Apple's Last M4 Mac: What's Rumored for the Mac Pro

[Unami]

As M5 design is probably already finished and they designed it before people started to buy their "ultra"-chips for AI, there'll probably be no M5-Ultra as well.

I wouldn't be surprised if the Mac-Pro completely goes away. They've killed it. I've yet to see one in the wild (in video or audio production). There's not really a point in this machine, except maybe for some very, very special case where you can't put a nowadays already rare PCI-IO-Card in a thunderbolt enclosure. Sure, pros like expandability - especially RAM and Harddrives - but at that price it's not really feasible, even for big businesses.

 

[cupcakes2000]

That was when you could upgrade this computer.
Now its an overpriced Mac Studio in a full size tower.

I highly doubt any Hollywood studios that buy macs are buying them with a view to upgrading them. They’ll just buy maxed out versions and then replace them. Indie musicians and film makers, maybe you’re right. But really it’s tinkerers that upgrade. These days people buy what they need and use it until it’s not good enough anymore, sell it or repurpose it, and buy again.

 

[Bug-Creator]

I wouldn't be surprised if the Mac-Pro completely goes away. They've killed it.

Thats what they said when the trashcan had no updates in 6 years.
Thats what they said when the cheesegrater_2.0 (Intel) had no updates in 4 years.

 

[Jperkinski]

People that will never buy a Mac Pro love to hate on the Mac Pro. Having owned a handful of them since the mid 2000s I am still a fan and plan to buy the next updated one. Aside from a few older G5 towers back in the day, I had a 2019 Intel (when it was the new design) and currently have an M2 Ultra.

For me, I have a handful of RME PCIe AES audio cards, and putting a few super fast OWC SSDs inside for my audio projects makes things super fast and tidy compared to a Mac Studio or Mini with a bunch of add-on stuff.

In a busy pro audio studio, stability and speed are a paramount so using an external Thunderbolt chassis (while still pretty fast) opens up some risk for things to run slower than if it was all in the Mac Pro. Plus, by the time I would get a Mac Studio and all the extra stuff the price is essentially the same.

So, while people that will never buy a Mac Pro love to hate on these, please know that professional audio (and video) engineers still rely on these types of machines so I am glad to see Apple still making them.

Sure, they are not as upgradable as they used to be but they are still more customizable than any other Mac desktop or laptop offered these days when you consider what you can put in the PCIe slots.

And yes, they are not the cheapest option but I'm not looking for the cheapest option. I'm looking for a powerful machine that boots up everyday, and doesn't get in the way of my work due to performance issues. I'm looking for a machine that can render big high-sample rate projects quickly and doesn't bog down or act up.

This is why we pay the extra price. Time is money.

I also have a MacBook Pro for simpler/mobile use and a MacBook Air (M4 now) for internet and light duty work so I get it, but for my main setup the Mac Pro is absolutely perfect for my needs, and I'm sure many other audio and video studios.

 

[patent10021]

If it doesn't allow for a minimum of 2048 GB of memory, it's a waste of time. Need a minimum of that for full-precision base models + simultaneous workflow.

 

[617arg]

I'm one of those few people who have a 2019 Mac Pro with maxed out MPX modules and just recently a base 2025 Studio M3 Ultra. I've was using the Mac Pro for 3D rendering and whilst I have a PC with a couple of Nvidia cards and it's quicker than both of the Macs, I just prefer using Mac OS.

The fact is those, this studio runs rings around the MacPro for straight performance. It also generates way less heat and is silent, noting my PC can claim. Yeah there are still some speciality audio cards but the majority of audio interfaces are still USB-C, there are only a few Thunderbolt interfaces. PCIe audio are pretty scarce.

Blow are my results. Unless they come out with some M4/M5 Ultra/Extreme chip with PCIe slots I can't see the point of it as much I hate to say it.

2019 MacPro with 2x 6800XT Duo's
Self Build PC with 1914900KF + 4070 Ti Super + A4000
Mac Studio M3 Ultra Base 28cpu/60gpu

Redshift Benchmark:
MacPro: 2:30
Studio: 2:03
PC: 1:37

Cinebench Multicore CPU
MacPro : 1169
Studio: 2666
PC: 2016

Rendering an actual scene in Cinema 4D and Redshift. The scene has a lot of translucent objects, think trees, grass etc so not the easiest thing to render. Output was at 3000px X 3000px.

