They’re going to show the current chassis during the unveiling, take the cover off, and pull out the new Mac Pro from inside.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Silicon Mac Pro Configurations: Everything We Know
- Thread starter MacRumors
- Start date
- Sort by reaction score
Writers don’t need to waste their time pre-fixing every title with "What we guess...". This is a damn rumors site. Everything here should be treated as a piece of fiction read purely for your own amusement. It’s obvious to everyone.
Last edited:
The old cheese grater and trashcan started at $2500-$3000 respectively and were just about affordable if you just wanted something with more oomph than an iMac, but didn't work in a corner office at DreamPixney. The Mac Studio is the new (& rather better) trashcan.
The 2019 Mac starts at $6000 for a worse spec than a top-end Intel iMac, and only begins to make sense as part of a $12k+ system that takes advantage of all those RAM and PCIe slots. Not saying that it's not what some people need, but it is not an impulse buy and not something you can assess by playing for 20 minutes on whatever configuration your local store happened to have on display aided by the local Genius' encyclopaedic knowledge of scientific modelling and professional HDR colour grading.
I assume that Apple have some sort of "serious callers only" sales team for when the DreamPixney-corner-office guy calls.
...again, DreamPixney probably got advance information via NDA. Apple don't want to see no stinkin' HP logo on the end of Despicable Toys 16...
More seriously, though, the primary market for an ASi Mac Pro will mainly be the folks who kitted out labs with 2019 Mac Pros with 1.5TB RAMs and quad high-end GPUs. I'm betting they won't yet have ripped out all those machines and replaced them with 128GB Studio Ultras & just hoped they could cope.
If Apple haven't been reaching out to key Mac Pro customers and giving them some roadmap information than isn't publicly available then the problem won't be customers who have bought Studios, it will be customers who have finally given up and switched to Windows or Linux after the third successive time that Apple effectively depreciated the Mac Pro without giving any clue as to its sucessor.
By giving the M2 Max an interconnect on both ends? By having an interconnect module that joins 4 M2s together, 2 to each side? The "4 chip" idea has been around since the "Jade 4C" rumours started, so they presumably have a plan.
Yes. The big questions (which this Fine Article doesn't really answer) are:
(a) Will it have PCIe slots? Bear in mind that the M1 Ultra has either 6 or 8 TB4 controllers (we know the Studio Ultra has 6, but since the Studio Max already has 4... ) so the extreme could have up to 16... so there's plenty of PCIe bandwidth there to use.
One possibility that occurs to me - which would be disappointing but kinda sensible - is if the "Mac Pro" were 'simply' a Studio Ultra (maybe bumped to M2 Ultra) in a 1U rackmount form, that could be racked up alongside Thunderbolt-to-PCIe enclosures and storage modules.
(b) Will it support PCIe GPUs? Or are Apple confident that their M2 GPUs will cut the mustard once the software gets optimised?
(c) Are people who currently actually need 512GB-1.5TB really going to be satisfied with - let's say 256GB of RAM - and somewhat faster swap?
Problem is, if you throw out on-die GPU and on-package 'unified' RAM then you're losing some of the killer features of Apple Silicon.
Apple Silicon Mac Pro Configurations: Everything We Think You Think We Know
It depends what you want to do. If you "need" to run the latest OS, then your machine's lifetime will be limited. But as a piece of hardware, most Apple machines will run for a very, very long time. For example, I'm 99% certain there are still audio professionals who run their DAW on old "cheese grater" Mac Pro towers running Mac OS X 10.6.8, with external DSP gear for the heavy lifting. Given that my current 16" M1 Pro MBP runs extremely cool, basically all the time, it too will run for a very long time. So, as a production machine I could run this for 10 years, easily, if I settled on whatever software still ran on the OS at that time. Being a developer and general geek I'm very likely to trade it in on the M2 version, but that's just a choice, not a necessity.Every single system I have disagrees with this so much.
Very well could be. My impression was that it is a hardware limitation (in addition to lack of software optimization), but that is just from random forum posts/tweets.
Other than the GPU scaling issue though, I can’t think of any reason why RobbieTT would say the M1 Ultra is fundamentally broken.
On what workloads? According to what benchmarks? Does it scale better or worse than other GPUs?
Basically nothing ever scales linearly with GPU core count. There's always a bottleneck somewhere. An Nvidia card with twice the cores never gives you twice the framerate.
You're going to have to be more specific with your complaint.
I have no complaints.
I was just trying to be helpful and share a common thing I hear from users and see in reviews, since this is something you asked about.
at least have the storage slots be user upgrades like the mac pro is now maybe have
2-4 disks or even an 8-16 stick mega raid 0 build in up to 32TB storage.
Thunderbolt-to-PCIe adds over head and can the video parts of the CPU push out that many links??
Will apple put in data/pci-e only TB ports?
will apple make poor choices like put in 4 10GB nics? over maybe only 2 and some other port with the pci-e io?
Apple can try something like SFP+ ports and then make it only work with apple plug ins.
Are you, though?
You're losing some of Apple's key PR campaign points, anyway.
For CPU-heavy tasks, it doesn't really matter what kind of memory you have. Might as well have a bunch of memory in slots, for pro users who want to upgrade.
For GPU-heavy tasks, there are advantages to the unified memory model, but the advantages aren't really that substantial when an M1 Ultra is only barely competitive with an RTX 3070 on most benchmarks. That's a discrete card you can buy for less than $600 online these days.
