Obsolete
but the expandable desktop (and laptop, and server) computer is obsolete
It was always obsolete if you bought maxed out system.
Apple has discontinued the Mac Pro – but it's just the first of the tower computers to go. The rest will follow soon. Fruit-sniffers extraordaire 9-to-5 Mac got the news yesterday, complete with official confirmation from Apple itself. It's official and it's happened, but there have been warning signs for months – in November …
Nothing's wrong with ARM CPUs. On the other hand the 2023 Pro has non-expandable non-replaceable RAM, non-expandable non-replaceable internal storage, nobbled PCIe... The performance was essentially the same as the Mac Studio. They couldn't design an update to the Pro which overcame these problems.
It sure as shit does for the target demographic for the "old-school" tower Pro users.
But that also highlights the issue that Apple doesn't care what that user-base wants/desires, since even before Apple moved to ARM everything, they loved to "integrate" everything to up-market the pro consumers choices, and ergo, the prices they could charge.
Just look at how absolutely tone-deaf the "Darth Vader trashcan" turned out to be in regards to what Apple wanted it to be, versus what the "traditional" Pro users wanted. Thing was a flop even before it went on sale.
But then again, we're talking about the company that quit making rack-mounted servers, despite there still being a lucrative market for it, due to certain groups of people (at least then) loved to 100% stay inside Apple's software ecosystem.
"They couldn't design an update to the Pro which overcame these problems."
The price and need for a dedicated socket to plug it in are also issues.
Apple is extremely proud of their RAM and charge the moon for it so it was much better to get the base level RAM offering and buy more someplace else. The same goes for storage. When a 1tb drive upgrade from a paltry 256gb one is the same as a 4tb drive elsewhere, the cry of rape is well founded. Graphics is something that keeps improving so locking in one version at the factory that can't be upgraded is a dead end for many. Apple, in their quest for the highest margins, has designed themselves out of the professional market that will be looking at things like expandability. My 2010 cheesegraters still work very well and it's mainly the lack of OS support that makes them a bit twitchy here and there. Frankly, most software that is requiring Apple Silicon Macs isn't the big of an upgrade from their last Intel-Mac version. At least nothing that impacts my work so I'll stay with that older version of software that I know quite well and can get around in much faster as the new version has probably had a whole UI redesign which will force me to spend a few weeks trying to find where my old keyboard shortcuts and menu items have gone.
Mr iPhone has been too long in the big chair at Apple and hasn't seemed to have any attention to give to anything but their phone and online services. They make a load of cash on the iPhone, I get that, but they had also been dominating the computing market for creatives and if Adobe were to feature linux versions that tracked with Mac/Win, Apple would be in trouble. DiVinci Resolve on linux might really scream.
Until the M4 Max (not M4) came along, you could have a problem docking a MacBook to a dock with three or even two monitors.
So I admit that I took my eye off the ball and I'm a year and a half out of date. But as I said, the non-swapable storage and RAM mean I wouldn't consider them for real work because they can and do go wrong. Am I supposed to have external storage hanging off it like ZX Microdrives and hope bought enough RAM for the lifetime of the device for all the software I want to run now and in the future to work around this?
Until I read the news that the you can swap out the RAM or storage and the DR scenario is not replace the entire computer with another one I have on standby, I wouldn't consider doing work I value on it.
I only have one external monitor and personally find multi-monitor setups annoying, less ergonomic and less productive. But each to their own. SWMBO has a setup which allows daisychaining between monitors. This is a neat trick, and reduces "unsightly" cables, and not supported by any of my Macs, but I don't miss it.
"daisychaining between monitors ... not supported by any of my Macs"
What's annoying is that the hardware supports it, it's just that MacOS doesn't - or at least that used to be the case with my 2016 MacBook Pro. If I booted linux from an external disk then I could happily have two monitors driven from a single port, but MacOS refused.
Now I have a TB3 dock which has multiple video outputs, and that sits nicely with the M4 Mac mini to give me a decent triple headed setup.
"I only have one external monitor and personally find multi-monitor setups annoying, less ergonomic and less productive."
It depends on the sorts of things you do. I'd never go back to a single monitor on my media production machine. I don't have to switch layouts to pull up tools from a palette. With video and image editing, it's extremely handy. One big monitor might work, but I shudder to think what a giant monitor from Eizo would run me compared to the 2x 27" that I have now.
I'm typing on a 27" iMac which is fine for many other things and it's less power hungry than the production machine to have on. For accounting, I have a headless older Mini that I remote into from the iMac and have no need for a dual monitor setup for that.
Some people editing video prefer 3 monitors so they can have an easier time by having a main work screen and can glance one way or the other to find listings of the media assets without switching layers/screen setups. Stock traders want to have multiple screens and process monitoring demands it.
"I only have one external monitor and personally find multi-monitor setups annoying, less ergonomic and less productive"
You are then rather the odd bird out there. Everyone that I know, in quite a variety of professions, that is using a dual, triple even more monitor setup (I have set up trading desk computers with 8 monitors, on 2 quad-head graphics cards) would never, ever, go back to a single screen setup. THAT is rather annoying, less ergonomic and less productive. And no, one large 42" monitor, for example, is NOT the same as having 2 or 3 24"/27" monitors next to each others.
"But as I said, the non-swapable storage and RAM mean I wouldn't consider them for real work because they can and do go wrong"
And so can motherboards, cpus, and other things. When the lifetime of said components is sufficiently long it doesn't matter.
The RAM is non replaceable on many machines, and that enables it to be physically closer to (or even on) the cpu package, which allows it to be substantially faster.
I doubt you'd say that a 5090 isn't "useful for real work" because you can't upgrade or replace the RAM.
