Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Maybe not ~2013... when all Apple really had to do was make an official Hackintosh.

...but now, the choice is between focussing on Apple Silicon SoCs optimised for mobile and SFF, versus a PCIe-based Mac Pro workstation aimed at a shrinking market niche which meant throwing out most of the USPs of Apple Silicon. It's not 2013 any more.

Doesn't really matter. They made the M2 Mac Pro which has PCIe slots/lanes and no MPX module support. They just needed a new SKU for the SoC for RAM/GPU expansion. Clearly they don't care about that super niche market and it would probably cost them a lot of money to figure that out. There are plenty of ARM processors out there with PCIe, RAM and GPU upgradeability, so it's more about the lack of trying from Apple's part.

What doesn't make sense to me is that they knew in 2019 that Apple silicon was in the pipeline and they still made that machine. It was basically obsolete by 2022 in many ways performance wise due to release of Mac Studio (with Ultra chip). A 3 year timeline is worse than the 6 year gap between 2013 MP and 2019 MP.

Maybe they were forced to do it because of some kind of contract they had with Intel to buy their Xeons....
 
Maybe not ~2013... when all Apple really had to do was make an official Hackintosh.

...but now, the choice is between focussing on Apple Silicon SoCs optimised for mobile and SFF, versus a PCIe-based Mac Pro workstation aimed at a shrinking market niche which meant throwing out most of the USPs of Apple Silicon. It's not 2013 any more.

exactly

apple silicon is just an entirely different thing. it's entire design is based around everything sharing memory on a single SOC
 
How were MPX modules backwards compatible when GPUs and drive carriages for example had their own power supply lane? (non standard). You can't take an MPX GPU and put it in another PC.

Sorry, yes, I was referring to the slots, not the cards themselves. The cards were obviously proprietary to the 7,1 (they were dependent on its airflow too).

They got rid of MPX in the M2 Mac Pro for probably this reason.

MPX was intended for GPUs. Given the M2 didn't support PCIe GPUs at all, it would not only have been pointless, it would have caused confusion / disappointment if people had tried fitting MPX modules.

It's not interest, just over-engineered for no reason whatsoever.

I meant Apple lost interest in the 7,1 full stop. They had no incentive to beef it up with new GPUs, when they wanted its owners to move to AS Macs.

They could've at least gave us a 9xxx series drivers as a last hurrah for a dead product years ago but they didn't.

Yes, I agree 7,1 owners were shafted. I gave Apple the benefit of the doubt above, when I said perhaps the 7,1 took longer than expected to develop, and the conditions were right for the AS transition earlier than expected.

More realistically, Apple were well aware of the transition and decided to build the most powerful Intel Mac Pro they could, to tide pros over until the Mx Ultra was ready to take its place. Whilst AS CPU cores were performant from day 1, it took the Ultra a few generations to catch up to high end PCIe GPUs (and still can't match Nvidia's best). Plus, there would have been some concern regarding UltraFusion scalability and yields - the 7,1 would be a hedge against these.

Either way they lost a lot of customers in this niche product space after the 2013 debacle, and then by then the 2019 wasn't a feasible option. Most moved to Windows if they were doing anything 3D related for example.

I agree. Forcing an unpopular form factor, then sitting on it for years with no upgrades and no communication, understandably torched Apple's credibility in the workstation space. It probably was too late by 2019 - why would anyone trust them? Whereas everyone else has published, long-term roadmaps, Apple is famously tight lipped. Fine for consumer products, but not when people are trying to plan business purchases.

They just went "too pro" on the Mac Pro. Should've just made the case and made the whole thing as affordable as possible.

They went too expensive. The 7,1 was a luxury product. The case alone would probably cost $1K. It seems like since the 5,1, Apple can only be bothered with the Mac Pro if they have the chance of a design award. They refuse to just build a sensible workstation, like Lenovo or Dell - or like Apple used to.

The big issue on Apple silicon is the on-die RAM/GPU, they probably saw no need to make a new SKU to support RAM slots and GPUs, they would lose performance anyway losing data on the logic board's tracers.