The one thing we do know, is that it will be hideously expensive.
Remember the days you could get in to a 'pro entry level machine for $2,999?
Remember the days you could get in to a 'pro entry level machine for $2,999?
Shoes and clothes that last a lifetime?! You haven't met my wife...
Electronics are evolving, that is the exciting era we live in
Man that's going back a fair way!
I can't see the "entry level" of a new AS Pro going for anything under about $8,000 or thereabouts.
You have to remember that the price of a Mac Plus, taking into account inflation, would be about $125,000 in today's money. So $8,000 is a steal for an AS Pro.Man that's going back a fair way!
I can't see the "entry level" of a new AS Pro going for anything under about $8,000 or thereabouts.
I realize you were joking, but:
Mac Plus was released in January 1986 for $2,600, the equivalent of $7,028.45 today.
Mac IIfx 4/160 was released April 1992 for $10,969, which is $23,296.43 today.
First Lisa was $9,995 in January 1983 -- that's $30,278.92 today.
Current Mac Pro (Intel) can be configured to cost $53,799. That's probably the most expensive Mac ever.
That's true. But, the average cost of a machine these days is significantly lower due to mass manufacture, global supply chains (ha!) and robust competition. I think a comparison with a circa 2008 Xeon Pro is the more telling comparison as it really highlights the massive increase in a 'relatively' short time period.
I've always bought Mac pros since the G3 days but I think they have become very much a niche within a niche so only accessible to businesses that will write them off against tax. For artists, musicians et al I suppose the Studio is now our option. Not a particularly cheap option though..
When I look at the list of things Apple ever supported as PCIe, the only ones that make sense anymore would be that I/O card (for those who want more T-bolt ports). Networking? Audio? Storage? While handy internally, as these machines don’t generally move around much, that can be external.
When the slots were for that PLUS GPU’s and acceleration cards, it was a fairly general purpose thing that everyone could use one or two of. In today’s world where acceleration is on the chip and so’s the GPU, I just wonder how many products would really utilize PCIe. Maybe where there currently aren’t any external options, but if Apple gives some of their partners a heads up, they could be well along the way of having an external Thunderbolt option available.
These are not meant for the consumer market, necessarily. A mini may be fine the average cat, but the pro has always been intended for pro shops and the like.
Not if the user needs to run Final Cut Pro… or Logic Pro.
It really doesn’t matter what’s happening on the PC side anymore unless the user is primarily NOT interested in macOS, BUT instead is interested in a contest solely consisting of ejecting waste water some distance and then seeing how it compares to others doing the same.
Ok, maybe art, if you really look after it. Maybe mech watches if they are extremely high quality and you look after and service them. Houses, well, only with constant expensive maintenance and renovation.
Shoes? Really? A lifetime! Clothes??? I get far less out of my clothes and shoes than I do my MBPs.
If you had the same access to parts and servicing for your electronics as you do for your mech watches, and houses, then you'd get a lifetime out of them too. You might however, simply find that the newer tech leaves the old stuff seemingly worthless, but it would still chug on, doing what it was originally designed for, just fine. I've found myself with old computers that I've left lying around, that still work, but are so damn old and out of date that I couldn't even donate them to someone. I've still got perfectly functional 1GB and 2GB RAM SODIMMs in my drawer, but can't think of anyone that I could give them to that could make use of them.
I don’t even expect it to double the power of the Mac Studio. The Mac Studio, after all, IS currently the fastest Mac Apple makes and faster than the Macs that came before it. Even if it’s only 20% faster, it’ll be the fastest Mac yet and, for those who want/need the fastest Mac, that’s what they’ll get.
In my mind, the differentiators will be related to RAM, storage, physical port options and other things above/beyond just CPU/GPU performance (maybe more ProRes encoders/decoders, stuff like that). I’m not thinking “how could this beat a Mac Studio”. I’m thinking more like, “Who, specifically, are the very few that need something that the Mac Studio doesn’t offer as options… and how many of that small group are not going to like what Apple presents?” I truly expect that some users that are waiting to see what it is (and have plans to buy it) will not like what they see because it drops some old “Mac Pro” expectation and they may drop macOS orrr… just use their Intel box until it dies. And I believe Apple’s factored in this loss of what can’t be more than a few thousand at this point.
They also upped the max GPU cores from 8 to 10, so a similar increase would see the Ultra with up to 80 GPU cores. Maybe they will, maybe they won't.
Also, the M2 might well have gained an 18% faster CPU, but they used a 9% increase in clock frequency to help that jump, which equates to a bigger power drain, and more heat produced. Are they going to do that across the board with the M2 line? It's a move of desperation.
Similarly, the 35% more powerful GPU was largely due to the 25% increase in GPU cores, and an unknown (as far as I can find) increase in clock frequency for the GPU cores.
The problems with increasing computing power with extra cores are: there are diminishing returns in efficiency with routing everything to the different cores; and you also need software that can efficiently split it's tasks into more and more threads, which suits some domains, such as the multidimensional arrays of graphics, but not so much the linear computations of many other tasks. Thus why the GPUs of the M1 line range up to 64 cores, but the CPUs mostly on only 8 or 10 cores, with only the Ultra with 20 cores. It would take quite a specific piece of software to make use of 20 CPU cores!