"The RAM is non replaceable on many machines, and that enables it to be physically closer to (or even on) the cpu package, which allows it to be substantially faster."
That's down to design choices. A bit of complicated packaging might force locating RAM somewhere that can have an access door. My iMac is super annoying that the taped on screen has to be removed to access anything inside so Apple could make it slimmer. That design choice doesn't benefit me at all. It could be 3x as thick and make no difference sitting on my desk. If it were thicker, maybe it could have had a standard VESA mount, which would be nice. I got the computer from a charity shop for $100 which is why I have it. If I had to pay more, I'd have gone with separates. Older Studios and Minis that can't be upgraded are getting to be very cheap second hand especially if they aren't maxed versions. If they were upgradeable, they'd be worth more.
RAM can be located very close to the CPU and processor designs could bulk up on-board cache. It's the same as storage, which may become the bottleneck. When I'm editing a video, the files are on a SSD for the best throughput. When I'm done, everything gets shoved over onto some spinning rust that's fast enough for requirements. For photo editing, I don't find that putting the files on an SSD gives me any more speed so I have cut out that in the workflow. I expect that it could be shown to be faster, but, in the real world, it's not buying enough to be worthwhile.
this has to be about the dumbest statement I’ve read on the register.
It’s clearly coming from a point of no experience or common sense from someone who likes to hate things because they think it’s cool or funny.
Of to slashdot with you, you’ll be welcomed by the grey beard linux trolls.
If your experience tells you to entrust everything to the locked-down device then go ahead.
People expected more from the Mac Pro and didn't get it so it didn't sell. The SoC design won't accommodate it and Apple couldn't or wouldn't improve the design to address these limitations, they would rather just kill the product line and sell more locked-down devices to be thrown away when RAM or storage is not enough any more.
So you throw away a computer when the RAM isn't "enough" (not that you consider what might be enough).
You might upgrade it by getting a later model, possibly with more memory, but you wouldn't throw away the previous one. It would either find a different use case for you, or you would sell it on so that someone else can continue to use the perfectly good machine.
The weakness here is the storage - that's the one thing where I do wish they at least used modules (like the Studio) across the whole range. Because I don't even care if the modules are NVRAM only or a complete NVME drive - but SSDs have wear limits, and so will need replacement at some point.
Having said that - I can't recall any Apple SSD failing in my use except for the one I massively abused by using it as a server cache drive for a couple of years - and if the SSD is going to last a decade or more... then do I really care that much, probably not.
I can accept soldered RAM, though grudgingly so. I can't accept soldered storage. Perhaps it's just my bad luck, but I've seen two SSDs fail on my own machines and, in my role as volunteer IT for some charities, storage is by far the most failure-prone component. It's also about neck and neck with the battery for the component that users most often want to upgrade when the rest of the computer is perfectly fine. For example, I was recently helping a user pick out a replacement for a Mac which, though some would probably consider it too old, the user didn't mind using except for the 128 GB of internal storage which was unsurprisingly confining, and if I could have offered her the option to continue using that as the main machine with a much larger disk, she would have taken that.
Taking a longer term view, if this trend of increased integration, and given the very fast interconnections that appear to exist now and in the future, and the ever increasing speed of WiFi connections, perhaps the market will come up with a self-contained storage extension pod that you switch on and either plug in or connect to *local* network (not cloud) and can use for slightly 'colder' storage of older material? Or does cloud storage crowd that niche out?
"perhaps the market will come up with a self-contained storage extension pod "
A laptop that's completely self-contained is the best solution. There's nothing else to remember to pack up other than the power supply. I'll often have external back up drives and such, but if I forgot any of that, I still need to be able to work. If my internal drive is getting squeezed for space, I want to be able to upgrade that. I usually max out my RAM as a first item, but doing that right at the beginning with Apple is horribly expensive and I don't buy new. It's as bad as buying a new car. The depreciation right off the lot is painful. I'd rather get something that suits my needs in most ways and upgrade RAM and storage if needed.
My cheesegrater is cable central and would be even worse if it didn't have all of the internal slots. It's been upgraded with USB3, a full install of drives (cd drives exorcised and SSD's put in their place), a PCI-NVME boot drive, Nvidia graphics card and probably one or two other things I can't remember. I could have gone up to 128gb of RAM, but 96gb is said to be faster due to the way it handles RAM addressing (or something like that). It has 12 cores of 3.46ghz processors. This didn't all happen at once or I'd still have a massive credit card balance. The older cheesegrater that I use as my media station isn't as upgraded, but certainly isn't stock. I might have bought one of the newer large hole cheesegraters, but the power draw would mean big money paying a sparky to run another circuit into the office.
"A laptop that's completely self-contained is the best solution."
...for you....
As recently discussed on another thread, it's not going to be long before the phone is literally all the computer most people need.
The A18 powered NEO is very clear evidence that there is no longer a significant performance penalty for the vast majority of users... I know it's M1 comparable and Apple just released the M5 ultra max pro wallet shredder... but in real terms, most people don't need that much power, memory, storage, connectivity.
Having a home NAS is not completely unreasonable - and there are some pretty fast NAS solutions out there - and with something like Tailscale then it's always available (albeit slower elsewhere).
You're also not then running the risk of the "next machine" coming with something newer and "better" than the nvme you currently use (heck, most of my bulk storage is still SATA)
As you mention - the cost benefits in terms of power efficiency are also not to be sniffed at.
Not for regular use, but I do use a Bluetooth mouse and keyboard with my iPhone connected to an HDMI screen with a USB AV Adapter. The system is OK as "a computer" and there is enough power with the USB-C to run a Raspberry Pi 2W "server" off the USB-A port for CLI stuff.
