Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
This whole Mac Pro discussion, on a higher level, showcases a potential long-term issue with current Apple hardware, that I had been thinking about a lot lately.

It is not merely an issue for the Mac Pro but for every Apple product line, including MacBooks.

While Apple silicon allows for amazing integration and with that comes high compute and access speeds, the fact that RAM and SSDs are now bundled into a package, which is no longer user upgradable or transferable, may end up biting Apple one day. And soon.
Especially if the SSD/RAM shortage due to high demand for AI data centers continues. And it looks like it will.

If one purchases a high-spec MacBook Pro today with 128GB RAM and 8TB of SSD storage - all that investment is basically turning into paperweights the moment a new hardware is purchased.

In the "old days" one could often transfer RAM and the HD from the previous device to the next hardware.
Just purchase new hardware with the lowest RAM and HD specs (or include none at all) and then install the previous hardware's RAM and HD into the new hardware.
Sure, the transferred old RAM and HD may not be as fast as the latest generation, but the former investment was preserved.
Especially with highest-end specs, like an 8TB SSD, it really hursts to not be able to transfer it over to the next hardware.

When Apple planned its move to integrated Apple silicon, over a decade ago, they probably thought that prices for RAM and SSDs will just continue to come down to the point that these will become cheap commodities, where it no longer matters much to throw the old ones away when new hardware is purchased.
I bet they did not see it coming that AI data centers would cause a sudden massive shortage.

Now imagine prices for 2TB, 4TB and 8TB SSD configurations will double in the next MacBook Air and Pro generation - something that many consider is likely to happen. Being unable to carry over one's current SSD to the new hardware is really going to bite financially.
Typically users are not able to downsize their SSD requirments. Data has a tendency to accumulate, rather than decrease. The typical case is that the next hardware needs at least the same if not more RAM and SSD storage space than the previous.


It feels a bit like a perfect storm brewing.
RAM and SSD prices are doubling while the worldwide economy becomes fragile, people losing their jobs or fearing they might - which all will likely lead to scaled down investments.

People won't be willing (or able) to pay the price for the same size SSD in their new hardware.
And at one point they will likely start to demand that they can once again carry over their old SSD investments from their previous hardware.
Which Apple silicon does not allow.

If Apple refuses to decouple RAM and SSDs, they might ultimately start losing their higher-end markets.
But that's where all their high margins are...

What do other people think?

The ability to retain investments in RAM and SSDs might become a really important argument for Intel/AMD PCs in the near future.
storage is on cards still tied to the host but you can move them somewhat with an full wipe and reload.

apple can add an m.2 slot to macs for an 2th disk but that will hurt there mark up.
 
This whole Mac Pro discussion, on a higher level, showcases a potential long-term issue with current Apple hardware, that I had been thinking about a lot lately.

It is not merely an issue for the Mac Pro but for every Apple product line, including MacBooks.
[...]

What do other people think?

The ability to retain investments in RAM and SSDs might become a really important argument for Intel/AMD PCs in the near future.
I don't see how this really matters. Most users don't migrate their components forward (even among enthusiasts, at most I see a PSU or GPU making the jump.) Unless you upgrade regularly sockets change, etc and your components won't be actually transferrable. More to the point even before the Apple Silicon transition laptops had basically been locked since the retina models outside of some very specific options, and that was mostly the case on most of the desktops as well. I don't think this really is a large market segment that Apple cares about catering too, and I don't think they're going to be inordinately harmed versus other OEMs in the ongoing memory crush.
 
  • Like
Reactions: JayKay514
I don't see how this really matters. Most users don't migrate their components forward (even among enthusiasts, at most I see a PSU or GPU making the jump.) Unless you upgrade regularly sockets change, etc and your components won't be actually transferrable.
Excellent point. I worked in computer retail for years, and that sort of "salvage upgrade" thing happened very rarely, and if it did, it was only the thin niche of high-end PC gaming users who had invested more in custom builds.

Remember also that since the 90s, the vast majority of brand-name PCs do not use off-the-shelf motherboards or other components either; their stuff is custom-built and optimized to hit certain price points. Things were so, so not standardized, and in many cases still aren't (for instance, if you buy a Threadripper Pro ThinkStation from Lenovo, its motherboard is very nonstandard and you can't upgrade it as easily as one built with commodity parts).

In any case, I think it's important to remember there's the phenomenon of psychological projection, in the sense that the problems of a statistically tiny niche of users are not the problems of the large majority of users, yet that tiny niche still believes that to be true.