Mac Studio 19:12 RTX ON
Mac Pro 2019 41:48
Windows 11 + 4070 Ti Super 14:58
Windows 11 + 4070 Ti Super + A4000 10:21

I prefer using Mac OS too, but can't justify a purchase any more without a real GPU. We do a lot of rendering using Twinmotion and not being able to use raytracing (and being limited in Lumen) is essentially a deal killer. We are looking for new computers now and will probably end up with Dell (Alienware) machines. This is based solely on being able to use the latest NVIDIA GPUs. It's sad that Apple seems to have abandoned this market.

 

[Velin]

We are looking for new computers now and will probably end up with Dell (Alienware) machines. This is based solely on being able to use the latest NVIDIA GPUs.

How many PCs do you need? If it’s just a few, you could do a lot better building them yourself. Much better components, cheaper price. And if you skip the RGB and watercooling nonsense, they aren’t too hard to build.

 

[AndiG]

Thats what they said when the trashcan had no updates in 6 years.
Thats what they said when the cheesegrater_2.0 (Intel) had no updates in 4 years.

But - and this is. a serious question - does the cheesegrater sell? (more than 5?).
In professional computing, Linux is definitively the place to be, in AI it is the place to be - but Linux may still lack some high demanded video editing tools.
Personally I like starr thinking to move even more to Linux (gnome 47/48 is. a pure joy) without artifical Apple restrictions and without macOS bugs ...

There will be a day. The Linux that runs on Mac hardware looks promising.

 

[617arg]

How many PCs do you need? If it’s just a few, you could do a lot better building them yourself. Much better components, cheaper price. And if you skip the RGB and watercooling nonsense, they aren’t too hard to build.

It's just a few (2 or 3). I would prefer to build them myself but my partners just want to get them up and running as soon as possible, which I understand. I already have a build list for what I would purchase for a build, which I might still do for a machine at home.

 

[drugdoubles]

Across 2024 and the early months of 2025, Apple refreshed all of its Macs with next-generation M4 chips, with the exception of the Mac Pro. The Mac Pro is still waiting for an update, but it is supposed to get an overhaul later this year.

M-Series Chip

The current version of the Mac Pro can be purchased with Apple's M2 Ultra chip, which came out in June 2023. It might seem logical for the Mac Pro to get the equivalent M4 chip, the M4 Ultra, but it turns out Apple might not have an M4 Ultra in the works.

When the Mac Studio was updated earlier this month, Apple announced a version with the M4 Max chip, and a version with an M3 Ultra chip, with no M4 Ultra unveiled. Apple told Mac Studio reviewers that not every generation of M-series chips will include a higher-end "Ultra" tier, so there may simply be no M4 Ultra that exists for the Mac Pro.

There's now a question over what chip Apple will use in the Mac Pro, and there are a few possibilities.

1. Apple does have an M4 Ultra chip coming, and it's not ready yet.
   
2. There's some version of a high-end M4 chip that is not technically an "Ultra" chip and is instead called something else like "M4 Extreme."
   
3. The Mac Pro will use the M3 Ultra chip.
   
4. The Mac Pro will get an M5 Ultra chip.

The M1 Ultra, M2 Ultra, and M3 Ultra chips that Apple has released have essentially been two Max chips linked together through an "UltraFusion" connector. The M4 Max does not have the UltraFusion connector available, so the first possibility seems unlikely.

Apple could be making an M4 Ultra or Extreme chip that is standalone and not a doubled up variant of the M4 Max, but Bloomberg's Mark Gurman recently claimed that Apple doesn't want to create an M4 Ultra chip from scratch because of costs, production challenges, and low sales of high-end and expensive machines.

Apple could refresh the Mac Pro with the same M3 Ultra chip that it put in the Mac Studio, but if that's Apple's plan, it's likely the company would have just refreshed the Mac Pro alongside the Mac Studio. The Mac Pro could be held back for other development reasons, but there aren't really rumors of notable new features coming.

We might be getting the first Macs with M5 chips later this year, but there's also a chance M5 Macs won't come until early 2026. Even if Macs with M5 chips do launch in late 2025, there's no guarantee that an Ultra version of the chip will be ready to go.