It's not just the silicon engineering. It's that software would need to cater for two memory paradigms. Perhaps the OS could make this transparent. Even still, you would likely wind up with situations where a MacBook Neo was faster than a Mac Pro.

The G4 Cube was also Steve's but it was never supposed to replace the Mac pro desktop variant. It was underpowered and just mostly aesthetics. The G4 Cube is somewhere between a Mac Pro and a Mac Studio. 2013 is not his baby.

It's hard to say where the Cube might have gone if it sold well. It has mountings for a base fan, the VRM has unpopulated spots for additional phases, and hobbyists have squeezed up to dual 1.8GHz chips in them I believe (from a single 0.5GHz). It could be specced with a GeForce 3, which was a respectable card in its day. Still, it had no expansion slots, only one drive bay, and only space for a short single-slot GPU. In todays terms, its perhaps a base Mac Studio.

After Steve passed, Apple felt some pressure to demonstrate they could still innovate. Schiller infamously unveiled the 6,1 as a riposte, not as the machine Steve had wanted to build.

All Tim Cook did was streamline the Mac product line more and gave birth to the death of the Mac Pro.

Cook famously does his work on an iPad (as CEOs can) and isn't really a computer enthusiast. I doubt the Mac Pro looks particularly exciting on the revenue spreadsheet, compared to say, sales of Lightning cables. He may also have been trying to do the 'visionary' thing - moving to the future that others can't quite see yet. Where desktop computers are dead, and we're all just swiping on glass (or God forbid, manipulating objects in AR).

What doesn't make sense to me is that they knew in 2019 that Apple silicon was in the pipeline and they still made that machine. It was basically obsolete by 2022 in many ways performance wise due to release of Mac Studio (with Ultra chip). A 3 year timeline is worse than the 6 year gap between 2013 MP and 2019 MP.

Yeah, it's almost like Apple can be a**holes.
 
Doesn't really matter. They made the M2 Mac Pro which has PCIe slots/lanes and no MPX module support.
Apple Silicon only has a fraction of the PCIe lanes/bandwidth that you’d get in a Xeon/AMD system, OK for specialist I/O or A/V cards (probably why the 2023 MP was made) - not enough to run multiple GPUs (which is what most people mourning the Mac Pro seem to want).
They just needed a new SKU for the SoC for RAM/GPU expansion.
No, it would need a re-designed Apple Silicon die with at least 32 extra PCIe lanes, DDR5 RAM interface in place of LPDDR5x… which was only useful in the Mac Pro (unlike the Mx Max which is used in the MBP and up & doubles up to make the ultra) and would throw away the advantages of fast communication between RAM, CPU, GPU and NPU that give Apple Silicon its advantage.

If you want that sort of “big box’o’slots” system, Intel/AMD is the tool for the job. Apple are right to get out of the market - which is getting eaten at both ends by flexible cloud computing at one end and increasingly capable mobile/SFF systems (where Apple Silicon has an advantage) at the other.
.
 
MPX was intended for GPUs. Given the M2 didn't support PCIe GPUs at all, it would not only have been pointless, it would have caused confusion / disappointment if people had tried fitting MPX modules.

Not necessarily. Pegasus R4i is a good example. While a good idea, no manufacturer did anything after these initial MPX releases. It takes too much effort to make these things.

More realistically, Apple were well aware of the transition and decided to build the most powerful Intel Mac Pro they could, to tide pros over until the Mx Ultra was ready to take its place. Whilst AS CPU cores were performant from day 1, it took the Ultra a few generations to catch up to high end PCIe GPUs (and still can't match Nvidia's best). Plus, there would have been some concern regarding UltraFusion scalability and yields - the 7,1 would be a hedge against these.
None of the AS Ultra's even touch anything NVIDIA has. They barely touch the 9070 from AMD. So in reality, even AS Mac Studios are not good for heavy 3d users for example, most of them moved to Windows. On the other hand, if they allowed GPUs and NVIDIA support with AS + PCIe slots, that's another story. A lot of people would be pleased and they wouldn't lose those customers. It's probably just a big technical undertaking for them to create GPU/PCie switches and they'll lose performance if they create a new AS SKU just for that.

