My guess is that the Mac Pro will be the last model to be replaced with Apple Silicon - that much has been predictable since the initial announcement since:
it is also what they did the "last time" (PPC -> x86) . ( and the time before that if I recall correctly 68K -> PPC)
However, question is whether it makes economic sense for Apple to make a "Xeon-killer" monolithic Apple Silicon CPU with dozens of cores that would probably have to concede the advantages of integrated/on-package GPU and RAM in order to offer the same sort of specs as the Mac Pro...
That would be tossing away the ability to natively run iPhone apps at best speed. That probably is not an option. There are software constraints here also. One of which is that Apple is dragging the iOS app store space onto Macs.
Apple tossing that leverage away just to cover the likely less than 1% space of the Mac pro is highly doubtful.
It would seem to make more sense to do something more radical that played to the advantages of Apple Silicon: maybe (as people have already suggested) multiple M1Xs with RAM and GPU power distributed between them. Trouble is, that's going to depend very much on properly optimised software, so it might make sense to keep the Intel Mac Pro on the books indefinitely to cater for people committed to a more traditional workflow, while releasing something more "courageous" in parallel.
This isn't really true if use modern 3d packaging and some slight augments to the baseline "M1X" die would be using to scale up.
The upcoming Xeon SP "Sapphire Rapids " packaging is illustrative.
Intel unveils its biggest architectural shifts in a generation for CPUs, GPUs and IPUs to satisfy the crushing demand for more compute performance.
www.intel.com
four dies on the same package with extremely short (distance) , narrow (small bumps) , silicon to silicon connections all with shared memory ( every core can make a request to each set of attached memory). The NUMA (non uniform memory access ) impacts aren't significantly different that those present in the Xeon 3200 28 (max) core configuration (internally that too has more than one internal bus and a memory request to the farthest away internal memory controller needs to two hops. )
Each one of these dies have a memory controller component and a PCI-e provisioning unit.
There is little "custom to Xeon 3200" software on the current Mac Pro. There could easily be little overhead for dies which are designed from scratch to be combined in a multiple die package. Four "M1X" dies that were design to be 100% independent... yeah that would be problematical, but Apple doesn't have to do it that way. it is actually probably overall cheaper to design them to be combined in 1 , 2 , 4 die packages. Once have that working they would have covered more Mac products with one single design ( so they can amortize out the costs over a much larger pool of systems then if did low volume specific dies. ). If just look at the augments as increasing the costs for the MBP 14 and 16" system then miss the point. The additional iMac 27 and "half sized " Mac Pro ( somewhat iMac Pro replacement) probably enough additional revenue to absorb and offset those costs.
A minor deviation that the Sapphire Ridge approach used above would be to split off the common baseline I/O systems ( the security processor , media engine , SSD controller ) into a four die ( perhaps stacked under 3D style) and still have the CPU/GPU and memory controllers spread out on slightly smaller dies with less "fused off " elements at higher multiples. )
There are Sapphire Rapids package variants with HBM memory attached to the overall package ( I suspect a larger socket size as likely not cooler compatible either. But 4 stacks on package ). Apple probably has four LPDDR4
per die. That will end up an even larger package ( although Apple's compute dies are likely a bit smaller than Intels here ) . It can be done... just a quite large overall package. ( which the current Mac Pro CPU thermal zone section could support with minor adjustments. "might loose some 'empty space' where optionally sticking 3.5" drives into. )
Where probably come up "short" is that the PCI-e provision complex on each die is likely not as much as Sapphire Rapids as the baseline die is targeted to a laptop with no dGPU present at all. So if Apple put one x8 PCI-e v3 ( or v4) on each die then would end up with less bandwidth to provision to slots . ( So getting rid of several slots wouldn't be surprising.... hence "half sized" Mac Pro. )
Similarly the 10 core laptop usage probably isn't gong to use ECC Memory. So decent chance the "half sized" solution also looses that in the sauce also ( along with capped top end RAM capacity ( sub 1TB and likely lower ).
Good question - and/or do they actually understand why people want them?
Apple understands that they are not building everything for everyone. Just because they don't build it doesn't mean they don't understand it. Apple buys gobs of servers for their web services . They definitely know why people buy them because hey have very similar problems. That is quite different from feeling compelled to make them as a product for "everybody".
After the MP 2013 there was threads with hundreds of comments for 6 years about how Apple "had to" do a Mac Pro update that de-emphasized Thunderbolt because that got in the way of buying the latest generic GPU cards off the shelf and tossing them into the Mac Pro "add-in card container". ( it is primary a card container). Apple "knows" those folks are out there. They are not necessarily going to kowtow to them. MP 2019 arrived with a seamless solution that combined in Thunderbolt. Apple isn't interested in completely detaching the Mac Pro from the rest of Mac product line up on certain aspects. It isn't about "know" . It is about a coherent product line up on certain aspects.