"If I can't do [niche, challenging computational task] then Macs are useless to everyone" is not a logical argument. Yes, it's an argument that maybe the Mac platform no longer caters to your specific needs and you should look elsewhere -- which is fine! Macs don't do everything for everyone.
 
  • Like
Reactions: BSDnostalgia
I don't see how this really matters.
Simple:

In a time when people need to tighten their belts, when businesses have to reduce expenses to the minimum, it hurts a lot if one needs to spend $4,600 extra on every new Mac Studio - just in order to retain the 16TB SSD.
It hurts equally as much if you need lots of RAM (e.g. for AI LLMs) and have to spend $4,000 extra on every new Mac Studio - in order to retain the 512GB RAM.

In the current economic climate it makes more sense to allow these super expensive SSD and RAM investments to be transferable to the next Mac Studio - as was possible with Macs 10-15 years ago.

Even if this transferred SSD or RAM is a tiny bit slower than the very latest generation would be, saving up to $8,600 on the asking price of a new Mac Studio makes a whole lot of difference these days.
Not every business has that kind of "spare change" lying around these days.

Sure, for the Average Joe it may not amount to a $8,600 saving, but even if it is "only" $1,500-$2,500 savings by not having to repurchase the SSD and RAM - that is a massive saving in this day and age for most people and businesses.

I am sure Apple can find a technical way to decouple SSDs and RAM from their Apple Silicon packages so that people can reuse these chips in their next Mac while retaining their fast connections.
 
Last edited:
Simple:

In a time when people need to tighten their belts, when businesses have to reduce expenses to the minimum, it hurts a lot if one needs to spend $4,600 extra on every new Mac Studio - just in order to retain the 16TB SSD.
It hurts equally as much if you need lots of RAM (e.g. for AI LLMs) and have to spend $4,000 extra on every new Mac Studio - in order to retain the 512GB RAM.

In the current economic climate it makes more sense to allow these super expensive SSD and RAM investments to be transferable to the next Mac Studio - as was possible with Macs 10-15 years ago.

Even if this transferred SSD or RAM is a tiny bit slower than the very latest generation would be, saving up to $8,600 on the asking price of a new Mac Studio makes a whole lot of difference these days.
Not every business has that kind of "spare change" lying around these days.

Sure, for the Average Joe it may not amount to a $8,600 saving, but even if it is "only" $1,500-$2,500 savings by not having to repurchase the SSD and RAM - that is a massive saving in this day and age for most people and businesses.

I am sure Apple can find a technical way to decouple SSDs and RAM from their Apple Silicon packages so that people can reuse these chips in their next Mac while retaining their fast connections.
This is fundamentally asking for Apple to produce worse products for most consumers. A lot of Apple silicon's advantages goes from the SOC setup. They weren't willing to figure out multiple computer GPUs or anything to make the Mac Pro more viable as a product, I don't think they're going to radically rethink their entire output due to memory shortages.

Beyond that, if the price is too high to upgrade to a new computer... why would businesses be buying new computers? What's more price-conscious than buying a lower-specced computer and transferring upgrades is just keeping the old computer running longer. Very few workloads out there actually require the latest and greatest tech.
 
  • Love
Reactions: Charlie Bonesx
that sort of "salvage upgrade" thing happened very rarely,

It happened often out in the field, though. And, companies using standardized PCI cards really facilitated it.

Remember also that since the 90s, the vast majority of brand-name PCs do not use off-the-shelf motherboards or other components either; their stuff is custom-built and optimized to hit certain price points. Things were so, so not standardized,

I recall stuff from Dell and stuff from the corner build-to-order place were pretty interchangeable back in the 90's. By 2010, you had to get a built-to-order tower to get commodity stuff, but, it was still there, e.g. with SuperMicro boards you could build anything from workplace servers to nicer PC configurations for gamers. So, I'm not really agreeing with this part of your argument. For about 15 years, there was a lot of interoperable stuff in the tower/server space. The biggest issue for gamers was getting enough power and cooling to those really hot GPUs that wouldn't work in a commodity PC tower.

In any case, I think it's important to remember there's the phenomenon of psychological projection, in the sense that the problems of a statistically tiny niche of users are not the problems of the large majority of users, yet that tiny niche still believes that to be true.

I agree mostly. I don't think it is that there is a single niche that is so small. Rather, I think the commonality among several niches that existing 1997-2012 or so fragmented. Something optimized for video production is not that similar to what gamers want anymore. CPUs with fewer but faster cores, that work for the general public and gamers, are optimized differently than CPUs for "bulk" cloud providers with 160 threads. What was once a single market for consumers, creatives, gamers, and business, has broken up again into more specialized markets. (Except we are all sharing a common market for RAM and flash storage!)