The reason AS is so powerful is because of unified memory and shared on-die high speed RAM and controllers. They focused on AI vs raster performance.

They went too expensive. The 7,1 was a luxury product. The case alone would probably cost $1K. It seems like since the 5,1, Apple can only be bothered with the Mac Pro if they have the chance of a design award. They refuse to just build a sensible workstation, like Lenovo or Dell - or like Apple used to.

I've been in the professional field forever (super high end stuff), and I remember the keynote for the 2019 MP and even I was baffled. I thought to myself who is this thing for, studios with hundreds of millions in budget. What made the older MPs successful was that high end studios and individuals (even prosumers) would be able to afford this thing and enjoy the same performance, but they went too far. Outside of the $999 wheels the whole thing was a joke. I basically got 3 years of use out of the Mac Pro, I can continue using it but it already feels dated in comparison with AS. They focused on businesses which lease these products and replace them every 3 years, which I think was a big mistake.

It's not just the silicon engineering. It's that software would need to cater for two memory paradigms. Perhaps the OS could make this transparent. Even still, you would likely wind up with situations where a MacBook Neo was faster than a Mac Pro.

It's doable, there are many workstation class ARM based systems out there with support for PCIe and GPUs. It's not totally a technical limitation, I just feel like Apple sees no upside in doing this.

If you look at the Mac Pro, even the older ones, there's PCIe lane switches all over the logic boards, these things are all doable. They can send video signals with MUX (which they used to do with iGPU + dGPUs on MacBook Pros). They can allow a PCIe GPU signal to pass through the iGPU without issues, they have the capability. What they could've done was have the iGPU (on-die GPU) and allow PCIe GPU through PCIe 5.0 (which is very fast btw) as needed.

After Steve passed, Apple felt some pressure to demonstrate they could still innovate. Schiller infamously unveiled the 6,1 as a riposte, not as the machine Steve had wanted to build.

Joke of an event tbh.


Cook famously does his work on an iPad (as CEOs can) and isn't really a computer enthusiast. I doubt the Mac Pro looks particularly exciting on the revenue spreadsheet, compared to say, sales of Lightning cables. He may also have been trying to do the 'visionary' thing - moving to the future that others can't quite see yet. Where desktop computers are dead, and we're all just swiping on glass (or God forbid, manipulating objects in AR).

I don't think it's that, he came from Compaq which was mostly desktops and some laptops. He knows his sh*t. He's just a master at streamlining product lines and manufacturing. Steve knew that foundationally if you cater to very specific target markets, you also elevate the brand they become forever Apple users and buy their other products, even if that product doesn't sell like hot cakes.

They could've kept the Mac Pro alive even if its 1% of their sales by at a minimum just adding Max and Ultra chips and keeping the same form factor, and allowing different options for RAM now that their SoC is semi-stacked. They could've went the extra mile and added GPU support if they trully wanted and collaborated with NVIDIA. It will take a long time for them to catch up to NVIDIAs software/raster hardware performance.

They are a 4 trillion dollar company with lots of cash on hand ffs.
Yeah, it's almost like Apple can be a**holes.

That pretty much sums it up.

Maybe Ternus will reintroduce it, he's a hardware guy. But I'm not holding my breath been burnt with the Mac Pro twice in a row already. I went the Hackintosh route between 2015-2017 and it was a huge waste of time, thousands of hours wasted "perfecting" a system that did work fine for real worl, but required maintenance and headaches all the time. 2013 MP was too crappy for my workflow.

Apple Silicon only has a fraction of the PCIe lanes/bandwidth that you’d get in a Xeon/AMD system, OK for specialist I/O or A/V cards (probably why the 2023 MP was made) - not enough to run multiple GPUs (which is what most people mourning the Mac Pro seem to want).

They can increase the PCIe lanes. AS Ultra has 32 lanes while 2019 Xeons have 64 lanes.

