In real render apps even no OC version i9-13900F easily outperform top M1 Ultra 20 cores.
When You add high quality liquid cooling and case You get very silent computer.
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 Reportedly Not Coming at WWDC, Mac Studio Refresh Likely Delayed Until M3
- Thread starter MacRumors
- Start date
- Sort by reaction score
When You add high quality liquid cooling and case You get very silent computer.
Until we see what Apple will utilize commenting that it Isn’t equal to the task is guessing. i imagine this far along the design just works as it should and it’s more to there being no issues with it running some version of MacOS without issues.
?? It's still a Mac Pro with Mac OS, only it still has expandability. That sounds like a good thing for someone that needs performance, and isn't a knock on Apple or Macs at all.
Power consumption isn't as relevant as performance for people that make their money with their computers, and it's still MacOS.
No, Apple isn't thinking this... Mark Gurman is the one thinking this. It's beyond stupid and yet people fall for his nonsense all the same. The Mac Studio has barely been out one year, so we'll probably get an update before we see a Mac Pro.
In the last decade, probably not. At least not in any serious way. Maybe someone threw that out there as a 'relive the past' thought experiment , but Apple is not structured that way strategy or tactically. The 'Clone Wars' are over.... been over for decades.
First, Apple doesn't directly charge for MacOS anymore. So any construct that involves going back to charging for it decoupled from Apple hardware has gigantic problems. The last "OS" that Apple charged for was macOS Server. That has been dead for years at this point as a separate OS. Even sputtered out as a set of extra apps and settings on the baseline macOS (let alone a 'forked' operating system).
Same thing for upgrades. Apple charges for several upgrades upfront with the price of the systems. If everyone buys the upgrades with no 'sales process' overhead , then they are cheaper to build and distribute. ( have a well know pool of money to get the upgrades out). Also happens to give Apple a solid revenue stream can recognize on a regular basis.
The licensing is solely done as a system. The hardware 'half' always comes inseparably tied to the software 'half'. Users buy the whole. Apple doesn't sell detached copies. That opens a door where they loose control over virtualization and several other modern factors. It probably isn't worth it ( the losses from revenues lost don't offset the revenues gained. The hypermodular crowd is likely less than a several 100K users versus main Mac base of 10's of millions. ). Even more so when have the money where the 10's of millions of systems per year are not doing incremental charges.
Second, how many folks are going to pay. If Apple charges $400-600 for his "macOS Pro" , then are most folks going to pay when the modern 'price' anchor for an operating system charge ( regular Windows or macOS or Ubuntu Linux ) is $0.0 ? Many folks who grumble that "Windows is so horrible" will change their tune if have to pay hundreds more for macPro.
Economies of scale matter. Just to put some concrete numbers to illustrate. If the nominal platform port costs are $50M/yr then if dealing with 20-25M uniform Mac sales per year that amortized cost is $2.50 - 2.00 per unit. In contrast if now flip to just having 200K units/yr that amortized cost is now $250 per unit. Burying $3 into Mac total system costs are pretty easy in Apple's pricing strategy. ( generally round up to x99 or xx99. bascially to the closest higher hundred more). Trying to 'hide' $250 is much harder. Still haven't covered burying upgrade costs ( if don't put the macOS Pro on a subscription plan ... which would likely not increase the 'fan base' either. )
Third, the third party application developers are not likely to spend more money long term chasing a niche of a niche. There are apps that several vendors don't even port to macOS as a whole when it was all on Intel. If divide the macOS market into even smaller fragments is Apple going to attract even more long term developers? Probably not. Windows with a 80+% market share possible could balkanize 70%-30% or 65%-35% and not collapse ( the 35 , 30 or even 20 or 15% of a Windows on Arm would likely all still be larger than the macOS entire share). But macOS already at 10% going to 9.1%-0.9% is pretty close to just being single 9%. The rest of the mac product line up is the software sales 'engine' is going. Most app developers are going to follow the money ( even if Apple engaged in some "rob Peter to pay Paul" to keep some "macOS Pro" thing around as a loss leader. Other partners are not going to sustain losses at the same rate that Apple might ).
