MP All Models MacPro 9,1 Wishlist

[mattspace]

the purpose of the Mac Pro was never performance, that was a bonus you got because of the components used for its main job. The purpose was to be the stable (Xeon. ECC RAM), post-purchase reconfigurable, generic computer, upgradable Mac.

The Mac Pro was almost always outpaced on specific tasks by other Apple systems, which never mattered because as long as it was fast enough, being the connectivity / peripheral monster was its role.

 

[JesterJJZ]

I think Apple releases at least on more Mac Pro. Since four Studios can be hooked together, why can't Apple configure the existing Mac Pro case for say 2 to 4 "slots" so to speak for multiple M5 Ultras? Even if inside the case, you just snap in an extra SoC into some sort of bay and manually plug it in/daisy chain it to the base/other SoCs... there's more than enough space and power in the Mac Pro case. Would be a cleaner solution than having four Studios, multiple TB cables, four power cables, etc. If Apple allowed that kind of expansion in the Mac Pro, it would be interesting.

I think I’d honestly be fine with just upgraded internals as is. Current form factor with M5 Ultra Combo or whatever.

 

[Nugget]

the purpose of the Mac Pro was never performance

I think maybe this is overstating what is your opinion about the product line. The Mac Pro as a product has never been about just one thing like expandability. That may be the aspect of the Mac Pro that you value the most, but it's hardly the only aspect of the Mac Pro. The Mac Pro has been about performance at least as much as it has been about expandability, both in practice and in presentation from Apple. Before it was the Mac Pro, this product was the PowerMac and even its name conveyed performance over expandability.

Past ad campaigns for the PowerMac and Mac Pro have emphasized performance and the Mac Pro has been positioned as the pinnacle of computing performance within the Mac model lineup at many times during its product life cycle.

The purpose of the Mac Pro has sometimes been performance.

 

[JesterJJZ]

The biggest issue Apple has faced with MacPro being positioned for performance is the lack of updates. Even when the M2 MacPro came out it was already on par with Mac Studio and outpaced within a year. At minimum there should have been an update every other year.

 

[conmee]

The biggest issue Apple has faced with MacPro being positioned for performance is the lack of updates. Even when the M2 MacPro came out it was already on par with Mac Studio and outpaced within a year. At minimum there should have been an update every other year.

Or if you're allowed to RDMA a few SoCs in the case (or even just swap SoC/board instead of buying an entire new Pro), you could at least scale the investment and not be easily outclassed by the following year's SoC.

 

[mattspace]

I think maybe this is overstating what is your opinion about the product line. The Mac Pro as a product has never been about just one thing like expandability.

Expandability was the only thing the Mac Pro did, that nothing else could do.

On every other metric, the Mac Pro was eclipsed by consumer systems within 12 months of launch, and usually earlier.

*edit* And we saw what happened in 2013 when Expandability was removed from the purpose of the product, and it was made all about "performance". It more or less killed the product, because the performance value of it ran out within the first year, and then from a performance point of view, it was effectively just a MacBook without the portability (and a multiple of the price).

Before it was the Mac Pro, this product was the PowerMac and even its name conveyed performance over expandability.

The laptop of the time was the PowerBook, and it was named as such before the PowerPC chips that gave the desktop line that name. That's not purpose, that's marketing.

And PowerMac was used for consumer PPC 603e products like the 6400/6500, that were definitely not performance machines, but were expandable.

Past ad campaigns for the PowerMac and Mac Pro have emphasized performance and the Mac Pro has been positioned as the pinnacle of computing performance within the Mac model lineup at many times during its product life cycle.

And yet, it has always been below the performance of cheaper consumer systems from Apple, while still a "current" model.

Marketing is not the same as purpose.

 

[krakman]

The biggest issue Apple has faced with MacPro being positioned for performance is the lack of updates. Even when the M2 MacPro came out it was already on par with Mac Studio and outpaced within a year. At minimum there should have been an update every other year.

How difficult would it be to manufacture a chassis and backplane with proprietary slots whereby every few years you can buy and plug in a new daughterboard with the latest CPU?

I suspect that the supply chain is geared up to make disposable products that need to be replaced every few years and "sustainability" is nothing more than a marketing slogan.

 

[JesterJJZ]

How difficult would it be to manufacture a chassis and backplane with proprietary slots whereby every few years you can buy and plug in a new daughterboard with the latest CPU?

I suspect that the supply chain is geared up to make disposable products that need to be replaced every few years and "sustainability" is nothing more than a marketing slogan.

I think it's less about difficulty and more about profitability, unfortunately.