No, it would need a re-designed Apple Silicon die with at least 32 extra PCIe lanes, DDR5 RAM interface in place of LPDDR5x… which was only useful in the Mac Pro (unlike the Mx Max which is used in the MBP and up & doubles up to make the ultra) and would throw away the advantages of fast communication between RAM, CPU, GPU and NPU that give Apple Silicon its advantage.

If you want that sort of “big box’o’slots” system, Intel/AMD is the tool for the job. Apple are right to get out of the market - which is getting eaten at both ends by flexible cloud computing at one end and increasingly capable mobile/SFF systems (where Apple Silicon has an advantage) at the other.
.

They can design a special SKU, they have 4 trillion market cap lol

All they have to do is continue their on-die RAM (if they don't want to have addon RAM slots), still do their on-die GPUs, but then allow the dGPU to send a video signal. They have done this before with Intel MacBook Pros.

I'm well aware that Windows has unlimited options, my point being if you are a macOS user and don't want to go to Windows at all (like me and a lot of professionals) you are going to have to work around Apple's Thunderbolt PCie workflow.
 
  • Like
Reactions: Flint Ironstag
Not necessarily. Pegasus R4i is a good example.

Does the R4i use the MPX slot for anything other than mechanical stability? Are the contacts even connected? It clearly wouldn't scratch the bandwidth of a PCIe slot. 4 HDDs would run fine off the 75W a PCIe slot can supply (except perhaps during spin-up, which could be staggered).

They focused on AI vs raster performance.

I'm not an expert, but I'm sceptical AS was designed with AI in mind. I think it just turned out to be a bonus of the unified memory model that you wind up with a GPU with access to 64GB+ of RAM. Which would otherwise be well into workstation graphics card territory ($$$).

Apple were very clearly caught napping with AI, whilst distracted with AR and electric cars. Most of their AI features are licensed from Google.

They can design a special SKU, they have 4 trillion market cap lol

Even if a Mac Pro wouldn't be great business, you could make a case for Apple spending some of their infinite wealth on giving the Mac a range-topping machine regardless. The issues would be a) the disruption of the unified memory model, used from top to bottom, across all Apple's products; b) Apple would need to demonstrate a serious, long-term commitment to the tower Mac before relevant professionals would even consider returning to the platform; and c) Apple would need to make friends with Nvidia.

Even now, many photographers are still burned by Apple dropping Aperture. Once lost, trust is hard to regain.

I went the Hackintosh route between 2015-2017 and it was a huge waste of time, thousands of hours wasted "perfecting" a system that did work fine for real worl, but required maintenance and headaches all the time. 2013 MP was too crappy for my workflow.

I'm well aware that Windows has unlimited options, my point being if you are a macOS user and don't want to go to Windows at all (like me and a lot of professionals) you are going to have to work around Apple's Thunderbolt PCie workflow.

At some point, you have to question the value of avoiding Windows. Getting Windows set up as you'd want it takes a bit of effort. But it's infinitely more practical than shaking your fist at the sky, cursing Apple for not providing hardware they clearly don't want to make, but is easy to order elsewhere.
 
They can design a special SKU, they have 4 trillion market cap lol
NB: Why do you keep throwing the term SKU ("Stock keeping unit") around as if it has some sort of significance?

They would have to design a new die - different from the Mx Max (2 of which form a Mx Ultra) - and find space on TSMC's already stretched chip production lines to manufacture it in tiny quantities even relative to the Mx Max. Chips made in small quantities like that are extraordinarily expensive because they have to pay of the huge fixed costs of design and tooling up.

Apple didn't get a 4 trillion market cap by throwing money on the fire.

They can increase the PCIe lanes. AS Ultra has 32 lanes while 2019 Xeons have 64 lanes.
Yes, by designing a complete new die at great expense.

Meanwhile, contemporary AMD Threadripper Pro and Intel Xeon-W processors offer 112 or 128 lanes because they're designed for big PCIe workstations.
All they have to do is continue their on-die RAM (if they don't want to have addon RAM slots), still do their on-die GPUs, but then allow the dGPU to send a video signal. They have done this before with Intel MacBook Pros.
(NB: The RAM is on-package, rather than on-die - but the memory controllers are on board and are designed for LPDDR5x, not DDR5).