Fourth, macOS doesn't have the max thread count to match the modern higher end workstation core/thread counts. macOS is capped at 64 threads. While conceptually the x86_64 version could be forked into something that does support a higher number. That would just add more to the cost issues outlined above ( diverge the code bases and also diverge shared costs.). So Apple would have to find some x86_64 vendor who would be willing to construct a 'Goldilocks zone' workstation that didn't overlap with rest of mac line up, but didn't represent too much more work for Apple. This is exactly what failed in the last century 'Clone wars'. Apple wanted clone vendors to only do the stuff they didn't want to do and clone vendor inevitability wanted to sell "more" systems and make moves to encroach on Apple's slice. It is already an overall PC market niche so there isn't much room for multiple players. Too many cooks in too small a kitchen is an unstable situation.
The objective is not to create some "Windows killer" macOS. It isn't put macOS on as many systems as possible regardless of profitability. Apple has openly and explicitly stated they are not trying to build everything for everybody.
Additionally, a major factor here is not about legacy PC form factors. Higher synergies with iOS and iPadOS apps ( e.g. learning to optimize apps on Apple GPU structure and Unified Memory) with macOS apps helps all three. Fragmenting mac apps on other , non portable optimizations doesn't enhance the other parts of the portfolio. I don't think Apple will drop macOS on Intel in next 2-3 years , but where does Apple want to be in 7-10 years. Where is the puck going? Even in the higher end HPC space hyper-modularization isn't a panacea going forward. (e.g., HBM RAM isn't modular. It is soldered on. )
At this point, after waiting so long for a "pro" desktop, I kinda agree. Seeing how AI whether we want it or not is becoming a part of creative process, its hard to imagine a "pro" machine that doesn't meet the standards needed by AI applications - for instance some of them need a lot of graphic card memory (I'm not speaking strictly about image generation but AI in general - it uses graphic cards for its computations). Some of the new AI tech NEEDS nvidia chip - they don't run on anything else. And let's not forget how all of market is developing - nvidia is more and more connected to AI and declared their commitment to keep this route for foreseeable future. Soon pro apps will have a huge support of AI stuff like generative image up-res'ing, denoising and all that midjourney bs (which I hate
). So to sum up - if apple doesn't figure out a way to have a PCI port for a peripheral graphic card in their Mac Pro and make their M-whatever chips work efficiently with external graphic cards, it's no longer a Pro machine just because of this. It's already obsolete. From a perspective of a "Pro" machine, this was a stupid, narrow minded choice on their behalf - making an all-in-one chip, with dedicated graphic cores, completely ignoring what nvidia is doing. It's apparent they totally missed the AI revolution that is unfolding right before us - just like few other compenies did, including Alphabet (Google). THAT is the visionary aspect of a CEO that Tim Cook lacks. He's good with numbers, but Jobs would have seen it coming and would direct Apple development accordingly.Considering the path they chose, maybe it's just better to leave the pro market than be a subject of ridicule. It's a niche market anyway. M chip machines are awesome for your everyday applications, including video editing. But it's not a future-proof machine and that's what one should expect from a "pro", expensive workstation. They committed to a certain architecture of M chips which is not suitable for cutting edge and future applications, unless they find a way to open the M chip ecosystem (I'm not a chip architect, I have no idea if that's even doable).
I wonder why noone is talking about this aspect of pro machines - maybe because youtube apple rumor bloggers don't know sh*t and just copy/paste stuff they read at twitter regarding next rumors, creating moronic content that contests their videos from just a few days earlier...
AI-powered processes in all our Creative Suites, color postproduction software, music production, video production, 3d / game dev is the future for a next decade. You NEED swappable graphic cards because thats their resource. Apple has to either pivot and create some open platform with a custom M chip or just abandon the true "pro" market and focus on nice web browsing laptops - which they have been doing for past several years.