I argue that, if the internal disk is broken or inadequate, then an upgraded internal disk is, unless it has to cost a lot more for some reason, better for every case. There are some people for whom an external one is acceptable and some for whom it's not, but I don't think you can find anyone for whom it's better. The cost example does come up, for example someone who uses a cheaper spinning external drive and doesn't need all that space inside their laptop, but that's as close as you can get.
For example, let's take that person who bought a new machine because her soldered disk was too small. She needs more storage to store large scientific software packages. A network disk isn't going to be good there; pulling 6 GB of code and resources across a campus network or occasionally an ocean whenever she wants to open some research data isn't going to be enjoyable. Fine, an external SSD then. And loading that example would work for that, but not so well for another piece of software which isn't a self-contained application bundle. That one wants to put some stuff in a hard-coded path on the root volume, some things in /Applications, many things in /Library, and a few things in ~/Library. External disks don't hold any of those. I could try to set up something with symbolic links for this application, but that would be tricky, it would be fragile, and neither I nor she wants to have her run to me when something breaks one of those and important work can't be done and she doesn't know a lot about filesystems and symbolic links. That's why people benefit when disks can be upgraded.
People benefited when laptops had a fully functional spare 2.5-inch drive bay, whilst some had a 7.5mm capacity, a few could take a 15mm drive.
In the race to “slimness” and cost cutting, very few laptops permit any internal storage expansion, forcing users to carry a bag of hubs, cables and expansion drives..
"I argue that, if the internal disk is broken or inadequate, then an upgraded internal disk is, unless it has to cost a lot more for some reason, better for every case."
If the drive is broken then that's obvious.
From those I know who use Macs in academic labs... they have their own local NAS and high speed networking, because the data is too valuable to trust to a single machine disk (of any sort).
Don't get me wrong - I'd love to see an name slot, or even a package slot like they did with the studio - because nvram does inherently have a limited shelflife... but it's just not an issue for the vast majority of users.
We've got those. They're called external hard drives, external SSDs, or NAS. None of these are new, and they all work. There are problems with them in the cases I mention.
Let's take only the failed SSD example. I've had them fail before. One failed outright, the other started sending SMART errors at me so I replaced it out of caution. If these had been soldered in, then I'd have to use my laptop with something dangling off it for the rest of its life. If I forget that one day or the USB cable fails, the laptop is unusable until I get it back. If I drop the laptop, the external storage device is likely to be more damaged than the laptop itself would be. This isn't good. With a modern Mac, it's worse, because the internal SSD needs to function for the machine to boot. I can no longer use it with an external device if the SSD failed. If I ask Apple to replace the disk, they'll ask to replace the logic board, meaning I have to buy a whole new expensive M* processor and unified RAM even though both of those are still working. That's why, since I consider the SSD a potential failure component, I do not accept it being soldered in unless there's another place for me to install one internal to the machine.
"Let's take only the failed SSD example. I've had them fail before."
Disk failures have always happened... but they are an awful lot rarer now than they used to be.
And the replacement being a larger piece - might have made that piece cheaper to start with... and with failure rates several orders of magnitude below certainty that saving doesn't have to be much to be worth it.
I'd like to see replaceable nvram (at least, a full nvme would be easier) but that's not important for the vast majority.
If a 192GB in a Mac Pro isn't enough, and for the target market, it is utterly inadequate, then you choose chose something else like the ThinkStation P620 which can be upgraded to 2TB.
The Previous Intel Mac Pro could be upgraded to 1.5TB, which at the time was comparable to the competition.
"You might upgrade it by getting a later model, possibly with more memory, but you wouldn't throw away the previous one. "
I don't throw them away, but I might have to as the closet where they live is getting rather full.
By the time I upgrade to a new computer, my previous machine is worth so little just spending the time to sell it isn't worthwhile. When I'm given some computers, many of them can be fitted with more RAM and a bigger drive and be resold or donated. Anything that can't be upgraded and isn't factory maxed isn't of interest. The demands on hardware just keep going up and while a slower processor isn't ideal, not having enough RAM or storage is a deal killer.
I also like the ability to take the storage out of one computer if the computer dies and install it in another for a quick recovery. That can be much faster than restoring from backups.
This post has been deleted by its author
There's the option of specialised hardware for this - we're already seeing this with the rise of compute modules in industrial applications. I don't think it'll be long until we start to see CM mounts (to allow SoC upgrades while breaking out the I/O separately) become the standard approach to this sort of use case. It's already commonplace in e.g. the Raspberry Pi ecosystem.
Old MythTV boz:
1s Mini-ITX case, board, disk drive and optical drive containing a Huppauge tuner card. The tuner card has 2 tuners, each able to extract several digital channels from a single mux so that it can record half a dozen programmes at once provided they're on a couple of muxes.
1x Powerbrick (powerbriks don't have fans to using a motherboard that takes external power means I don't have the noise of a fan and the box is lower profile).
New MythTV system when I get round to finally putting it together (or it might be some other S/W).
1x Pi in its box
1x Powerbrick for the Pi
1x Silicondust tuner box containing 4 tunsers. AFAICS although each tuner can be on a different mux, so recording from 4 muxes at once, only one channel per tuner so no more than 4 channels at once.
1x Silicondust power brick
1x Netowrk hub - the Silicondust tuner is networked, not USB connected
1x Powerbrick for network hub - can I work out which is for the hub and which for the tuner as they both has the same size plug but different volteges
1x External drive for the Pi
Verdict: old - neat, new- a visual mess
"1x Netowrk hub - the Silicondust tuner is networked, not USB connected"
SO none of the rest of it needs to be anywhere near your TV.
And in fact this bit can go anywhere there is a signal.
So... You're going to be able to put that all in a cupboard/cabinet somewhere and see none of it.