As of right now, there's no clear indication of what's in store for the 2025 Mac Pro's Apple silicon chip.

Design

There aren't rumors of design updates for the next Mac Pro, so it's not likely that Apple has anything planned.

Ports

Apple added Thunderbolt 5 to the Mac Studio and MacBook Pro, so the Mac Pro will likely get Thunderbolt 5 ports too. Thunderbolt 5 will allow for more high-resolution displays to be connected to the Mac Pro.

RAM and SSD

The M3 Ultra chip supports up to 512GB RAM, so if the Mac Pro gets the M3 Ultra or something similar, it will support a lot more RAM. The current model is limited to 192GB.

Storage maximums will also double, as the M3 Ultra supports up to a 16TB SSD, while the Mac Pro is limited to 8TB.

Launch Date

Bloomberg's Mark Gurman said last year that the Mac Pro will see a refresh toward the end of 2025, but given the chip uncertainty, Apple could be planning to hold it until 2026.

Article Link: Apple's Last M4 Mac: What's Rumored for the Mac Pro

I expect m3 ultra will be used but nothing special like no new case or even no new name.

 

[TEG]

Just a reminder, when the M3 Max was originally released and dissected, the UltraFusion connector wasn't found on it either. I suspect Apple is trimming it off on the Max chips to make the die smaller. There may be an M4 Ultra, or given how the Ultra chips don't scale well, the M4 or later Ultra chips will be an entirely new design, that can then be fused to make the Extreme chips... if those are in the works.

 

[tenthousandthings]

I honestly don't think Apple’s not going to bother with an M4 Ultra for the Mac Pro. It’s pretty clear by now that not every chip generation is going to get an Ultra version, and the M4 Max doesn’t even support UltraFusion. So forget M4 Ultra. And they’re not going to use the M3 Ultra either—if that was the plan, they would’ve launched the updated Mac Pro alongside the Mac Studio just a few weeks ago.

What Apple will do is build a completely new chip for the Mac Pro—let's just go with that M5 Extreme naming. It’ll be a standalone, high-performance chip that doesn’t rely on fusing two Max chips together. It’ll be the top of the line, no compromises. Apple will likely update it every 3–5 years, which is perfectly fine for the Mac Pro’s niche. But when it drops, it’ll crush everything else in their lineup, including the Ultra chips in the Mac Studio for years.

My bet is they’ll announce it at WWDC this year, alongside refreshed Pro Display XDR and Studio Displays. Then, a couple of years down the line, we’ll get the M5 Ultra—a cut-down version of the Extreme—for the Mac Studio. That’s how they’ll stagger their chip tiers going forward.

This would finally solve the long-standing issue of the Mac Pro being left behind. It puts the Mac Pro back at the top, where it belongs, and keeps the Mac Studio in a solid middle tier with regular Ultra upgrades every other year.

This is the only move that makes sense in my opinion. If Apple’s serious about differentiating the Mac Pro, they need to give it a real performance lead—not just recycled silicon that is glued to another chip.

I think you are right in general (I’m still hopeful for a true two-year cadence, with marketing aimed at a four-year upgrade cycle), but you are wrong about what the silicon will be.

The key thing you are missing is the increasingly-essential role of advanced packaging. When you dismiss it (as “recycled silicon that is glued together”) you underestimate the importance of Apple’s initial, pioneering foray into this realm, the world of so-called “chiplets,” with UltraFusion (InFO-LSI, the same technology Nvidia uses for Blackwell Ultra).

M5 Ultra and Super (since Nvidia is using Ultra now, Apple should feel free to use “Super” instead of the rumored “Extreme”) will introduce both SoIC in the Pro/Max, and CoWoS-L in the Super. I think they will still use the Max as the basic building block, but it will be an SoIC (instead of SoC), with integrated chips, with more flexibility, allowing for two or more Max variants, leading to two or more Ultra (via InFO) variants that would then be combined into the Super (via CoWoS).

The Max is already near the reticle limit. It can’t get much bigger without using advanced packaging. The rumors are SoIC and CoWoS, along with InFO, they make perfect sense, and they also mesh with Apple’s needs for its PCC (Private Cloud Compute) silicon in its custom servers for that.