Linux doesn't have the software I use. And it's at this point I am so fed up with Windows. I wish macOS is in a better state with more software and more support. But if there is no competitor to i9/4090s, this will never happen.
M2 Max has 96GB of RAM. Two of those (you know what makes up the Ultra) would give 192GB of RAM. That is a significant upgrade and would make me jump on a new Mac Studio ASAP.
...etc.
Apple Silicon SoCs have been a great success in the laptops and tablets which are Apple's bread and butter. ASi has been pretty successful in the small-form-factor desktop range, too - and it does this with only two basic laptop friendly CPU die designs - the regular Mx and the Mx Max: the Ultra being two linked Maxs, the Pro being a cut-down Max). For Apple, the benefits of that probably outweigh the risk that the Mac Pro will be a casualty.
Unless the M2/M3 Ultra die has some hidden depths - e.g. the ability to support significantly more than 128GB of regular, expandable DDR5 RAM and enough PCIe bandwidth to run workstation-class GPUs - producing something equivalent to the 2019 Mac Pro is going to require designing a custom CPU die just for the Mac Pro. That would be massively expensive with such low volumes.
...and to what end? An ARM-powered Mac Pro's performance on rendering, ML training etc. would be capped by whatever AMDs latest GPU offerings (also available in cheap commodity PCs) could do and would throw away the relative benefits of unified RAM. Apple Silicon/ARM's big party trick - unbeatable performance-per-Watt - is at its least relevant in a single user tower workstation - it gets relevant again in high-density computing and cloud servers, but but whereas the M-series is currently the only laptop/desktop-class ARM processor worth a mention, Ampere, Amazon etc. all have the jump on Apple when it comes to server-class ARM chips.
What does Apple (strategically) need a 2019-like Mac Pro for?
- Halo product? Apple already have a halo product - the iPhone pretty much codified the term 'halo product' - and the Apple "brand" is as established as it is going to get.
- Maintain support for 3rd party "pro" applications? Not much use to MacBook Pro owners if they are optimised for AMD GPUs rather than M-series/Metal.
- Development platform for the AR/MR goggles and ML-based developments? Please, its 2023 and developers won't get out of bed if they can't work remotely (and even when they are in the office they're collaborating using the same devOps tools that they'd use at home) - the development system for any new platform is most likely to be primarily cloud-hosted. All these products will have to be delivered on iPhone-class hardware via mobile connections, and any extra horsepower developers need to pre-render content, train up ML models etc. can be provided on-demand in the cloud (which is where all the assets, training data for models etc. will live anyway). OK, that might not suit everybody just now - today there may be enough people working the 'old fangled way' to sell a few super-powerful personal workstations - but that number is only going to go down, so now is not the time to invest zillions developing a chip tailored for a dying sector when you could be developing the ultra-mobile SoCs for the delivery platforms that are going to sell by the millions.
- Hardware for running cloud services? The 2019 Mac Pro already isn't designed as a data centre server, and when it comes to running their own online services, Apple haven't eaten their own dogfood for years (if ever). MacOS is perfectly capable of being a server OS, but has nothing to offer over Linux once you unplug the screen and mouse. The xServe died for a reason.
So, Apple don't really need a 2019-style Mac Pro. If that's your current workflow, it sucks, but then moving to the ARM instruction set for you niche software - or even a forced upgrade to the latest MacOS - were probably going to suck, too, and considering that the last 3 Mac Pros (4 if you count the iMac Pro) all ended up being abandoned for years and then replaced with something radically different, I'm not sure what combination of misconceptions would make "real pro users" back the next Mac Pro horse.
That's not saying Apple won't release a Mac Pro. They might decide to burn the money anyway and do it properly. OR a Mx Ultra with PCIe slots for A/V cards and storage will tick the box for some people (but not the very high-end market the 2019 was aimed at). Or, something more adventurous with multiple Mx Max/Ultra 'compute modules' (again, unlikely to be a drop-in replacement for a 2019 MP). We'll see.