"Verdict: old - neat, new- a visual mess"
Verdict: Complete lack of thinking about where to put things.
A TV signal tends to be available where the TV is, or vice versa, which amounts to 2 places in the house,, each with a TV there. The solution is likely to be a rat's of cables pushed behind some item of furniture, not good for reliability and not good for working on.
This approach to system-building is a lash-up. It doesn't look as if it's been properly engineered and possibly the best solution would be ot grab an old VCR case or the like and hide it all in there.
So grab an old VCR case... you seem to forget that cables exist, that digital TV signal won't originate from your TV location - there will be some reception equipment somewhere...
Before I ripped them out we had a feed from the aerial into the loft space, where there was a splitter from which the signal was then dropped to the myriad places people thought having a TV was a good idea... so there was basically completely free choice of where to put a tuner - just drop a cable from the loft, or put the tuner in the loft.
There are two problems with this theory. Problem 1 is that that's almost entirely back to the thing that we're theoretically replacing. A replaceable CPU on a board that has the connectivity to peripherals in a box designed to contain them sounds a lot like a desktop computer with cards and a CPU on a motherboard. The connectivity might go through different types of ports, but otherwise, it's the same concept.
The second problem is that this is less common than we'd want for similar reasons why it's less common in the desktop form. Manufacturers don't often want to do it, because it reduces the sales value of new hardware. Apple could easily stop soldering their storage in, in fact they have on some models, and put the necessary firmware on a separate storage device so that you could replace it with someone else's storage, which is still a problem on those models. Unlike with the RAM, there's not a speed difference involved. They don't want to, so they don't.
But even in the Raspberry Pi world, there's a problem. The Raspberry Pi CM5 doesn't work in the same devices designed to take a CM4, and that doesn't work in something designed to take a CM3. Every time they've released a new module, it requires new carrier boards. That's not really a surprise because the new processor tends to have more and faster peripheral interfaces and require more power, but it means that the idea of replacing one CPU with a faster one without ripping out the rest of the hardware is not an option and hasn't been except for the one time when you could replace the original compute module with the CM3. The Raspberry Pi module doesn't work much for forward compatibility and is only popular because it speeds up testing for manufacturers.
How many boxes do you have on your bench that used to be PCI cards?
And there is no reason to have a tangle of leads anywhere, that's just messy layout.
You need a PCIe device, that's a box which probably takes the following cables:
- Power
- Thunderbolt
And for many cards the power cable should be optional - just power it over the thunderbolt cable
So with one cable your specialist gear can now be anywhere on the bench, and facing in a convenient direction, so your signal cables don't need to be too long, or awkward to get at to plug and replug.
That thunderbolt cable isn't "a mess of cables" it's the freedom to locate and orient your specialist harware
"it's the freedom to locate and orient your specialist harware"
Actually, it's not. It doesn't include the freedom to locate the specialist hardware in a single box, securely plugged into the processor in a unit that can be tested in a workshop, picked up whole and positioned in a production environment sll ready to go without having to be reassembled and retested. Anybody needing that is going to tell you that your separate boxes are an unacceptable mess of cables and they're certainly not going to buy it.
50 years or so down the road and we're being offered something that looks like a hobbyist solution as the future.
Of course you can - if you want a custom enclosure just print one.
But for the majority of cases it's actually far more convenient to have your custom hardware not tucked with the only accessible portions be under a desk, in the dark, where you can't see what you're plugging in.
What bit of hardware are you actually thinking of?
Again - I would stand pretty confident that one cable isn't "a mess" of cables, and the ability to use that expensive specialist hardware with multiple machines is probably an advantage.
Exactly what I came here to say. Thunderbolt/USB-C is fast enough for most use-cases that aren't already being integrated into SoC's, and the PD capability is also sufficient for powering most of those too. I've no doubt PCIe will remain a thing that's available for certain niche applications, in much the same way you can still buy serial adapters or external SATA docks, but if the majority of consumers don't want, need or care about such things, it'll disappear from the mainstream - and pretty much already has. Business customers are a different kettle of fish, but with the prevalence of leasing arrangements these days, they're also caring less and less about this stuff for the bulk of their fleet. Gaming was one of the last fortresses when it came to consumer uses, but the writing's on the wall there, given the climbing price of GPUs, coupled with the ever-rising TDP for diminishing returns in performance and capability. Moving them into an SoC that offers better power efficiency and performance for very little difference in cost (at the mid-high end of the market) is the obvious place to go, as Liam is arguing.
Good luck finding USB/Thunderbolt adaptors for
Satellite with gotoXY motor
Traditional wired SCSI
Firewire
Floppy for arbitrary filesystems. USB floppy drives are very limited.
GPIO cards
GPIB for old but expensive to replace test gear.
Sometimes serial or parallel interfaced devices need a real I/O port on bus and not via USB.
I remember around 2006 or 2007 the PCMCIA slot was ditched on laptops, but stupidly the the replacement had also USB as well as some form of PCI. Almost nothing was migrated to the the new card format.so it was soon abandoned for USB only. The USB to PCMCIA adaptor only supported storage devices.
Also ethernet should be built in. I resent having more space used in the bag for adaptors than the laptop!
I've no objection to super slim laptops for people that only want WiFi, wireless mouse and one extra screen via USB-C DP Alt mode. Having no separate power port is too limiting. Most have too few USB ports now and need a USB 3.0 hub. Some have ports where the power can't be disabled and when off the battery is run flat by the HDMI or ethenet adaptor.
We need choice. Fatter laptops for engineers etc with HDMI, ethernet, three real audio ports (out, stereo in, and mic/headset switchable), a real parallel and serial port. Maybe even bring back PCMCIA.
We need PC cases & MoBos with PCI and PCIe slots, IDE, floppy and SATA ports.
