tenthousandthings

Not sure how it happened, but a key piece of info got edited out of this summary. The NY Times is the source of the story about China and AI being at issue, while Variety is the source of the story that says the parting is amicable. 

We still have about a month before I will feel comfortable saying this, but it’s really starting to look like iMac will skip M2. This goes against everything I understood about Apple silicon for macOS. To the point where I still don’t really believe it. It’s unbelievable. I just don’t get it.

cpsro said:

ApplePoor said:

There is also an iPhone 15 Pro series CPU processor change mid year which might be a negative for the early adopters.

There is or will be a change to the process used by TSMC to produce the A17 Bionic? What do you know about this?

ApplePoor said:

Was posted on forum several weeks back. Initial ones until inventory is depleted then new "improved" model probably starting early next year, Apple had purchased "all" of the production of the first chip.

I think this is a reference to this rumor from June: Apple iPhone 15 chip manufacturing may shift to a less expensive track (Apple Insider)

But that is unsubstantiated and a typical sort of connect-the-dots rumor that often turns out to be wrong, because Apple's special, disruptive, and wildly-successful relationship with TSMC is still not very well understood, even though we are ten years into it at this point (it started with TSMC 20nm, with iPhone 6). Yes, it is confirmed that TSMC's mainstream 3nm production will shift to N3E, and the second-generation refinement N3P will be built on N3E, not N3 (the name of which has been changed to N3B to differentiate it from N3E). But that does not mean Apple will necessarily be included in that detour. It's entirely possible A17 Pro, M3, and M3 Pro/Max will all be N3, and not N3E. Not to mention the uncertain possibility of a non-Pro A17 next year (basically A17 without the "Pro" GPU, mirroring the now-established approach to M-series silicon).

For this perspective, see Daniel Nenni's comments on the June rumor: TSMC’s 3nm Output Could Reach Up To 100,000 Wafers Monthly By The End of 2023 (SemiWiki main forum post)

What gets lost behind the veil of secrecy is the reason why TSMC is sending everyone but Apple over to N3E. N3 is specific to Apple. As a result, it's not flexible enough for a wider array of customers. There's a deep dive into what is known here, but keep in mind that TSMC's published papers just sort of chart differences, they don't give away trade secrets: TSMC N3, And Challenges Ahead (WikiChip Fuse)

There's a very good chance Apple is happy with N3, and they won't be going over to N3E. The mysterious N3S could also be Apple-specific. Regardless, it will all merge back together with N2 and M4.

Well put. Quality bloviation, thank you.

While it would still need to be accompanied by the commitments and efforts you mention, there is a way to answer the question posed by Ternus. He’s right that there’s no optimal way to integrate AMD’s current PCIe GPU architecture, for example. Plus, I don’t think he only means unified memory. It’s also about TBDR (tile-based deferred rendering) and Apple’s approach to GPU architecture.

https://developer.apple.com/documentation/metal/tailor_your_apps_for_apple_gpus_and_tile-based_deferred_rendering

Nonetheless, it could still be done, by creating PCIe components that use this approach. Imagination Technologies is one company that could do this, not to mention Apple itself. PCIe 5 and Thunderbolt 5 make this practical.

I think Apple has gotten itself into marketing trouble. Those marketing handlers who appear with the Apple engineers when they do interviews should be spending less time worrying about what John Ternus or Anand Shimpi might say and more time developing effective marketing that accurately reflects what those engineers are doing.

The root problem is that M1 and M2 could easily be called A14X and A15X, but they are marketed as being something more than what they are (I'm not suggesting that what they are is problematic in any way, just they they are being marketed as something they are not). It’s an inaccurate portrayal of what Apple is doing. It matters because it disguises the actual transformation from A-series to M-series, which is the jump from the A/M to the M Pro/Max/Ultra.

This problem leads to confusion. The M3 will launch in October as the start of the next generation of Macs, and of course it is, but it’s also not. That transformation won’t actually happen until the M3 Pro/Max is launched.