The only urgency is that the M1 Max Mac Studio is currently in collision with the M2 Pro Mac Mini which, with the best CPU/GPU option, costs about the same as the base Mac Studio, has a better CPU and is swings-and-roundabouts on the GPU (fewer cores, but individually faster - so your mileage may vary depending on the application). Must be decimating sales of the base Mac Studio (24 old GPU cores vs. 19 new cores in the top Mini Pro), but the 32 GPU Studio Max, not to mention the 64GB RAM option still seem viable. However, a M2 Max Studio would restore the balance so it seems a pity if that's being held back waiting for the Ultra option.
Trouble is, a kludgey Gurman-esque M2 Ultra Mac Pro that cannibalised a couple of TB4 ports to provide some (non-GPU capable) PCIe slots probably would clash with a new M2 Ultra Studio (c.f. a $4000 M1 Ultra Studio + $2000 Sonnet rackmount/PCIe enclosure). One possible move for Apple is to simply re-badge the M2 Studio as Mac Pro (I mean, it's more credible than the Trashcan in that respect).
One edit to your comment though. It originally started with 256GB of SSD, not 512. I know because even with the high price I was going to buy it day one until I saw the specs.
The point is x86 server chips with many cores exist (Genoa 96 core or even Sapphire Rapids 56 core) with many pcie5 lanes for expansion, and a ton of ddr5 memory channels as well.
And Even better x86 chips are coming (which is probably in apple’s calculus). Granite rapids and zen 5.
Those x86 chipsets support memory expandability and pcie that “Pros” want. It is rumored that the apple silicon Mac Pro won’t support memory expansion or add in pcie graphics cards. That’s a no go for many pros. So if apple silicon isn’t expandable as an SoC, why not just use an x86 server platform for its expandability, and throw in an apple silicon-based coprocessor or accelerator over pcie5 to accelerate the tasks that creatives use like Final Cut video etc. there is more than enough bandwidth on the pcie5 bus for an apple silicon accelerator to flourish along side an x86 server platform.
Or apple could just do what intel amd and Microsoft have done for decades which is to develop an expandable platform with pcie, ram, and storage expansion.
It seems game over for the Mac Pro. They should announce, that the Mac Pro is EOL and if anyone needs the Xeons, should look elsewhere.
PCI would be problem if there were many modern usages for PCI. They aren't. PCI-e yes. PCI no.
There is no huge disconnect with PCI-e with the current chips. Apple implements four x1 PCI-e v4 connection on M1 and M2 chips so for. It is what Apple uses to hook up Ethernet and Wifi/Bluetooth , USB controllers. So far Apple is mainly focused on PCI-e v4 to reduce the number of lanes coming out to of the chip and not as much on cranking up maximum bandwidth. But they do have a presumably mostly working PCI-e v4 implementation. They just don't have larger aggregate bandwidth ( e.g., one or two x16 PCI-e v4 controllers).
Similarly they have on die Thunderbolt controllers so basically have four x4 PCI-e v3 controllers embedded inside those TB controllers. Again they have done some PCI-e work so not a "moon shot complexity" problem to put any PCI-e on the dies.
Two low complexity changes ( that require work but no where near painted into a corner problems. ). One path would be to take a 'Max sized' die and replace the four x4 PCI-e controllers and surrounding TB controllers infrastructure with one x16 PCI-e v4 controller. Run the one x16 PCI-e v4 controller out to a PCI-e switch ( MP 2019 has one) and ta-da can provision 5-6 slots that way.
Second low disruption path would be to put a small, narrow PCI-e controller shim between to Max dies and connect all three with two Fusion Connectors. Part of the issue is that most of the die edge pace for external connection on Max dies is primarily thrown at Memory and the Fusion connector. If just leverage the Fusion connector then don't have 'conflict'. Again one or two x16 PCI-e v4 controllers run out to a one , or two , input PCI-e switch ( the MP 2019 switch is a two input switch).
Once have the PCI-e slots then multiple compute GPUs are far more a matter of just drivers than anything else. Multiple internal storage drivers is basically trivial ( drivers already exist). There are already 50+ cards that would work in PCI-e slots.