Someone needs all of that, but so few need most of it that it's not very likely. A bigger machine with HDMI, ethernet, separate audio in and out, yes, people need that and enough of them that you can actually buy it. Parallel ports on a laptop? So rare that someone wants to buy that that nobody makes it anymore. Same with PCMCIA because nobody makes anything even slightly new that uses it so, unless you absolutely need something old which used that method, it's a completely useless slot which takes lots of space and internal components. You can hope for that all you want, but it's not surprising that you'll never get it.
On the basis that thunderbolt exposes PCIe, I don't think any of those would be impossible to find, and the market for most of them is incredibly small.
I rather think that the age of the floppy disk is well and truly over - the handful of things that still use it should probably be virtualised already, they're just not going to last much longer.
And no, I really don't want to go back to the days of having a breeze block of a laptop with an onboard parallel port.
Again, dongles exist, and no they aren't the spawn of the devil. They're what allows engineers to use a laptop which has the economies of scale that are afforded to everyone else.
You wouldn't sell any laptops if you included a parallel port, and likely none with a serial port (though that's probably more likely) because the cost would be *way* higher than an adaptor that can be left on the cable you're using to connect to whatever ancient doohickey you're thinking of.
there was one in 1995 the Tulip Omniversa in the ISA slot.
That is, according to the folder, this computer. It's a Pentium 75 based computer with a quad speed CD-ROM drive, a Soundblaster 16, 2GB HDD, 14k4 modem and a Haupauge TV-card.It also comes with an infrared remote and this was all combined in one neat package.
Even the fastest mobile GPU (5090 mobile) is more than 50% slower in actual game play than its desktop conterpart and is roughly on par with a mid-range card. There will always be a market for discrete components and no amount of Apple-style "you will eat our dog food and enjoy it" will convince me otherwise. Of course no one is gaming on a Mac...
Apple has for years provided excellent hardware for professional video. The integration of RAM on the die makes a lot of sense for transcoding and other processes where you want to avoid copying between processors. It's a niche market which is happy to pay a premium for stuff that just works. And for a while Apple was one of the few companies speccing machine suitable for local LLM work.
Think is, are we not fast approaching the point of diminishing returns for gaming graphics power? Once that gets hit, then the discrete power-hungry GPU just starts to shrink, both literally and from a market perspective.
Though I feel I probably said that about the original GeForce 256 Annihilator Pro (where I originally got this username I think - or certainly the same sort of time) with its glorious 32MB of DDR memory that really made Quake 3 Arena come alive.
So this is heading for workstation class desk side units, like HP Z series. The people who want such things want customisation and therefore sockets and stuff. That’s not economic for the great unwashed so it will be SoCs and more integration for them. An Apple M series laptop must be pretty much a chip on a board with very little else now. But then a Mac Mini is much the same performance as a PC costing four times the price.
Yes, let's get the really obvious one, "GPUs", out of the way. Yes, that's the big one everyone will name first. Thing is, Apple Silicon Macs don't support third party GPUs, and Apple doesn't offer GPUs in a discrete PCIe form factor, so PCIe slots have never been usable for GPUs on the Apple Silicon Mac Pro.
And GPUs are the ONLY mass market use left for PCIe in 2026. Anything else you can name is either handled just as well by Thunderbolt or is a niche case in 2026.
Apple will have had data on how many Mac Pro buyers used their PCIe slots, as well as how much Mac Pro sales were impacted by the Mac Studio which is essentially identical except for the slots and internal space in the case. The kind of people who need to install outdated PCIe cards that aren't made anymore or wanted a quad 100GbE interface card or wanted to create an array with a couple NVMe drives weren't buying Macs in the first place, so there is no point to Apple catering to them.
Which version of Thunderbolt?
Plus if I’m running a bunch of LAN interfaces, storage devices and multiple monitors, I’m going to need a multi-port Thunderbolt hub and power supply…
Remember only v3 is royalty free, and it took a while to get non-Intel hardware eg. AMD “Thunderbolt” certified given Intel retains the certification.
Not seen any wide use of Thunderbolt 4 and 5.
Basically any version over 3.
But here's the trick... a thunderbolt 3 device will work in a thunderbolt 4 or 5 port.
Shock, horror... Almost all of my devices are actually plugged into my TB3 dock, makes it very easy to take my desktop with me when I need to (extended family breaks at the beach when I don't want to take pto)
Your needs for PCIe slots are because you have a crappy outdated board that skimped on USB ports and has a slow builtin NIC. If you were buying something new in 2026 you wouldn't need PCIe for any of that stuff.
Thunderbolt does 80 Gbps bidrectional, which is enough for a 100 GbE interface (it won't hit the full 100 GbE but people who truly need a full 100 Gbps rather than 80% of it are well into the server or highly specialized workstation market) or connecting an external array or adding USB ports (I've never used close to half of the USB ports on any desktop type PC, they always have more than you need)
Anything which needs lower latency than Thunderbolt, as that ends up doing a CPU <> PCIe bus <> PCIe bridge <> Thunderbolt <> PCIe bridge <> external device <> external I/O dance.
High end sound cards come to mind (well, they would, if anyone still made them!)
High end storage as well (like NVMe over Fibre Channel, or maybe NVMeoPCIe RAID array, if low latency is required for workload)
High end networking like Infiband, RoCE, NVMeoEE, even plain old Ethernet
Maybe, just maybe, AI coprocessors (with real neural networks, not just GPUs with their video ports removed)
Even high end VR/AR, assuming full I/O ("gaming" plus real time audio/video inputs and full multichannel A / dual or multichannel for "array of monitors" V outputs)
Or even a combo: AI card plus GPGPU card co-driving VR/AR I/O (with Thunderbolt driving the optional sex toy or fembot chassis)
True, any dedicated, mass produced appliance would integrate to drive down costs, but, hobbyist/developer/creator types would need something flexible, low latency, not mass-produced, and would pay a premium price for it.