 

[keksikuningas]

How difficult would it be to manufacture a chassis and backplane with proprietary slots whereby every few years you can buy and plug in a new daughterboard with the latest CPU?

I suspect that the supply chain is geared up to make disposable products that need to be replaced every few years and "sustainability" is nothing more than a marketing slogan.

with how much the motherboard itsels probably changes, it might as well be you buy a whole motherboard.

and to answer your question about the daughterboard, I imagine it would be exceedingly more difficult to spend time on that for 0.1% of their mac userbase compared to just selling motherboards the users would need to screw in themselves (or go to apple store for that)

i wonder how many Mac Pros were actually upgrades throughout their professional lifetime compared to just staying at their ordered spec before being retired

 

[avro707]

One here, went from 16 to 28 core and afterburner added. The other was 28 core and 2x W6800X Duo from the start.

Both however were very expensive to begin with.

 

[keksikuningas]

One here, went from 16 to 28 core and afterburner added. The other was 28 core and 2x W6800X Duo from the start.

Both however were very expensive to begin with.

yeah but we are on a tech enthusiast forum, although you could make a point that someone spending that much money on a computer probably cares a whole lot about it to the point of bothering to hang out on a forum like this one

 

[avro707]

yeah but we are on a tech enthusiast forum, although you could make a point that someone spending that much money on a computer probably cares a whole lot about it to the point of bothering to hang out on a forum like this one

I was only here because I remembered I still had an account and it was useful to find information I needed to get an old 5,1 Mac Pro upgraded.

If that makes me not a proper professional user of the Mac Pro then so be it.

What mine are used for costs multi millions of dollars…

 

[keksikuningas]

I was only here because I remembered I still had an account and it was useful to find information I needed to get an old 5,1 Mac Pro upgraded.

If that makes me not a proper professional user of the Mac Pro then so be it.

What mine are used for costs multi millions of dollars…

I did not mean to say that you are not a professional. I meant to say that being on an enthusiast forum self selects for people who might bother upgrading and researching more into the system they are using 🙂

 

[Flint Ironstag]

However efficient all this stuff gets is great. But for a long time, there's going to be a demand for a solid local box that you can cram full of stuff.

 

[SDAVE]

the purpose of the Mac Pro was never performance, that was a bonus you got because of the components used for its main job. The purpose was to be the stable (Xeon. ECC RAM), post-purchase reconfigurable, generic computer, upgradable Mac.

The Mac Pro was almost always outpaced on specific tasks by other Apple systems, which never mattered because as long as it was fast enough, being the connectivity / peripheral monster was its role.

What are you talking about lol
Mac Pro was always the fastest Mac on the block, and Apple always advertised it that way...until Apple Silicone.

 

[SDAVE]

I've come to the conclusion that Apple won't abandon the Mac Pro, it will just keep updating with M class series.

There is a very niche market that still needs PCIe slots, like Production Houses, etc.

 

[mattspace]

What are you talking about lol
Mac Pro was always the fastest Mac on the block, and Apple always advertised it that way...until Apple Silicone.

For about 3 months, until the next laptop update was faster on everything other than big parallel multicore workloads, which most people don’t do.

That was always the pattern.

 

[Apple Knowledge Navigator]

I think the concept will remain the same, but they’ll shrink the enclosure down significantly.

That 3D ‘honeycomb’ milling is likely the most expensive and time consuming component of the device to produce, just as with the Pro Display XDR. Simplifying that aspect of production would not only reduce costs but allow the enclosure to be designed a fair bit smaller.

I can see the number of PCIe slots being reduced too and potentially having just one double height slot.

 

[Nugget]

For about 3 months, until the next laptop update was faster on everything other than big parallel multicore workloads, which most people don’t do.

That was always the pattern.

You are agreeing that Apple have historically positioned the Mac Pro as a performance solution and marketed it that way.

The fact that it also historically fails to hold a permanent spot at the top of the heap is a secondary observation.

 

[Apple Knowledge Navigator]

I don’t think video editing is what’s driving workstation performance innovation any more. I think a lot of the confusion I see about the future of the Mac Pro stems from a failure to see this new reality.

Video editing isn’t on the frontier of desktop/workstation performance any more.

Very well put, and correct. The fact that we can now edit 4K video on a phone should say a lot about how fast the performance bars have raised.

Mac Studio will serve the majority of media professionals well, and I believe Mac Pro will be all about AI, science, and those wanting the neatest solutions for connecting to a variety of storage and PCIe needs. It may continue to be a niche for Apple but, if they simplify the production of the enclosure and continue to develop chiplets for the M-series chips, it should be a lot more cost effective and quicker to assemble and hopefully better value than the current product.