[ display GPU are a bit more dubious due to entanglements with the new macOS features of transparent iPhone apps and implicit assumptions of macOS on Apple Silicon apps. ]
RAM slots are different , but unless Apple goes to supporting ECC then higher capacity isn't the primary issue. ( M2 Max is 96GB so 192GB. There is about in range where no ECC is a reasonable boundary. Go > 200GB and then in dubious zone. Apple's soldered RAM is generally faster. It is not as much as a 'painted into a corner' situation as a trade-off taken one. [ At some point though Apple should tackle ECC though. RAM densities are going to get bigger and chiplets for scaling are going to make lots more sense. The "Max" almagation isn't a good chiplet. It physically doesn't scale well. ]
While that comes across as AI is here and now, it’s also indicative of it isn’t. It seems that we are advancing on using the term, but it’s not really AI. All those AI is the future need to be looked at as another form of data mining by google and others. Now as far as separate graphics cards the last thing I saw was Apple making their own external cards rather then partnering with nvidia because of how they are advancing metal3 to be their graphics engine. Supposedly we expect MacOS 14 to add to that. The comparison of what metal3 provides vs what it lacks is best seen with resident village playback demos. Still a WIP. As far as AS Macs being the same as traditional PC workstations that is still not known, because we haven’t had any technical discussions representing a Mac Pro, just extended conjecture on what might be used.
I really like Mac OS. It's easily Apple's best product... but for some screwed up reason, half the population will gladly get ripped off on the price of iPhone but they will NOT get a Mac computer to go with their phone? WTF. Why aren't Macs more popular?
What makes a computer decent isn't subjective; we can pretty easily give a value to a number of factors and pretty much all of these are in the hardware. And with all the hype Apple tried to sell us years ago about how great M1 chips were, they still haven't delivered a solid desktop for power users.
Because the macOS without the other 20M/year other systems of the rest of the Mac line up isn't really sustainable. Mac Pro sales are likely two orders of magnitude smaller. macOS isn't an extremely large system segment to begin with. 0.9% of 10% is approximately zero in terms of whole percentage points.
The dubious assumption here is that this faction stripped off the rest of the mac line up is itself a viable collection of users. Likely looking at 100K/yr or less folks. Even if collected 5 years of those folks together (0.5M), you would still be in the less than 1M users zone. At the start of the transition there were 100M Mac Intel users supported. That number is shrinking. The number of Intel Mac Pro sales is not outstripping the number of Intel Macs falling onto the de-supported vintage/obsolete list each year. That won't change if Apple continued to sell a speed bumped Intel Mac Pro.
In 3-4 years the real base of system that viably support macOS on Intel is going to disappear. There is nothing a Mac Pro all by itself can do to fix that. It is a niche of a niche of a niche that is likely only going to get smaller over time.
macOS doesn't support over 64 threads so pointing at AMD/INtel SMT enable systems that blow past the 64 thread limit doesn't mean much. It isn't a commerically viable place for macOS to be anyway. [ Forking macOS on Intel even farther away from the Apple silicon version is only going to make the economies of scale worse ; not better. ]
3rd party GPU drivers very similar issue. Without the unit numbers of the dGPU used in MBP 15" and iMacs this is interesting to AMD how? The largest GPU vendor over last 10 years on Macs was Intel GPU and those got dumped right along with the Intel CPUs. The dGPUs on the Mac Pro was a "fall out" of the embedded dGPUs of other Mac systems. ( NVIDIA faded when they stopped getting embeded GPU wins. And sealed their fate by throwing gas on the fire and blowing up bridges behind them as they left. (e.g., "halt and catch fire" drivers decoupled from maintream macOS updates. )
Again ... economically viable how? Apple already have accelerators that don't need PCI-e v5. In fact, use a bus that is faster than PCI-e v5 ( in bandwidth and latency). So where is viable to build yet another accelerator with different and slower bandwidth/latencies for a diminishing smaller market? ( the scale on Apple's market crosses most of their entire product line; not just macs. )
Apple could perhaps sell a "Mac on a Card". A Mn Pro probably could fit on a 75W bus powered card.