Given that most places use network storage I think the assertion that you need better than thunderbolt latency for your storage is harking back to the days when external storage was actually slow.
Sound cards have no issue with thunderbolt latency, hell they had no issues with the latency on USB2, even running 32 channels at full quality (as many professional mixing desks do)
High end networking over thunderbolt is also fine, but seriously, how fast do you want to go with a human facing device?
What do you reckon is the minimum latency of a TB5 connection?
Given that Apple expose RDMA over thunderbolt - it's got to be pretty decent.
Apple's data from 2013 (https://developer.apple.com/library/archive/documentation/HardwareDrivers/Conceptual/ThunderboltDevGuide/Basics01/Basics01.html?utm_source=chatgpt.com)
- Devices on a Thunderbolt bus exhibit higher latency than devices on internal slots—about 1.5 microseconds of round-trip latency per hop.
So yes, you're adding a bit of latency, but PCIe is already looking at .5-1us anyway (Cisco Nexus K3P-S data sheet), so we're in the same ball park.
https://www.cisco.com/c/en/us/products/collateral/interfaces-modules/nexus-smartnic/datasheet-c78-743827.html
If you're really that latency sensitive then you'll want to build a processor into your "whatever" to manipulate the data, and either control that from your computer or send a processed version to your server farm to play with later.
Anything which needs lower latency than Thunderbolt...
OK, and which of all these things you've listed are anything a consumer or small business would be doing? They aren't using Fibre Channel, Infiniband or AI coprocessors. Though when you get to talking about fembots I wonder if you're just trolling here.
Hobbyists aren't using most of that stuff either. Now yeah some hobbyists want PCIe slots even if they don't NEED them for anything in a modern PC, "just because", and that's fine. The kind of people who want that were never in the market for a Mac in the first place, so the existence or non existence of the Mac Pro doesn't matter to them or help Apple sell any of them.
"And GPUs are the ONLY mass market use left for PCIe in 2026."
You don't do any video work, do you?
Thunderbolt is great, but external stuff connected with cables is a liability. If that external box can be de-packaged and put inside the computer, it's one or several less cables to worry about.
When I do photo jobs, I have a bunch of things attached to my camera. Just recently, I've put the camera in a SmallRig cage and stuck the Camranger and other stuff onto the cage so when I get to a job, I'm not spending time assembling/disassembling and cables are fastened down so I'm not tugging on the camera's ports and working the cable ends. It's impossible to find quality cables so I have to order a bag full of cheap Chinese crap and find one that works and swap it out when it eventually become intermittent. I would love to have a camera that doesn't need external bits added to it, but makers are keen on keeping their ecosystems closed/proprietary. Canon does have some wireless remote capability, but it's crap. If they published the specs and/or provided a SDK, that would be one more dangly bit that could go away. I could use the CamRanger software on my tablet that connects via the wireless capabilities that the camera already has. It might also allow me to interface a product I have for photographing rocket launches up close (wired since wireless is prohibited).
I run video distribution and streaming systems using a thunderbolt connected Decklink Duo 2.
Quad SDI connections with the connections at a place that's convenient (not round the back of a tower computer) and therefore doesn't need a patch bay to be useable.
I'm sure you'll have considered whether there are any shoe mounts available for your various bits and bobs on the camera itself - but that's not really the fault of thunderbolt.
I hate to agree.
Yes, there are niche-market applications, like high-end sound or video cards. Yes, there might be someone that sees a use for specific interfaces, but that is not mass market.
On a side note: MACs have always served the artistics that needed calibrated video or specific audio. That may become less an argument to buy Apple hardware now.
Not just apple, amd and presumably intel if they can manage to get in the game.
I think this shift to Soc will do to Nvidia what digital photography did to Kodak unless they can get a revenue stream to replace GPUs. And as a keen film photographer and very early digital adopter, the lesson was that the vast majority of people will accept good enough (or even crappy) over perfect if the price point is right, which we all probably already knew...
"unless they can get a revenue stream to replace GPUs"?
They've got one. AI. They're going to milk that one for all its worth while they can.
Also, NVIDIA kit isn't just in the consumer space. Their GPUs are used in datacentres to provide graphics capabilities for virtual desktops. That business won't be disappearing as quickly.
I know they've got the AI revenue stream, and it's probably much more important to them than GPUs for now, at least until the bubble bursts. And maybe they can fall back on crypto mining. And maybe they can license their tech into a SOC provider. Or become a SOC provider (which I assuming is what they were angling for when the bid for ARM was knocked back). I still wouldn't want my retirement fund loaded up with their stock though. Doesn't matter either way, if NVidia tanks it takes everyone else with them.
"the vast majority of people will accept good enough (or even crappy) over perfect if the price point is right, which we all probably already knew..."
Unfortunately, that can be the bulk of the market. As a professional, I'm ok with spending more if it means I can work faster or not have to work as hard getting around limitations. Like my cars, I use my cameras until the wheels fall off if I can. I'm not going to add up what I've spent on my view camera as that's a hobby and I don't want to know as it might rub some of the happy off of it. Hmmm, how much is a 6x17 roll film back.......
Quote: I think this shift to Soc will do to Nvidia what digital photography did to Kodak
Maybe 5+ years ago, but unlikely now, their gaming revenue (i.e. discreet GFX cards) was around $11.35B in FY2025 (ending Jan 2025). A decent chunk of money, but that was only 8.7% of their total revenue.
