while also optimizing communication between them.

TSMC plans to start manufacturing 2nm chips in late 2025, and Apple is expected to be the first company to receive chips built on the new process. TSMC generally builds new fabs when it needs to increase production capacity to handle significant orders for chips, and TSMC is expanding in a major way for 2nm technology.

To serve its major client Apple, TSMC has established a dedicated production line at its Chiayi P1 fab, where WMCM packaging monthly capacity is expected to reach 10,000 units by 2026, reports DigiTimes. According to Apple analyst Ming-Chi Kuo, only "Pro" models in the iPhone 18 series are likely to use TSMC's next-generation 2nm processor technology because of cost concerns. Kuo also believes that the iPhone 18 Pro will feature 12GB of RAM as a result of the new packaging method.

Terms like "3nm" and "2nm" describe generations of chip manufacturing technology, each with its own set of design rules and architecture. As these numbers decrease, they generally indicate smaller transistor sizes. Smaller transistors allow more to be packed onto a single chip, typically resulting in increased processing speed and improved power efficiency.

Last year's iPhone 16 series is based on an A18 chip design built using a second-generation "N3E" 3nm process. Meanwhile, this year's upcoming iPhone 17 lineup is expected to use A19 chip technology that's likely built on an upgraded 3-nanometer process called "N3P." Compared to earlier versions of 3nm chips, the N3P chips offer increased performance efficiency and increased transistor density.

Article Link: iPhone 18's Advanced A20 Chip Packaging Gains Momentum at TSMC
 
Last edited:
More than being an advantage for the A20 chip, if WMCM is utilized for the M6 series in 2026 this could potentially be a gamechanger for Apple's desktop chips, as it provides an alternative to creating larger and more powerful workstation-grade chips than Apple's current approach of stitching two Max dies together. It would also allow Apple to do things with the regular chips that are impractical to do with monolithic dies, such as offer beefier GPUs or more memory.
 
More than being an advantage for the A20 chip, if WMCM is utilized for the M6 series in 2026 this could potentially be a gamechanger for Apple's desktop chips, as it provides an alternative to creating larger and more powerful workstation-grade chips than Apple's current approach of stitching two Max dies together. It would also allow Apple to do things with the regular chips that are impractical to do with monolithic dies, such as offer beefier GPUs or more memory.

It will be interesting to see the impact of using WMCM packaging across the A as well as M series of chips. While lower power consumption leading to longer battery life is nice, I suspect the battery life is more than adequate now for most users. If Apple does decide to use the tech to build workstation-grade chips, it will, as you suggest, be game changing, if the incorporate more powerful chips across teh entire M powered iPad and MacBook line. Keeping current battery life with much more powerful chips would be quite a feat. Besides the obvious advantages for gaming and power users; Apple could make AI actually useful the average user while maintaining privacy by doing everything on machine in terms of personal data. A spell checker could adapt to your writing style, such as if you use multiple languages; tech support for Apple supplied programs could be not only more robust but actually perform tasks beyond Siri. Spam filtering could be more robust as it gains info on typical spam messages or emails, especially ones that use very similar text but different phone numbers or email addresses.
 
Last edited:
SoC is a failure for M series and now, they are using MCM at last.
This is about as bizarre of a comment as stating that Apple should have stayed with Intel. When my dreams turn into nightmares, I still hear the Intel MacBook Pro fans blowing at full speed every time I launch an app (yes, I have weird dreams).

Engineering is all about trade-offs...much like design, policy, and corporate strategy.

For all its faults...Apple has been killing it with its M chip strategy.

Do I wish there was an M4 Super-Extreme Ultimate Octo-Starkiller Edition? Sure. But the lack of one doesn't mean that their chip strategy hasn't been superb, far exceeding the needs of 95+% of its users.

Now, it seems that Apple can finally make its Super-Extreme Ultimate Octo-Starkiller Edition for the M6. I don't need one, but I might buy one just for the giggles and spits.
 
TSMC plans to start manufacturing 2nm chips in late 2025, and Apple is expected to be the first company to receive chips built on the new process. TSMC generally builds new fabs when it needs to increase production capacity to handle significant orders for chips, and TSMC is expanding in a major way for 2nm technology.

Expected by who? Those not paying attention?

"... next-generation AMD EPYC™ processor, codenamed “Venice,” is the first HPC product in the industry to be taped out and brought up on the TSMC advanced 2nm (N2) process technology. ..."

1347696.jpg


Literally, AMD standing there with a finished development N2 wafer in their hand. In TSMC's terminology Apple's A/M series is 'HPC' class product.

Is Apple far behind? Probably not. Decent chance they have in initial 'development' wafer by now. However, TSMC isn't exclusively doing things only for Apple. Throw enough moony up on the table to get an early wafer, TSMC doesn't discriminate. AMD isn't the 'broke and barely keeping the lights on' company they were 10 years ago anymore. Nvidia is flush with cash also. There are multiple players in a hurry and don't have "fixed in stone" , "every September" dog and pony shows.

Epyc has a longer (more complicated) product qualification cycle than A-series sized die, so who ships at relative max volume first may be tighter. AMD may have enough lead time here to ship product to retail end users first though. (certainly if count shipping to system developers for test builds which go to customers. )

AMD recently finished talking about Venice Epyc in June 2025 and put a '2026' delivery date on them.

There is a decent chance they are shooting for a May-July rollout.
 
  • Like
Reactions: Annihilator_47
This is about as bizarre of a comment as stating that Apple should have stayed with Intel. When my dreams turn into nightmares, I still hear the Intel MacBook Pro fans blowing at full speed every time I launch an app (yes, I have weird dreams).

Engineering is all about trade-offs...much like design, policy, and corporate strategy.

For all its faults...Apple has been killing it with its M chip strategy.

Do I wish there was an M4 Super-Extreme Ultimate Octo-Starkiller Edition? Sure. But the lack of one doesn't mean that their chip strategy hasn't been superb, far exceeding the needs of 95+% of its users.

Now, it seems that Apple can finally make its Super-Extreme Ultimate Octo-Starkiller Edition for the M6. I don't need one, but I might buy one just for the giggles and spits.
Then how come Apple is NOT able to make Mac Pro grace chip? It only proves my point and if you have a problem with that, prove me wrong.
 
Last edited:
SoC is a failure for M series and now, they are using MCM at last.

The primary additional component that will be added to the package is probably the modem. This does not herald any sort of disaggregation of either the A or M SoC.
 
Last edited by a moderator:
  • Like
  • Disagree
Reactions: NetMage and BNBMS
‘monthly capacity is expected to reach 10,000 units by 2026’

That seems like at least a factor of 10-100 too low. If that was daily then it could maybe support an iPhone release.
 
Wafer is the operative word here. Wafer scale processes allow for wafer to wafer bonding with microbumps (and usually TSVs). It should allow for large data (memory) interfaces that focus on minimizing joules per bit. It will certainly help with memory intensive tasks common in AI applications (or games).
 
The primary additional component that will be added to the package is probably the modem. This does not herald any sort of disaggregation of either the A or M SoC.
The die size is too big for its performance. Do you not understand? For that size, it supposed to be way better than RTX 5090 series but it never did cause its SoC's inefficient design. Guess what? Even Mac Studio is not even close to RTX X090 grade performance despite their advertisement.
 
  • Haha
Reactions: NetMage
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.