The Intel mobile chips used in the Mac Pros had pretty weak integrated GPUs but provided 16 lanes of PCIe direct from the CPU deliberately designed for connecting dGPUs. Intel laptops did the switching thing so you could use the low-power integrated GPU on battery and switch to the more powerful dGPU when on power. Apple Silicon Macs dumped that by just having a far more power-efficient integrated GPU.

you could make a case for Apple spending some of their infinite wealth on giving the Mac a range-topping machine regardless
...but it would be a "me too" machine who's performance could never be better than whatever AMD and NVIDIA GPU-de-jour could provide (even assuming Apple could bury the hatchet with NVIDIA), only of interest to a dwindling market who won't/can't use Windows or Linux.

Apple's "flagship" is now the Apple Silcon Pro/Max MacBook Pro - because it can provide industry beating levels of performance - including local AI - for an ultraportable laptop. A Mac Tower with AMD/NVIDIA isn't going to stand out because you will always be able to build something with more cores, more GPUs (even an ARM system from Ampere etc.) using generic PC parts. Plus, the "big multi-dGPU slotbox" is probably the one section of the market where power consumption and thermals matters the absolute least.
 
A Mac Tower with AMD/NVIDIA isn't going to stand out because you will always be able to build something with more cores, more GPUs (even an ARM system from Ampere etc.) using generic PC parts. Plus, the "big multi-dGPU slotbox" is probably the one section of the market where power consumption and thermals matters the absolute least.

not only could you build something with more cores and more gpu's using generic PC parts, as long as macOS is supporting x86 still, you could run macOS on it!

then we're just back to being able to build your own upgradable Mac tower, which does what for apple exactly?

and once x86 support is gone, all of the code to deal with non-UMA and dGPUs is gone with it
 
Last edited:
...but it would be a "me too" machine who's performance could never be better than whatever AMD and NVIDIA GPU-de-jour could provide (even assuming Apple could bury the hatchet with NVIDIA), only of interest to a dwindling market who won't/can't use Windows or Linux.

Sure. It wasn't a serious suggestion (as the rest of my post should have made clear). More that if money were the only factor*, supporting a range-topping loss-leader shouldn't be out of the question for a $4Tn corporation.

*I realise it's not.
 
NB: Why do you keep throwing the term SKU ("Stock keeping unit") around as if it has some sort of significance?

It's a new SKU buddy lol (do you know how manufacturing works?)

They design a new/modified chip just for the Mac Pro and it becomes a SKU.

Yes, by designing a complete new die at great expense.

They can do it. They just won't because they're looking at their bottom line.

(NB: The RAM is on-package, rather than on-die - but the memory controllers are on board and are designed for LPDDR5x, not DDR5).

That's what I said lol

...but it would be a "me too" machine who's performance could never be better than whatever AMD and NVIDIA GPU-de-jour could provide (even assuming Apple could bury the hatchet with NVIDIA), only of interest to a dwindling market who won't/can't use Windows or Linux.

Apple is still hurt from the dead GPU issues in the 2010s with NVIDIA hence the end of their partnership. Also they are competing with them in the LLM space (just a different approach).

I'm not sure where we are headed with this conversation but you are reiterating literally everything I've been saying.

I'm done with this convo.
 
They design a new/modified chip just for the Mac Pro and it becomes a SKU.
Yeah, and it's the "design (and manufacture) a new/modified" chip bit that's expensive, not assigning a SKU code number and putting it in the inventory database.

That's what I said lol
Nope
All they have to do is continue their on-die RAM [my emphasis]
You explicitly said "on-die" (i.e. integrated into the actual processorchip) which is not the same thing as "on-package" (separate chips soldered on top of/alongside of the chip when it is put on it's carrier) (LOL, do you know how manufacturing works?)
 