"If I can't do [niche, challenging computational task] then Macs are useless to everyone" is not a logical argument. Yes, it's an argument that maybe the Mac platform no longer caters to your specific needs and you should look elsewhere -- which is fine! Macs don't do everything for everyone.

I think Apple, and, "creatives", seem to be "divorcing" unnecessarily. I think Apple could support an Xserve/Xserve-outboard-chassis arrangement that would not compromise the essential "Studio Ultra" model but would support outboard storage and outboard PCIe-slots for I/O and computation devices, including even GPUs used for computation (but not graphics output directly). To connect the main "Studio" chassis with the outboard chassis would require something like this cable/QSFP pair:
https://www.naddod.com/products/102069.html
connecting to a chassis something like this with 2-4 slots with 8 say PCIe 5.0 lanes
https://www.naddod.com/blog/naddod-dac-cables-for-nvidia-dgx-spark-connectivity

It should be possible to use AMD and Nvidia GPUs, just without the direct-to-screen graphics. Apple could support this without too much cost, as long as the creatives don't insist on GPUs for direct graphics output so they can game in their spare time.

"Power" gamers are now a different market at this point, and, I don't see how the old gamers x86 niche can make sense for Apple at this point. Game software companies will selectively port some stuff to Apple Silicon, but, the big single software market is diminishing anyway.
 
It happened often out in the field, though. And, companies using standardized PCI cards really facilitated it.
True. That said in the mid-90s at our store we were selling Compaqs and Aptivas and other PCs. In Canada at least, Dell wasn’t selling via retail, just direct.
I recall stuff from Dell and stuff from the corner build-to-order place were pretty interchangeable back in the 90's.
I don’t doubt it, but when I moved up to corporate in the retail computer company, we did research that showed Dell would often swap out different models when fulfilling large corporate orders which (and this is the Win95 era) often caused issues with batches requiring different drivers, creating conflicts. (Remember IRQs?)
I agree mostly. I don't think it is that there is a single niche that is so small. Rather, I think the commonality among several niches that existed 1997-2012 or so fragmented.
Great point. I think we’ve now hit the point where most devices natively handle the common multimedia tasks that used to strain older devices - video, audio, light 3D rendering. We used to have to buy DSP accelerator cards for ProTools or UAD plugins, multi-SSD PCIe cards, or Blackmagic or Afterburner stuff for 8k video etc.

So the market for a big box with a fast CPU and lots of expansion has shifted to different niches, whereas we would have all needed Mac Pros before.
I think Apple could support an Xserve/ Xserve-outboard-chassis arrangement that would not compromise the essential "Studio Ultra" model but would support outboard storage and outboard PCIe-slots for I/O and computation devices, including even GPUs used for computation […]
Apple recently approved a driver to allow Nvidia eGPU use for running LLMs in Docker VMs, so that’s something.

The rest kinda exists - there are definitely Thunderbolt PCIe expansion chassis for cards and drives, though none with that nice Apple design.
“Power" gamers are now a different market at this point, and, I don't see how the old gamers x86 niche can make sense for Apple at this point. Game software companies will selectively port some stuff to Apple Silicon, but, the big single software market is diminishing anyway.
I don’t think it was ever Apple’s goal to compete with the big PC / console gaming giants, but to create and dominate their own market for phone and tablet gaming.
 
  • Like
Reactions: BSDnostalgia
This is fundamentally asking for Apple to produce worse products for most consumers. A lot of Apple silicon's advantages goes from the SOC setup.
But really?
That is what Apple tells us and wants us to believe - so we are forced to pay thousands of $ more than we would actually need to.
I am not convinced that there isn't a technical approach and solution that could satisfy both goals: the SOC integration yet still make RAM and SSD upgradeable / transferrable.

The fact that some people successfully upgrade SSDs in MacBook Neos and Mac Studios proves that at least that procedure is technically possible. Apple could easily invent a user-sevicable socket mechanism for these. They just do not want to. For purely profit margin reasons. Not because it is technically impossible, or would downgrade performance.


why would businesses be buying new computers?
Because your hardware and software manufacturer and OS provider cleverly builds-in "planned obsolences" into their products - forcing people to buy new hardware.
Either by dropping security updates (which businesses are required to be up-to-date on these days, one cannot skip them as one would open oneself up to liabilities in case of a hacking incident).
Or by requiring certain new "security" features in new CPUs in their latest OSes, which are then required by some software vendors too.
Or by just not supporting browsers of a certain age anymore.