 

[SDAVE]

For about 3 months, until the next laptop update was faster on everything other than big parallel multicore workloads, which most people don’t do.

That was always the pattern.

Apple silicone closed that gap though.

 

[deconstruct60]

Less than 3 months ago, in mid-December, Apple sent M3 Studio clusters to a handful of tech YouTubers along with a giant box of thunderbolt cables and instructions on how to enable RDMA over thunderbolt. This is a new feature that was just added in macOS 26.2.

...

This hardware stack of Studios and a spaghetti mess of thunderbolt cables delivers significant AI runtime performance even with the relatively modest thunderbolt interconnect speed and rivals NVIDIA architecture solutions that require 10kilowatt+ three phase power and datacenter-level cooling. Four 512GB Studios (RIP) give an addressable 2TB of space to run large, unquantized, frontier-caliber models.

One major reason the 512GB M3 Studios are sold out is this campaign to draw out a substantive number of buyers. In some respects it may have worked too well as the RAM supply for the largest capacity disappeared (and it isn't like extra RAM is going to fall out of the sky these days. If haven't ordered previously it at the system vendor level , then you probably are not getting it. )

The 'plain' M5 was out and the M5 Pro/Max are coming. The M5 Max covers the binned down M3 in many GPU respects ( especially in this AI space). If they had not sold those M3 Ultras off before the real stats on M5 Max came it probably would have gotten harder to get rid of them for anyone not completely maxed out on RAM needs. (Yes that is lots of these AI folks. )

Now imagine a hypothetical future Mac Pro solution where the slow thunderbolt cables are replaced by motherboard traces, or pluggable GPU/NPU daughterboards to significantly increase the RDMA interconnect speeds in between M5 Ultra chiplet resources.

The highly custom AI servers that Apple built for themselves do not decouple the GPU/NPU, but do get rid of loose network wires for networking to the card. sub pod clusters of 8 SoCs are hooked together with a custom logic board. A higher focus on RDMA cluster interconnect isn't going to be better or enhanced focus on video out. For a single person, GUI workstation that is pretty much as Mac . None of those 'compute cards' are consumer PCI-e lane provision focused at all. It is Apple's 'war' on wires more so than trying to provision maximum consumer commodity device plug-in ability. Custom cards to hook to their own stuff.

Apple is reportedly working with Broadcomm. That could easily get them 100-400GbE RDMA connections. But that primary is only cluster or data center useful. Apple going down that part is drifting away from the core Mac Pro market.

Generating tons and tons of AI slop doesn't require any GUI interface at all ( no macOS needed. Hence no Mac Pro necessary. )

Thunderbolt 5 RDMA is likely way more affordable.

 

[Nugget]

Apple going down that part is drifting away from the core Mac Pro market.

Curious what you think “the core Mac Pro market” means in 2026. It’s not video editing any more. My view is that the core Mac Pro market has shifted and some haven’t recognized that.

 

[deconstruct60]

Curious what you think “the core Mac Pro market” means in 2026. It’s not video editing any more. My view is that the core Mac Pro market has shifted and some haven’t recognized that.

What I think probably diverges a bit from Apple . I'm going to start off with a couple of 'nots' that are probably highly aligned with Apple.

1. Not pick GPU PCI-e card first and then build system around that.
( so it isn't primarily a container box for a favorite PCI-e card. It is not a generic 'box with slots". )

2. Do not want to run Windows and/Linux raw a significant amount of time.
( pre 2006 Mac were not 'go to' machines for Windows. Some intel era stuff was purely side effect of being on
Intel x86. None of that is essential 'Mac' properties. There is a sizable group who folded into "Mac Pro"
userbase who bought deeply into those side-effects , but wasn't core to Mac Pro design.
So it is not the ultimate tinker with everything imaginable box. )

3. In 2026.. not primarily focused on x86 apps. Users who are actively seeking to unload as many of those possible.
( not primarily backwards focused looking software users. )

4. Not doing long computations sensitive to data integrity. [ It doesn't look like Apple is going to be bring ECC to
their systems any time soon. ]

5. Not on some insatiable CPU core count quest. Has to be some number larger than 64 or 'bust'.

6. Not some 'bragging rights' , 'trophy' box. (***). [ I don't think Apple has 'conspicuous consumption' as a driver of Mac Pro. Nor would it work anyway. ]



So with that filter.

1. Single person and typically multiple screens standard. ( at least a couple of screens loading up the output on
a couple (or more) Thunderbolt ports for primarily video/audio out. ). Yes, this appears somewhat disconnected to the rack orientation , but it isn't datacenter (large or local closet) deployment that rack is oriented toward. It is still a macOS GUI usage, the system is just being fixed located and integrated.