They design a new/modified chip just for the Mac Pro

it's the "design (and manufacture) a new/modified" chip bit that's expensive

I'm guessing that they are going to do another round of "Ultra", and, if they do, I'm hoping that the design includes more I/O. Specifically, enough to support more/faster SSD internal to the, say, "Studio Ultra", and, enough extra to support several higher speed connections to card-oriented external PCIe 5.0 chassis'. I'm thinking 12 lanes of PCIe5.0 bandwidth per higher speed optical connection. Something like what Nvidia already ships for its connections to DGX Spark products. Those are QSFP56, but QSFP112 are also available.
 
Apple is still hurt from the dead GPU issues in the 2010s with NVIDIA hence the end of their partnership. Also they are competing with them in the LLM space (just a different approach).


Regarding the second sentence, that is not my impression at all.

To me it seems Apple is content to wait and see how the initial AI arms race will shake out, and just pick a strategic alliance with whoever is left standing. Probably the most sensible stance to take for them.


I'm guessing that they are going to do another round of "Ultra", and, if they do, I'm hoping that the design includes more I/O. Specifically, enough to support more/faster SSD internal to the, say, "Studio Ultra", and, enough extra to support several higher speed connections to card-oriented external PCIe 5.0 chassis'. I'm thinking 12 lanes of PCIe5.0 bandwidth per higher speed optical connection. Something like what Nvidia already ships for its connections to DGX Spark products. Those are QSFP56, but QSFP112 are also available.


I think with John Ternus as incoming CEO, this might be something Apple will do in the future, even if it is a net loss leader. But my understanding is that products have -on average- a two to three year lead time, so I wouldn't expect it to happen anytime soon.

Of course, by that time, the potential user base for such a product may have abandoned ship altogether.

Like I said before, I'd already be happy with a Mac Studio that offers a few M.2 slots.



edit:

As an aside, I'm unhappy with a lot of Apple's decisions, currently my main gripe is closing off the kernel to third-party audio developers and forcing them towards DriverKit. This has significant impact on stability and efficiency of audio streams at low latency, and makes the Mac less professional for audio as a result.

Apple is lucky that the competition (if you can call it that) is MS Windows.
 
Last edited:
You explicitly said "on-die" (i.e. integrated into the actual processorchip) which is not the same thing as "on-package" (separate chips soldered on top of/alongside of the chip when it is put on it's carrier) (LOL, do you know how manufacturing works?)

Who cares buddy you know what I meant.
 
But my understanding is that products have -on average- a two to three year lead time, so I wouldn't expect it to happen anytime soon.
Absolutely, but, I'm guessing that they are partly through that lead time and already have a design. In that case, they just have to decide whether or not to execute. The outcome could be the world's fastest single-SOC server, in the form of a Studio Ultra. With very fast single thread speed, and, more throughput than an AMD EPYC 9755. Not as much I/O, of course, but, at least enough to support state of the art A/V applications. It would be nice to see Apple push for more than just the next iPhone.
 
Apple is lucky that the competition (if you can call it that) is MS Windows

Yes

It’s a shame that their isn’t a real OS competitor

Edit:

I’ve stayed with Apple this whole time (on and off since system 7) for the OS

I’ve tried many windows and linux in between at various times

(I also like playing with different OS for kicks. Haiku, bsd, whatever. I still have a red star os VM 😉 )

Apple silicon has been a great payoff,

but it’s all the OS for me

That’s why I did hackintosh for so long
 
Last edited:
XP was why I switched to Mac!


When Intel Macs and Bootcamp were mature options (OSX Tiger), I made the switch. I’d been using Macs at work and with Bootcamp I didn’t need to discard all the -ahem- “evaluation” software I had accrued and relied on.

Bootcamp made the switch a no-brainer, but after about a year and a half I realized I hardly ever booted up in WinXP anymore, despite my initial reliance on the aforementioned software.
 
I didn't know anything about Mac's back then. But I sure did know how to tweak XP to run like a beast.😉
I thought I'd get a cheap secondhand Mac off eBay to see what all the fuss was about. Once I got OS X running on my $90 G4 Power Mac, and it was so much smoother, stable, and so much more proficient at multitasking, that my much, much faster, carefully built, overclocked XP PC got relegated to glorified games console.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.