If Apple added PCI-e slots internally they would have internal expansion. The drivers for that are already working (been working for years on macOS on Apple Silicon). Apple Silocon has PCI-e . It is more a matter of "how many lanes provision" than "do you have it". There are four x1 PCI-e v4 lanes now. Apple would need something like one , or two , x16 PCI-e v4 controllers to provision out decent bandwidth. That doesn't really require any Earth shattering, "moon shot complexity" changes to what they already have. A bit more work that will cost some more money. ( really need a more desktop oriented chiplets. But don't need to throw the baby out with the bathwater. )
Thrown and 'extra' x1 PCI-e v4 lane at a SATA controller and/or USB controller and can provision the 1-2 SATA/USB internal sockets the MP 2019 has. Again wouldn't be major change from what they have to do legacy interface internal storage expansion with the 'building blocks' they already have.
Intel, AMD , and Microsoft are trying to make "everything for everybody". Apple is not. Apple could cover more folks with some internal PCI-e slot provisioning. Not everyone , but more even if don't chase every possible hyper-moldularity option.
Probably - It certainly didn't bother Apple that the $6000 2019 Mac Pro could be beaten in a sprint by $3000 iMac, because the MP was all about extreme expandability and Xeon/ECC stability.
Gurman's prediction does make sense if you believe his (I think?) other prediction of a M2 Ultra Mac Pro with all the same limitations as a M2 Ultra Mac Studio beyond the capability for a few non-GPU PCIe cards. Now, I wouldn't say no to a M2 mini-tower that could take a couple of PCIe cards to add extra USB controllers and some M.2 slots but that's not going to address the same market as the 2019 Mac Pro. If Apple wanted $6000 for that they probably would have to resort to shenanigans such as holding back the M2 Studios. We'll see - my guess is that someone saw a M2 Ultra test system lashed together in a Mac Pro case and got carried away.
Looking back, it took about 4 months between the launch of the M1 Max in the MacBook Pro and the launch of the M1 Ultra Mac Studio (and the M1 Max studio presumably had to wait for that). The M2 Max was released in late January, so its really not surprising that we've not seen any M2 Ultra action just yet.
From Digital 2023 - Global Overview Report.
Left to our own devices
But even as our average mobile time increases, GWI’s latest data shows that computers still account for an important share of internet users’ connected time.Mobile phones have long been the dominant device for internet use across most developing economies, but GWI reports that PCs, laptops, and tablets still account for more than half of people’s connected time in 18 of the 46 countries for which this data is available.
And crucially, computers still dominate online activities in the United States and Canada, as well as across most of Europe.
Belgians and Danes spend the greatest share of their internet time using larger-screened devices, at 42.5 percent and 45 percent respectively.
At the other end of the spectrum, internet users in Indonesia, Thailand, India, and China spend more than 60 percent of their connected time using a mobile phone.
-----------
So this is very much a regional/cultural preference also observed with travelers/family's from those countries working in other countries. You can also link it to the two main camps of users types which is Internet users vs Social media users.
I'm going to say the iPad is at fault for being more in one camp versus someone else using laptops. They are quite a bit cheaper to own, and like the phones they can be used around the clock compared to someone using Macs.
Ok, but iPads are just glorified iPhones. In some ways they're worse (iPhones got the better camera). There really isnt THAT much of a difference between the two.
The "AI revolution" will be delivered on mobile & embedded devices, for which Apple Silicon SoCs - with powerful on-die GPUs and neural engines - are not only ideal, but are already in the hands of millions of potential users.
Developing AI applications, training models, pre-rendering content and providing the server-side processing clout can happen on generic PC boxes stuffed with NVIDIA GPUs for all Apple cares. It's not like Apple services like Music/Maps/TV are being run from old xServes or racks of Mac Minis...