Apple Silicon Timeline (past and future)
Hi -- I'm creating this discussion thread so we can resurrect it every time Apple releases a new iteration of Apple silicon, updating the timeline and commenting on what we've learned from the new development(s). I've placed it under "Future Apple Hardware" because it's really about the future, even though its starting point is the past. This timeline doesn't follow every Apple-silicon product, just those I've chosen to illustrate the progression and to show how the consumer (= not "pro") devices fit into Apple's roadmap. Once I've introduced a device, I continue to follow it. The products for each generation of silicon are listed in the order they were released/refreshed, but I don't specify the dates. In M-series hardware, I continue to list the product where it was first introduced.
Here is the current timeline, to date. I have observations about what it means for the future, but I will save those for the comments...
Here is the current timeline, to date. I have observations about what it means for the future, but I will save those for the comments...
A4 (March 2010) iPad 1 :: iPhone 4
A5 (March 2011) iPad 2 :: iPhone 4S
A5X (March 2012) iPad 3
A6 (September 2012) iPhone 5
A6X (October 2012) iPad 4
A7 (September 2013) iPhone 5S :: iPad Air 1
A8 (September 2014) iPhone 6 [TSMC 20nm]
A8X (October 2014) iPad Air 2 [TSMC 20nm]
A9 (September 2015) iPhone 6S :: iPad 5 [TSMC 16nm]
A9X (November 2015) iPad Pro 1 [TSMC 16nm]
A10 (September 2016) iPhone 7 :: iPad 6 :: iPad 7 [TSMC 16nm]
A10X (June 2017) iPad Pro 2 [TSMC 10nm]
A11 (September 2017) iPhone 8, iPhone X [TSMC 10nm]
A12 (September 2018) iPhone XS :: iPad Air 3 :: iPad 8 [TSMC 7nm gen1 "N7"]
A12X (October 2018) iPad Pro 3 [TSMC N7]
A12Z (March 2020) iPad Pro 4 :: Mini (Developer Transition Kit) [TSMC N7]
A13 (September 2019) iPhone 11 :: iPad 9 [TSMC 7nm gen2 "N7P"]
A14 (October 2020) iPhone 12 :: iPad Air 4 :: iPad 10 [TSMC 5nm gen1 "N5"]
M1 (November 2020) Mini :: iMac :: iPad Pro 5 :: iPad Air 5 [TSMC N5]
M1 Pro/Max (October 2021) MacBook Pro [TSMC N5]
M1 Ultra (March 2022) Mac Studio (aslo M1 Max) [TSMC N5]
A15 (September 2021) iPhone 13 :: iPhone 14 [TSMC 5nm gen2 "N5P"]
M2 (June 2022) MacBook Air :: iPad Pro 6 :: Mini (also M2 Pro) [TSMC N5P]
M2 Pro/Max (January 2023) MacBook Pro [TSMC N5P]
M2 Ultra (June 2023) Mac Studio (also M2 Max) :: Mac Pro [TSMC N5P]
A16 (September 2022) iPhone 14 Pro [TSMC 5nm gen3 "N4"]
Comments
On my prediction of "Max+" and "Ultra+" -- sure, it could happen with M3. However, as I understand it based on what others have said, a die shrink is hard enough, and wildly expensive, without introducing major changes. So that's why I'm predicting the major GPU step forward in complexity will come with M4, not M3. But I'd be happy to be wrong, and I acknowledge that my forecast here is almost certainly so. Luckily, I'm not burdened by any sort of insider knowledge, let alone scientific knowledge, so I don't have anything to lose. Keep on Truckin' (Blind Boy Fuller, R. Crumb) ...
I'm a bit more optimistic about the M3 timeline as well. Earlier in 2024, perhaps. End of 2023 seems too optimistic... although we do know that Apple has booked to use all of TMSC's 3nm capacity in 2023. So either their capacity isn't too much, or Apple is expecting to sell a lot of new A17s (but the smartphone market has been softer the past few years), or it will be a combination of A17s and M3s.
I could see M3 (on TSMC N3) shipping with the Vision Pro in early 2024, adding to its allure, and the MacBook Air not getting it until June (on N3E instead of N3, per that rumor). I guess the iMac, iPad Pro, and iPad Air could all get it in October, but that still feels like a long shot.
Marvin, in the thread on Malcolm's front-page article about what can be surmised about M3 (as of July 2023), identified some new information: A16 is now thought to be TSMC N4P (5nm fourth generation), and not N4 (5nm third generation) as was thought at the time it was introduced. The underlying source is TechInsights: https://www.techinsights.com/products/dfr-2209-801
That site specializes in "reverse-engineering." I'm not quite sure what that means in this context -- they examine the silicon used in shipping products. Thus, the final line of my initial post should read:
A16 (September 2022) iPhone 14 Pro [TSMC 5nm gen4 "N4P"]
I think I would rate this as "likely, but not confirmed" -- it's information from a single source. But it's not surprising Apple would get production from this process before anyone else, and before anyone on the outside even knew it was ready to produce.
We do know that Imagination (now owned by China, but still a British company) signed a new licensing deal with Apple in 2020, but that might have just been about ongoing or past use of their IP, and not future implementations.
A10 Fusion was the first to have a label, as it was the first A-series SoC to have two types of cores, combining two high-performance cores with two energy-efficiency cores. Thus, "Fusion."
A11 continued this but also introduced the Neural Engine, so the label was changed to "Bionic." This continued until the A17, which introduced the change to "Pro" yesterday. It will be interesting to see what they do when the M3 comes out. I've updated my list to include the A17 Pro and these labels:
Probably it's just so in the future it's clear which devices have that (hardware raytracing) capability.