Even if your hardware is fully capable to do what you want it to do, but if your browser version is no longer supported by most websites, or if your OS no longer receives any security updates and an OS update requires a new "more secure" CPU, or if one of your key softwares requires certain CPU "security" features - as a business you are forced to upgrade your hardware.
Whether you actually need it or not.


You just think the customer is in control.
Not really.
The software, browser and OS manufacturers are in control.
They indirectly dictate when people have to purchase new hardware.
And Apple on top also technically - or conveniently? - requires people to re-purchase SSDs and RAM each time anew (at inflated Apple prices aka super high Apple profit margins). Coincidence? Or deliberately engineered to work in Apple's favor like that?
 
Last edited:
So the market for a big box with a fast CPU and lots of expansion has shifted to different niches, whereas we would have all needed Mac Pros before.

Exactly.

And, thank you for pointing out the Nvidia eGPU link. I had not seen that.

Apple recently approved a driver to allow Nvidia eGPU use for running LLMs in Docker VMs, so that’s something.

The rest kinda exists - there are definitely Thunderbolt PCIe expansion chassis for cards and drives, though none with that nice Apple design.

It does exist, although, to really support full use of that type of GPU, you are going to need a lot more bandwidth. That is why I would suggest 400+ Gbps QSFPxx links such as I referenced. Those are off the shelf in the network world. And now, being used already for chassis-chassis links. And, from what I have read, the I/O capability is already in the previous Apple Ultra chipsets. I think Apple could put a couple of QSFP slots in a Studio without upsetting the cart.

I don’t think it was ever Apple’s goal to compete with the big PC / console gaming giants, but to create and dominate their own market for phone and tablet gaming.

I agree. But, a number of people still think in those terms. e.g. In the article that you linked to, the very first comment started out, "Although this is a far stretch from being able to use eGPUs for gaming ... " Followed by a discussion similar to those in MR. Some folks just can't seem to let that go.
 
Apple has become a major consumer brand but it willingly gave up big parts of the pro segment it used to own.
In the very old days many creative industries, graphic designers, scientists, music, and even accountants had (back then) high cost Macs. Both for performance and prestige. I bought my first Classic II (It came with a heavy duty carry on back and was considered to be portable!) back then second hand from a big consulting company that had just switched to "IBM" hardware. This separate pro market as a high paying niche for extra capable hardware might just not exist anymore or not be relevant compared to consumer mass market revenues.

Apple had several hardware revolutions over and over again from 68K to Power PC and Intel to Silicon (plus changing connectors) that might have hit the pro people harder that need specialist software to work. They had to buy stuff new every time again and again. Many moved on to Windows or Linux.

Forgetting about expansion slots over the quiet airflow design was a big pro mistake Apple made, likely the final nail in the coffin for the Pro line. Practically, stock MBP are very capable and fast and cover pro uses well. To me it feels like more people use Apple again after the iPhone and Apple Watch ease of use convince them to try.
 
I think Apple overdid it with the Mac Pro. Too expensive. The idea of having a tower with extensions is great, but how many PCI do you need when you have Thunderbolt. With the rise of AI and Mac mini, Studio sucess, i think offering a more affordable smaller unit (2U max), for homelab, small businesses would make sense. I don't see small business running Mac Mini or MacStudio. It just look too small. Something you can rack on a server rack makes more sense.
Lower the PCI, and boost the GPU, Ram, and disk options. (2.5 SSD, NVme). MacStudio was meant for the "Studio".
 
I think Apple overdid it with the Mac Pro. Too expensive. The idea of having a tower with extensions is great, but how many PCI do you need when you have Thunderbolt. With the rise of AI and Mac mini, Studio sucess, i think offering a more affordable smaller unit (2U max), for homelab, small businesses would make sense. I don't see small business running Mac Mini or MacStudio. It just look too small. Something you can rack on a server rack makes more sense.
Lower the PCI, and boost the GPU, Ram, and disk options. (2.5 SSD, NVme). MacStudio was meant for the "Studio".
All my PCIe slots are full, and I have a full PCIe expansion chassis.
 
how many PCI do you need when you have Thunderbolt.

Am I the only one in the world who hates noisy external boxes that do not even (color) match the Mac anymore, let alone add tons of cables and power bricks everywhere?

Sure, you can do a lot with Thunderbolt.