In short, needs users highly focused on macOS. No macOS focus then not going to sell the Mac (Pro or not).

2. Probably people who like the build approach of the rack Mac Pro more than the desktop tower one. Folks not big fans of the 'nifty' twist handle and lift after having to disconnect all the wires coming out of the Mac Pro.
While Apple may 'hate wires', I think the core of the Mac Pro market that is left don't have as high a hate of wires inside and within limits coming out of the box.

It isn't a lateral 'desktop' system. It should be rack first and 'desk-area' second (e.g., optional '$999' stand to be deskside, off the floor, and vertical). [ literal desktop footprint wise the Mac Pro can't compete with the Mac Studio and trying to shrink the Mac Pro to try to compete that way is dubious. ]

3. Folks with either more than one relatively high bandwidth PCI-e card or a bulk card usage ( of low to medium bandwidth).

Examples
i. More than one storage device. One and only one internal disk was an issue in 2017 and it is a still an issue now in the > $7K systems space. Lots of data being visualized locally. ( that includes 'synthesized' more 'drawn' data visualization as opposed to classic raytracing. )
[ NOTE: SATA is basically dying in high performance contexts. Some SATA SSDs are being discontinued. I don't think Apple wants to track keeping them around in but previous systems of this class has multiple drives. It is a different connector and storage technology, but is the number really going to drop at the high end? RAID-0 drove a larger number of SATA drives to get to one volume image, but storage is still way slower than RAM. It is still a bottleneck. ]

ii. better than 100GbE networking. ( Apple needs to so substantive work here on driver support and cooperation). [ there is a part of Apple that thinks Wi-FI 7-8 is super fast and that isn't as true in certain segments. ]

iii. Input/Output standards that Thunderbolt doesn't directly support. legacy and future audio . legacy and future video . etc. ( a focus on legacy , 'sunk cost' cards is somewhat what appears to be Apple's no hurry to get new Mac
Pros released. Part of the product marketing is looking 'backwards' in time as opposed to forward. )
[ there is overlap here with an. xMac Studio enclosure wrapped around the mac Studio. ]


iv. AI/computational inference ( not video out GPU cards that happen to do AI Inference, just straight inference.)_ Computationally could build back in ECC compute coverage if offloaded here. )
[ again Apple is lacking on some foundational driver work here. So software holes to fill.
Can compromise without giving up the alignment of the SoC with the rest of the apple products.

At the moment some inference is heading to the 'infinity and beyond' power consumption levels.
Apple got bit before with more reasonable power projections on add in cards.
There is an option that they can take it back into their own hands though if they wish. ]

except for M.4 SSD storage most of those mean 'more wires'.

PCIe v4, v5 , v6 , v7 keep progressing and Apple appears to be drifting toward Thunderbolt 5 is really good zone; we can hit the snooze bar for years. For the laptops that is probably OK. For the $6K desktops that seems a bit dubious over a very long term.


4. Folks who run multiple instances inside of one 'box'.

i. multiple instances probably drives more than one storage drive and/or faster networking to boot from.
[ might be good to put the occasionally psychotic, hallucinating agent into more than just a 'software' jail to contain it when things go very sideways. ]

ii. better virtualized hosting. Direct mapping of cards/resources that Apple want to punt on in the mainstream macOS
but can be utilized by other OS that want to put do the work.
( Apple needs lots of work here software wise. And agin not the x86 is insanely great crowd. )

iii. the folks who need to keep an partially sealed, VM time capsules of old hybrid macOS around for apps just can't replace.

Other Macs could do this at limited scale also, but it would be a matter of "more" virtualization and containers.


If Apple picks a 2, 3, 4 year frequency and sticks to the schedule that would significant help also. Mac Pro has folks exiting because there is little predictable about what Apple is doing. That is mostly not a 'hardware' or 'software' thing that Mac Pro is missing. The volume is going to be too low for yearly. It is probably too low for every 2 years.

(***) This slower pace is won't be a bragging rights box in most years.

 

[Nugget]

I'm really struggling to imagine there's more than a handful of users out there who want a rack mount machine that can also support a large number of direct-attach displays. Also skeptical that there's any need at all for local spinning disks, especially in your hypothetical box that also has "better than 100GBE" networking. Even at 10gbit networking you've completely removed any incentive to have slow local storage, ignoring the lack of large filesystem support in macOS.

It's hard for me to see much of any market for the machine you're describing here, much less view it as the "core Mac Pro market."

Thanks for sharing your thoughts, either way. Always interesting to hear what other people are dreaming up.