M3 Max (October 2023) MacBook Pro 16" (also M3 Pro) [TSMC N3B]
Mainly what was blown to smithereens by this move, releasing M3, M3 Pro, and M3 Max all at once, was my sense of the cadence. As I argued then (above), I thought I saw a two-year refresh cycle emerging. This was, obviously, wrong. The MacBook Pro was refreshed after almost 10 months, and the iMac was refreshed after almost 30 months (almost 36 months after the release of M1).
One thing I said, though, still rings true. It's about products. The M2 launch was tied to the (then) new MacBook Air. The M3 launch is about getting the iMac back into the fold (and straightening out the iPadOS lineup, which didn't happen, but they went forward with what they had, which was macOS), yes, but more importantly it's about keeping the MacBook Pro on the cutting edge. Qualcomm has a bunch of high-end comparison slides to the M2 family that were current for, what? One week? I'll bet they thought they'd have months of unchallenged/unverified claims about the Snapdragon X Elite, due "mid 2024." Apple just laughs and says, "We ship products, not benchmarks."
We should expect M4 next year due to:
- TSMC introducing cheaper 3nm technology
- Thunderbolt 5 (at least for Max and Ultra)
- M3 Pro being penny pinched to below M2 Pro level
- Wifi 7 + Bluetooth 5.4 (Broadcom is ready with chipsets for Apple)
- AI: M3 is delivering just 50% of A17. That could indicate a boost to AI in M4, and it might be required for next-gen Siri.
Too bad that Apple stopped offering motherboard swaps.Adding some information I didn't include in my timeline about the A9. You'll see why I'm adding this info when I reply to Nubus's comment...
The A9 was dual-sourced, from both TSMC and Samsung, due to the introduction of FinFET transistors. Apple hedged its bet, sourcing from both. So the A9 line of my timeline should read:
A9 (September 2015) iPhone 6S :: iPad 5 [TSMC 16nm; Samsung 14nm]
Well, definitely Apple and TSMC are capable of maintaining an annual cadence from the standpoint of silicon fabrication -- they've been doing it together for a decade now with the A-series and iOS, with Apple receiving the first high-volume production on TSMC's latest and greatest process node every year but one.
But there may be reasons why the M-series and macOS will be different. An annual cadence is rigid -- it doesn't allow for the kind of flexibility Apple needs to introduce a variety of new macOS devices. Setting aside Apple's long history before M1, I think we've already seen at least two examples of this in the Apple silicon era. First was when they held back the M2 launch, nine months after the A15. M2 waited for the (then) new MacBook Air, which was then given seven months to establish itself in the spotlight before the M2 Pro/Max launched. Second was the iMac remaining on M1, allowing the Mac Studio sixteen months to establish itself, a period that included an M2 refresh for the Mac Studio, while the iMac sat still.
So I'm thinking an 18-month cadence overall (in line with the original rumor), with some flexibility built into that, pending new-product and competition-driven releases. Thus, for example, I think there a good possibility the Mac Studio and Mac Pro will see the M3 Ultra refresh in Q1 2024, but I'm still predicting WWDC 2024.
The latest on TSMC's 2nm Nanosheet development is here: https://www.anandtech.com/show/21091/tsmc-ecosystem-for-2nm-chip-development-is-nearing-completion
Re: 2. Thunderbolt 5 would be nice, but another necessary jump is to PCIe 5 in the Mac Pro. Together, these two advances will allow for a new generation of Pro Display(s), plus a second Studio Display. I think the 27" Retina 5K is here to stay, but a larger Retina 7K would be awesome and require Thunderbolt 5.
Re: 3. I don't know, there are some improvements, as I understand it. But yes, it's about hitting their price points.
This is interesting, and we won't know for some time. Keeping the M3 for 18 months will be difficult. Christmas, end-of-year sales, back-to-school,... there is a reason why iPhone is on a 12 month cadence. Doing 18 months updates and skipping a generation will make it a 3 year update. Introducing upgrades every 12 months will make it easier to skip a processor generation. Perhaps we don't need a new Studio or Mac Pro every year? For 2024 we need to see Wifi 7, DDR5X, and preferably Thunderbolt 5. Apple could name such devices "MacBook Pro M3 rev 2", but processor naming is a lot of marketing as seen on Apple Watch. Better to name them M4. M3 could be moved to N3P... a little bit slower but much cheaper. So many ways forward...
A17/M3 TSMC N3B
A18/M4 TSMC N3E
A19/M5 TSMC N3P/X
And possibly even going to 4 generations if TSMC 2nm is late, and I think it will be.
N3E is a "cheaper" fab than N3B, and it is scheduled to start mass production for 1H 2024. Apple will use this cheaper fab to add more transistors to its A18/M4 generation of chips. So, dies get bigger, more CPU, GPU, NE cores, etc. Waiting on a measurement of M3 Max die size to size if there is room for die size growth there though.
I believe N2 uses the same lithography hardware as N3, right? [I might be wrong.] So this awkward N3 B/E shift won't be repeated in N2. This may be wishful thinking, but I think Apple and TSMC learned some lessons around 2015 (during the FinFET transition) that are being applied now. I read somewhere (TechInsights, I think) that GAAFET is required for going smaller, so planning for this has been in the works for years, and it may have always been in the frame for macOS silicon.
I don't really understand if the "X" process tier could be applied to an M4 Extreme, with everything else on the "P" tier -- I gather it is more for low-volume, focused HPC. N4X didn't fit Apple's timeline, but it looks like N3X might. It's a tantalizing idea, and it's not like an M4 Extreme Mac Pro is going to be high-volume.
Note:
I love the ASML site: https://www.asml.com/en/news/stories/2021/semiconductor-manufacturing-process-steps