On the other hand, their data-centre side generated $115.19B in the same period, representing 88.3% of total revenue.
Gaming revenue actually grew in FY2026 (ending Jan 2026) to $16.04B, but data-centre grew more to $193.74B, so its revenue share also grew to 89.7%.
The only way Nvidia are going to end up in trouble, is if/when the AI bubble bursts.
The path to performance is increasing integration.
No, the path to decreased manufacturing costs, planned obsolescence, and more landfill economics is increasing integration.
Increased integration can get you some increased performance, but with the limitations of the SoC manufacturers' design.
No company is omniscient and can see the most-cost-effective way of achieving performance level X. A PC purchaser is locked-in and limited by the SoC design.
Further, many companies do not want consumers to get the best-possible performance for price Y. They want consumers to buy a(nother) spendy PC from them to get their desired performance level.
Contrast that with, "O.M.G. look how more smoothly, quickly, and beautifully Quake runs with a 3dfx Voodoo auxiliary video card, than it does without! I am saving up for one of those!"
But you couldn't use a card like that without a PCI slot (later 3dfx also produced cards which worked with AGP slots).
SoC improves interconnects at the expense of thermals. Shared RAM and shorter interconnects greatly improve performance of a given puece of silicon - as long as the problem still fits.
So it means less performant silicon can do more - raising the floor.
That makes a lot of sense for a consumer device.
It also means the silicon must be lower power and thus performance, or the smoke comes out - greatly lowering the ceiling.
And once it starts to run a bit hot, it must throttle aggressively - and the machine becomes almost unusable.
Worse, the moment the task touches the edge of the memory limit, the performance of the entire machine crashes and the SSD wears out.
You cannot buy more RAM or silicon performance. Your only choice is whether to make the task smaller, or to rent someone else's computer.
As Moore's law runs out and technology approaches the limits of physics, other things have to be done to create enough old computers to fill up the landfill.
On the other hand, the economies in post industrial countries are no longer productive enough to support the waste of planned obsolescence.
I recall owning a truly ancient PC (somebody else's cast-off) whose hard drive was mounted on an expansion card.
That would be an oddity.
Hard disk controllers were ISA cards at least in the PC/XT(ST506?) and AT(ATA) and I imagine in the 386 etc machines which were mostly AT compatible until mainboard board IDE controllers were standard.
I guess some clever vendor saw a market for a controller card with a thin, small laptop disk glued on. Given I have never seen such an adaptor (I have seen some peculiar hardware) I might conclude it was a rather small market.
Unix kit was all SCSI (and other oddities like IPI) so I didn't have much to with PC tat after the AT but I suppose the adaptor might have been SCSI too. HPUX workstations of the period sported (E)ISA slots and I could imagine the card was for a wider range of hardware than the x86 kit.
A thin laptop disk? These were full 5¾" jobs. The thickness of the drive mean that the adjacent slot was blocked. Looking back they may seem like an oddity but they were real. Look up old copies of Byte on archive.org &you'll find the adverts. This was before HP/UX and maybe before SCSI.
I'm glad I have sacks of old PCWs and Bytes around the place - all this sort of thing is part of our history and all too easily forgotten.
Simpler times. A bought a "25 MB Winchester" that was 19" rack mountable for a DG Nova. DG sent someone out to see why I was using such a big disk.
The voice coils "sang" when it was running, and after a while I learnt to work out where it was in a sequence by the sound it was making. One day the sounds were interrupted by a nasty scraping sound and everything stopped. Through the transparent plastic top I could see the gouges in the oxide when the head crashed, the inside was coated with it. Fortunately it was still under warranty, as it cost about the same as my new car.
To quote from your link in another post
"While sources inside the company during the launch of Hardcard claim it was the first HDD controller integrated into the drive printed circuit board,[6] Xebec, a HDD controller manufacturer in the early 1980s, had already done that with their Owl product around August 1984. It was a complete 5.25 inch half-height HDD with an integrated controller and drive electronics on the same printed circuit board with a SASI interface.[12][13]"
Perhaps what the other poster was referring to?
@ Bebu sa Ware:
No, the "HardCards" pioneered by Plus Development (and later produced by many competitors) were not an "oddity".
They allowed increased permanent storage in PC- and XT-compatible computers which typically had two full-height drive bays, and later, just two half-height drive bays.
Almost exactly 40 years ago, I added a 20 MB hard card to my Tandy 1000.
Which I loved!
On the same Saturday morning visit to my local Radio Shack, I also expanded the RAM from the 384 KB that I started with in April of 1985 to the maximum of 640 KB.
In the day, the ability to expand a PC or an Apple II was a huge advantage.
(As was Radio Shack’s ubiquity.)
https://en.wikipedia.org/wiki/Hardcard#Tandy_1000
The Amiga 2500/30 was a stock Amiga 2000 (with M68000 in a socket on the motherboard, along with 16 bit "chip" RAM), a native Amiga Zorro II bus (M680x0 bus) with slots, and an initially empty set of ISA slots in-line with the Zorro slots, which were connected to each other, but nothing else.
It shipped with an expansion card containing a M68030 CPU(including MMU), 68881 MPU, and 32bit RAM (with connector for more 32 bit RAM).
There was also a SCSI card with a bolted on HDD.
I added an x86 bridgeboard card which, as the name implied, contained an x86 (8086 I believe) and "bridged" the two busses on the motherboard. Only the x86 has access to the ISA bus slots.
Also had a genlock card (poor man's Video Toaster, from before Toaster shipped)
Expansion cards ruled in the late 1980's-1990's.
I was impressed by the hardware in a ~2010 mac pro tower that I inherited as a disgarded six year old box. Admittedly the original owner had maxed out the ram and disk storage (2Tb drives.) I was more surprised that I could install RHEL7 on to the system without any gymnastics at all. I used it as test machine few a few years with extra disk storage set up as ZFS raid5 volumes which were used as isci volumes for machines on a test network hanging of one of its ethernet ports.