The thing that you get, which a lot of others don’t get, is that Apple has conceded the modular-internal-systems market to AMD’s Ryzen. When Apple’s executives repeat the company mantra, that their silicon is tailored for their products, that’s what they mean. There’s a fundamental difference between how AMD approaches this and how Apple does. Strangely enough, Nvidia’s approach is much closer to Apple’s in this respect.

 

[XboxEvolved]

I miss the days of buying a PowerMac G4 and being able to upgrade essentially everything inside of it. Somewhere along the way their main desktop got incredibly expensive. Even adjusted for inflation, the PowerMac G4 would be less than $3k. I'm not holding my breath that we will ever get prices that are lower than the $7k minimum that we see now for them, but I don't see why you can't at least be able to add internal SSD, and graphics cards.

 

[nathansz]

but I don't see why you can't at least be able to add internal SSD, and graphics cards.

Apparently their is no way to add a gpu to Apple silicon

 

[theluggage]

They'll probably try to release some sort of trash can version again where things are locked down, which will flop (because it's not the intended audience, if you want a locked down one you can just buy the Studio)

That literally is the Mac Studio - and it doesn't seem to have flopped.

The current Mac Pro is for people who need more PCIe slots. lanes & bandwidth than you can sensibly get using Thunderbolt enclosures (which are 4 lane only and usually a generation of PCIe behind the MP). For internal storage & specialist I/O cards - probably a lot of legacy audio & video cards that could theoretically be replaced with Thunderbolt-specific stuff by people who would completely re-tool if they had the time and money. If that's not you, then you can stop worrying about the Mac Pro.

If it doesn’t have an M5 Ultra chip then what is the point of it?

6 full length PCIe slots sharing 16 lanes of PCIe 4 (versus 2-3 slots sharing 4 lanes of PCIe 3 per external Thunderbolt PCIe enclosure - PCIe 4 maybe soon with TB5). That's it. That's the point. I don't need that, but apparently others do.

I'd quite like a Mac mini-tower in the ~$2k price range lie the good old days - mainly so I could fill it with USB 3 cards - but it isn't going to happen with Apple Silcon since you need the second die of a Mx Ultra chip with its unused SSD interface to provide those PCIe lanes.

It is meant to be the flagship model,

Says who? It's a specialist model for a niche who need the slots, and has been the last in line for updates since about 2012. It will go away as soon as that "niche" has moved on to a Thunderbolt-based workflow.

If anything, the MacBook Pro is the "flagship" - laptops have been Apple's strongest point since the 1990s, and its the laptop market where chips like the M4 Pro and Max really give the MBP a power-consumption/performance edge over x86. - and a the syetem that "everybody wants" (the latest NVIDIA and AMD discrete GPUs) would throw away that advantage and never be any better than those GPUs (which are also available in PCs). Apple Silicon is terrific for laptops and small-form-factor workstations, which is where Apple will make their money, but if you want a big box'o'slots it just isn't the best tool for the job.

Upgrading the MP to the M3 Ultra might just happen - the motivation would be that 512GB RAM option (maybe for people who needed a 100 piece orchestra worth of samples permanently loaded into RAM) but you'd really have to be committed to a Mac workflow to pay that cash.

 

[theorist9]

I prefer using Mac OS too, but can't justify a purchase any more without a real GPU. We do a lot of rendering using Twinmotion and not being able to use raytracing (and being limited in Lumen) is essentially a deal killer. We are looking for new computers now and will probably end up with Dell (Alienware) machines. This is based solely on being able to use the latest NVIDIA GPUs. It's sad that Apple seems to have abandoned this market.

While not as powerful as the top GPU offerings from NVIDIA, the M3 and M4 GPU's do have hardware ray tracing. Is there some reason their ray tracing doesn't work for you?

 

[theorist9]

In professional computing, Linux is definitively the place to be, in AI it is the place to be

That may be true for general GPU computing, but I'm not sure about that for AI specifically:

"...every AI developer I know uses a Mac! Essentially, and I am generalizing: Every major lab, every major developer, everyone uses a Mac."