But chances are, these external boxes are black painted or black plastic, are really cheap looking and not at all matching the silver Mac Studio or silver Mac mini.
Then they usually come with really cheap fans, that are not temperature controlled and spin on high speed (meaning high noise level) 24/7. They should come with ear plugs.
Add to that, that these boxes often use external power bricks, which are not also ugly but just lay around always in the way with their cheap looking cables which do not match the nice (often braided) Apple cables, and which all require extra power strips and other unsightly stuff that just collects dust and is a nightmare to clean.

Brrrr.

Who wants that?
I do not.

Having had a Mac Pro I can attest that this was a much better solution.
Everything was internal in one box. Ahh!
No cable salad. No extra power bricks. Just one cable and the Mac Pro.
And the Mac Pro's fans were very efficient and temperature controlled, most of the time you could not hear them - as opposed to my current external disk drive enclosure which sounds like a jet engine...
And if you look for silver colored external expansion boxes or disk enclosures ... good luck! Most are either only USB-C (not Thunderbolt) or they have really poor reviews, or both. You're basically stuck with buying ugly, noisy, black boxes with cheap looking cables and power bricks next to your shiny, nicely designed silver Mac Studio or Mac mini...

Why is this never a concern for anyone?

Is everyone really totally fine with ugly, external crap lying around everywhere in a bed of spaghetti cables and power bricks that are a nightmare to dust?
I am not.

I want a Mac with at least space for two internal 3.5" HDs (for large Time Machine backups and a media/data disk) and at least one or two PCI slots for internal expansions.
I don't want a plethora of ugly, noisy, black, external Thunderbolt connected boxes with cheap looking cables and power bricks cluttering my desk area.
 
Last edited:
You aren’t the only one.

But obviously there weren’t enough of you actually buying Mac pros to make it worthwhile for Apple to keep selling them

The bar is rather higher on the Mac platform compared to x86, though. If Dell had to fund Xeon CPU development themselves, rather than buy them off the shelf, they might not bother with workstations either. A PCIe-heavy AS chip, especially one that could take GPUs, would be largely bespoke for one Mac model.

A downside to using your own architecture rather than an industry standard. Obviously, there’s lots of upsides too.
 
The bar is rather higher on the Mac platform compared to x86, though. If Dell had to fund Xeon CPU development from scratch, rather than buy them off the shelf, they might not bother with workstations either.

A downside to using your own architecture rather than an industry standard. Obviously, there’s lots of upsides too.

it’s not about cpu development. The same cpu’s are in the studio

The upside of not using the “industry standard” x86 is that they don’t need a giant tower full of fans, a separate gpu (with its own bank of fans) and a giant psu (with its own fan, to cool itself from powering all those fans 😉) to reach the same performance

Thus the only purpose of the fancy steel tower is for people that want internal expansion, which between thunderbolt and lack of discrete gpu support, is obviously a very small number
 
it’s not about cpu development. The same cpu’s are in the studio

Yeah, but the CPUs in the Studio have naff all PCIe lanes. You’d need something that does to make a tower worthwhile.

Plus, a tower would also need to support dGPUs too, or it wouldn’t have much appeal. Requiring significant architectural / OS changes.
 
Yeah, but the CPUs in the Studio have naff all PCIe lanes. You’d need something that does to make a tower worthwhile.

Plus, a tower would also need to support dGPUs too, or it wouldn’t have much appeal. Requiring significant architectural / OS changes.

pretty much exactly what I said
 
pretty much exactly what I said

Not really, since a Max or Ultra doesn’t match the GPU performance of a comparable PC tower.

Saying no one wants a tower is a bit meaningless if the only available tower is hopelessly limited (as well as very expensive).
 
Plus, a tower would also need to support dGPUs too

But, people wanting that capability is exactly what kills the Pro. If you just wanted an AS tower with PCI slots for a bunch of specialized A/V cards, no problem. But, they did that, and, few people wanted it.
 
But, people wanting that capability is exactly what kills the Pro. If you just wanted an AS tower with PCI slots for a bunch of specialized A/V cards, no problem. But, they did that, and, few people wanted it.

Probably talking at cross purposes a bit here. I was saying that making a tower workstation is non-trivial for Apple, since to do it properly would require a dedicated chip design. None of their other products, which are either laptops or headless laptops, have plentiful PCIe lanes or use dGPUs. I understand why Apple don’t bother - I wouldn’t either.

I was pushing back against the idea that Apple doesn’t make such a machine because no one wants one. Obviously they do, as the PC workstation market demonstrates. It just doesn’t suit their mobile-focussed system architecture, or ‘whole widget’ business model.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.