Also, they might see value in having a "flagship" high-end workstation in the same way that (say) VW sees value in making Bugattis. If it's helping keep that Apple logo on the end credits of major movies & TV shows then, maybe.
Again highly doubtful. Apple has explicitly said they aren't out to build everything to everybody. VW adding brandsto cover a wider set of markets is a "sell everything to everybody" kind of strategy. Selling Bugattis really doesn't help them sell more Golfs to highly limited budget buyers.
that whole lego sells really doesn't work as well has the concept is hyped in these forums. ( or for any of these car companies used as example. Chrysler has had lots of halo cars and has been given sweetheart loans to keep them afloat multiple times. ) .
The iPhone has failed to serve as a gateway product to the Mac, Apple TV, and HomePod for over 50 percent of iPhone users, while the iPad, Apple...
www.macrumors.com
AppleTV is a 'dog' compared to Disney Plus in terms of subscribers ( so is HBO ) . If Apple wasn't handing out massive 'freebies' subscriptions it would be in worse shape. the content of the service is a far better seller for the service than the logo on top. Similar what the product does matters more than the logo.
3. Trashcan then get's zero updates (apart from rotating out the entry-level model) between 2013 and 2019 and was probably going to be replaced by the iMac Pro in 2017 until that was rejected by key customers/partners resulting in that unprecedented press conference where Apple announced the (then) mythical modular Mac Pro.
4. iMac Pro gets launched in 2017, is never significantly updated (apart from a new GPU option?) and dropped in 2021... by which time the low end models were out-performed by the top-end iMac but if anybody had committed to the all-in-one format and needed ECC ram or the better cooling they were out-of-luck.
That press conference introduced the concept of the iMac Pro as much as the much later Mac Pro update. I think some folks overblame Intel for slow updates but the iMac Pro and the M-series transition is one of those where Intel's and AMD's stumbles on timelines and/or thermals probably caused problems for the iMac Pro's "one last update".
if W-3300 had come in 2020 (like it was suppose to) and at thermals similar to the W-2100 at similar core counts Apple could have used it. Instead, the 2100 got a price cut and labeled the w-2200 with almost zero uplift at constant thermals. Similarly Navi was late ( slid out 8 months and into much later 2019 ) which has a knock on effect of sliding out big Navi.
The additional "problem" was that Intel's mainstream processors were released from the 4 core limit. So the 'regular' iMac 27" highest BTO configurations encroached on the iMac Pro space. Then in late 2020 the M1 Mini also encroached on some old iMac Pro workloads also. That is even before the 'M1X' Mini arrives (and 20-40 core iMacs arrive later).
Throw on top Apple's limited ability to walk and chew gum at the same time. The iMac 27" update work was pulled to get the iMac 24" out the door. ( in fairness the pandemic was likley a contributor there too with the physical design labs closed for a while. )
Once the iMac Pro missed the 2nd Half 2020 update window , it was a much "squeezed out" as Apple was delinquent.
5. New Mac Pro gets announced in 2019, ships in early 2020 (along with a 2x price hike c.f. previous Mac Pros that simply threw lower-end Mac pro users, who just wanted some PCIe slots and internal storage, under a bus) and... six months later Apple announces that they're moving everything to Apple Silicon by ~2022 while not breathing a word, publicly, about any sort of roadmap for how they will cater for Mac Pro users (they haven't even clearly said that they won't support discrete/PCIe GPUs - that's just a, not unreasonable, extrapolation from comments in developer docs).
By 2017 Apple knew there were looking at end of 2020 ballpark for Apple silicon. Apple doesn't talk about future products are part of an explicit corporate policy. This isn't some "mac pro" thing.
Apple "leaked" through folks like Gurman that they were moving to their own chips long before 2020.
2-3 year roadmaps with an official Apple logo in the corner of the Keynote (PowerPoint) slide. ... when has that happened since Jobs returned? That just isn't what Apple does. Folks can moan and groan that it is missing, but they have said in official corporate statements that they just don't work that way.
Apart from being licensed to run MacOS, it really is just an Intel tower that mainly just took advantage of being an early adopter of the latest Xeon-W chip to make sure it won a game of Top Trumps when it came to PCIe slots and max. RAM.
early adopter?
Launched Q2 2019.
https://ark.intel.com/content/www/u...-xeon-w3265-processor-33m-cache-2-70-ghz.html
Apple shipped in mid-late Q4 2019.
An early adopter would have been Q3 2019. ( or Q2 but I don't think there was any substantive volume then. )
...on the bright side, Apple Silicon might be Apple's chance to offer something a bit more distinctive that doesn't invite like-for-like comparisons with commodity workstation/server hardware.
Apple doesn't make commodity "container" boxes. There are folks who attach Apple to the corporation the existed around the time Steve Wozinak worked there full time. That isn't Apple. Hasn't been since Jobs came back.
Even the MP 2019 isn't primarily a commodity container. MPX and SSD modules that don't fit in anything else.