Max Weinbach, Creative Strategies

A weekend with Apple’s Mac Studio with M3 Ultra: The only real AI workstation today

Discover how Apple's Mac Studio with M3 Ultra transforms AI development, offering unmatched GPU power, vast Unified Memory, and optimized MLX integration.

creativestrategies.com

 

[macduke]

I could see them differentiate between the Mac Pro and Mac Studio by intentionally updating the Mac Studio a little earlier while leaving it a generation or two behind when the Mac Pro updates. Sucks because the Studio was such a good deal. I would get one except that I prefer a laptop. I did the whole two Macs thing for four years and it was a pain. I have way too many different things going on at once to switch from my desktop over to a laptop and back without it causing a lot of friction in getting everything set back up on one or the other. So I ended up hardly ever using my MacBook Pro. Thats why when M3 Max came out and I saw how big of a leap in performance it was, I went for that in 14”. Maximize performance when docked while maximizing portability when not. I freaking love it! Can’t imagine going back to a desktop with performance and battery life this good in a laptop.

 

[Maigrais]

Mac Pro 2019 model will be remembered as a crown Apple Mac desktop forever. It will retain its title as the most expandable Mac that Apple has ever built. None quite like it before & none quite like it after. It is desktop nirvana.

 

[617arg]

While not as powerful as the top GPU offerings from NVIDIA, the M3 and M4 GPU's do have hardware ray tracing. Is there some reason their ray tracing doesn't work for you?

Ray tracing / path tracing is simply not available on the Mac for the rendering software we use. We end up having to use a lesser quality/limited version as compared to machines with NVIDIA GPUs. While I appreciate Apple pushing forward, it seems like it will take a while for developers to catch up and create an equivalent experience.

 

[Squidiey]

Plot Twist: the M4 Ultra Chip will be a completely new chip instead of 2 M4 Max Chips

 

[arobert3434]

Option 5: The current Mac Pro is the last one.

And honestly, they probably only released that one just to officially close out the Apple silicon migration and bring it to a graceful end.

 

[deconstruct60]

As M5 design is probably already finished and they designed it before people started to buy their "ultra"-chips for AI, there'll probably be no M5-Ultra as well.

Seems unlikely. The "Ultra" isn't a Mac Pro only chip. The M3 Ultra has a narrower range of workloads that it does better than M4 Max. Once the M5 Max arrives on the Studio, the M3 Ultra pretty likely won't be viable anymore for the Studio.

If Apple is going to keep two SoC options for the Studio they will need "something else, that is bigger than a Max" SoC. Perhaps it won't be called "Ultra" , but it probably will be an more capable option. ( Apple already has the name .. so pretty likely they will just reuse it. )

I suspect the Studio systems aren't going to get yearly updates. So even if there are once again two variants of the Max Max design ( one without and one with UltraFusion) the second version would just come substantively later and stay on market longer (i.e., until a M7 Ultra-class solution arrived. ).

I wouldn't be surprised if the Mac-Pro completely goes away. They've killed it. I've yet to see one in the wild (in video or audio production).

The primary missing piece for Mac Pro is more software ( drivers updates/broadening) and I/O bandwidth upgrades.

There's not really a point in this machine, except maybe for some very, very special case where you can't put a nowadays already rare PCI-IO-Card in a thunderbolt enclosure. Sure, pros like expandability - especially RAM and Harddrives - but at that price it's not really feasible, even for big businesses.

If Apple tweaked the 'desktop' variant of the Max die to be able to drive the dual input PCI-e switch for the Mac Pro that would directly address the price 'problem'. For example if the 'desktop' Max die connected to an I/O die that had two x8 PCI-e v5 lanes to supply backhaul to the switch they could pull the MP price closer to a Studio ( e.g., $1,500-2,000 gap over the Ultra Studio versus the $3K gap now. ) , but a chopped down Thunderbolt I/O card.
[ the ultra configuration would get better backhaul maybe two x16 PCI-e v5 of x&+x16 provisioning. Along with the additional CPU/GPU/NPU/etc units. ]


Open up the drivers so that someone would put an AI Accelerator in the systems and it would be bigger still. (i.e., 'grow' the PCI-e card ecosystem for macOS ).

 

[deconstruct60]

Had an interesting conversation with our Apple rep this week.
As you know, Apple have AI servers crunching their LLMs using Apple silicon, apparently running a special version of iOS.
I jokingly asked if they were xserves! But they are some bespoke server.
I wouldn’t think they’re rack-mounted Mac Pros… they could be I suppose…but wouldn’t it be better and more efficient all round to go with a “slot in” M4? A few SoCs to a unit? Perhaps ultra fusion has evolved into a socket?!

Not a 'socket' but into a custom/bespoke logic board. Strip off the 3-4 USB-C style sockets. Drop any chips for WiFi and Bluetooth. Perhaps add 1-2 more 10GbE sockets (and discrete Ethernet processing chips).

Could also shrink the local SSD storage. This custom PCC operating system doesn't persistently store any user data ( once computation is done, any local user data is wiped. There is no long term file storage. )

In short, it doesn't mean the SoC is different from the chips sold in end users Macs. It would more so have impact on the other 3rd party chips that Apple places around the SoC and the size of the logic board.

These could be set up akin to PCI-e cards or just follow the OAM
https://www.opencompute.org/blog/ne...oject-to-launch-within-the-ocp-server-project

The Apple Mn whatever is still soldered to the logic board. It just isn't a logic board that fits inside of external Mac container. To limit the board to some container than Apple Design came up with it what doesn't make much sense. It is only deployed in a locked data center. Nobody is going to spend lots of time staring at the enclosure contemplating its beauty.

 

[deconstruct60]

That would not be ultrafusion.

But.... it is plausible that they use custom silicon in those servers that as such isn't used in any consumer product.
It also not completely out of the question that the MacPro we know is dead and eventually comes back later as a dedicated AI workstation with just those chips.

If the silicon is specifically (narrowly) custom just to run Apple Intelligence subset workloads then that doesn't seem likely. Apple won't be trying to clone/compete against Intel Xeon and EPYC chips. It would be datacenter Google Tensor and AMD MIxxx and Nvidia data cards (not gaming, commerce cards).

If so skewed that drops all the display engines ( generates no video out), then it would be pretty useless as a Mac Pro SoC. macOS is a Graphical User Interface and if there is not graphics output ... not much of an interface.

Ditto with keeping around several Thunderbolt controllers. Pretty likely would toss those for 100-200GbE controllers for datacenter deployments. (especially if all the persistent file storage you want to access is off in the network (non local. remember PCC OS doesn't keep local files.). This farming out workload to 'cloud' means moving lots of network data in the aggregate. )

Indeed, one of the rumors.
" ... At least a trio of companies are believed to be involved with the chip. Apple is said to be handling the overall design of the chip, while Broadcom is said to be providing some networking technology for it. TSMC is expected to begin mass production of the chip in 2026, using its third-generation 3nm process, known as N3P. ..."

Apple Intelligence Servers Expected to Get All-New, Turbocharged Chip

Apple Intelligence servers are currently powered by the M2 Ultra chip, and they are expected to start using M4 series chips next year. In an eventual...

www.macrumors.com

The network tech that Broadcomm is going to bring isn't going to be THunderoblt or USB-C port based. What is the primary day-to-day use of USB-C in a AI datacenter farm? Not much.

If Apple is trying to do a monolithic chip then the hope would be more dimm; not brighter for a Mac Pro reuse. [ Apple's high level goals to have 'totally green powered' datacenters and so far avoidance of nuclear suggests they would be leaning toward monolithic to save power. )

If Apple is actually getting on the chiplet bandwagon and Broadcomm networking is via a UCIe or UltraFusion connection to a separate I/O chiplet . Then maybe Apple could substitute another desktop I/O chiplet for Mac Pros (and would have removed the desktop ports I/O from their base 'compute cores' chiplet) . But technically that is a different SoC. Some chiplet reuse there , but not all.

The Max isn't really a good chiplet design. It is a bit too chunky.

Could be "slow" at running normal apps (if thats even an option) but crush it on purpose build SW.

I suspect the 'server' deployment is more so about the 3rd party chips around the SoC ( or possible chiplets made by other folks in the SoC package) than in other stuff.

Just one of the many far out possibilities on the future of the MacPro with the other extreme an M3Ultra one being silently dropped within the next few weeks.

For the immediate term, it doesn't appear hyper customer server chip is coming. 2026 for the rumored one with Broadcomm contributions.

If Apple had for make a M3 Max+ die to get to M3 Ultra, then it is unlikely they are going to drop it any time soon. Or that the "Studio Ultra" would sell in sufficient numbers to justify the doing another mask/validation/deployment.