Yeah the Mac Pro towers (the one made up through 2013) was impressive. Lots of RAM expandability, lots of storage expandability, and plenty of PCIe slots. The modern one looks like the old one, but the RAM is soldered, storage is soldered, it inexplicably has all these storage bays, but no way to actually hook them up to anything. And the PCIe slots, but there are VERY few PCIe ARM drivers for macOS. (But, if you put Asahi Linux on there, well, PCIe is PCIe so almost anything that'd work on any PC will work in there. Even the unnatural combination of ARM CPU + Intel ARC GPU.)
Not for servers.
Most datacentre servers are sold purely on the ability to swap out failed components easily, and to upgrade in place. Unless they were very, very, very cheap, no-one is going to buy a server with soldered in, non-upgradable RAM, or non-replaceable or expandable storage.
Apple have a longer term dilemma given their direction of travel. When do they fold the Mac, Macbook, and iPad into a progressively single product line?
For the rest of the PC market there is still a long way to go, and competitive platforms, PC/Windows, Tablet/Android, Tablet/Chrome book, PC/Linux.
My guess is that tower type PCs will move towards server componentry. Expensive, but expandable.
Not only servers, byt many industrial PCs and often workstations - I just built a workstation with PCIe cards to connect to avionics buses - for which we needed the lower latency available - of course this is a market Apple is utterly not interesed in, and if it killed the mac Pro line, the reason is they weren't selling anymore.Probably they asked too much for a system that didn't deliver enough, since they switched to their own CPUs. The fact Apple introduced a lower-cost macBook shows IMHO they are changing strategy, for some reason.
> Most datacentre servers are sold purely on the ability to swap out failed components easily, and to upgrade in place.
I think this is the reverse of reality. In fact, for datacentre deployment, each server is a disposable fungible unit.
It is, as I have said elsewhere, well over a decade since I did field support and closer to 2, but I've worked in a few datacentres and with a few rackmount boxes. I've never even _heard_ of anyone upgrading one of the things.
> I've never even _heard_ of anyone upgrading one of the things.
No, they never *upgrade* them - that happens with lease refresh cycles, but I have know bad ram to be exchanged a lot along with anything else that can be unplugged, and if you have a machine with 256+GB on board you are not simply going to replace the whole machine (at least not with todays prices !)
Probably 95% of people, if not higher, NEVER upgrade their pc.
People on here, and in other tech oriented places just forget there is a world out there that is not theirs. It's like watching Linus Tech Tips and many on his channel complaining that Linux isn't good enough because it can't play AAA games. Guess what, most people are not fucking gamers. Or like the poster above saying it can't run multiple monitors (they can). Guess what, most people don't use multiple monitors and if they do,it's usually just two
For the VAST majority of people SoC are absolutely fine. The really issue is if you buy one that is under spec'd from the start (think those horrendous HP things).
Exactly - just because there's still a niche market for something, doesn't mean it's going to stick around for mainstream hardware. Liam gave plenty of examples of this happening in the past. Startech's whole business model revolves around this process, after all - they specialise in catering to this niche market, much like Radioshack and Maplin used to.
We have the contrary situation here among people who are apparently IT pros. They think it's acceptable to put together a system with a lot of little boxes strung together with leads. It isn't going to play well controlling industrial machinery or a medical scanner, or the like.
> It isn't going to play well
I think you're wrong. The trend is already visible: millions of little Intel NUC style sealed boxes, often screwed onto the backs of screens via the VESA mounts, in the baseplate of standing digital signs, and otherwise built into the gadgetry they are controlling.
Plug it in, it picks up an IP address, downloads an image of the installed system drive from a server, configures itself, and sits there waiting to be told what to do. When they fail, due to flgas being clogged with dust and so on, you don't fix them you just replace them.
It's 18 years since I was in field tech support but we were already rolling these things out then.
Even NUCs have replaceable RAM and storage, unlike Apple.
Certainly NUCs and similar have their place as thin clients but not everyone can use one for the work they have. There will be a place for the expandable PC or server because work always expands and needs more resources. So much for the 15-hour work week we were supposed to be working.
There are still a fair few applications where tower PC's and the like win out. My repair bench always requires multiple ethernet connections to run diagnostics on company kit. As a result the 2014 era HP tower (Core i7, 32GB ram and a nice big SSD) mounted under the bench is fitted with two ethernet PCI-E cards alongside the motherboard NIC and I've yet to have any connection issues with the setup.
When asked to set up a similar bench for someone else, the bench PC provided was an intel NUC. This involved a mess of USB hubs and ethernet dongles to get a setup together that included two monitors plus keyboard, mouse and spare USB connections for hardware programmers and the like.
Not only is this unwelcome on the bench but the USB dongles can be flaky and lead to wasted time troubleshooting.
> tower PC's and the like win out
If you can buy them cheaply.
You can still buy PCs with ISA slots -- but they're very expensive specialist kit, and both dear _and_ slow.
I think a lot of people are misinterpreting me. This won't be instant and it won't be all at once.
A lot of laptops have come with soldered RAM for 5-10 years now. Especially the faster ones, and the slimmer/lighter ones. The bulk of the end-user work PC market is laptops now, I suspect. Desktops are a declining niche.
A bigger SOC package with CPU, GPU, RAM, and most of the chipset comes first, with NVMe slots for storage that's still offboard.
Some posters said the point of servers was to be repairable. They haven't read a word of how Google has been operating for 20-30 years then. Racks of bare motherboards at first; if it fails, sling it. Now, blades, doubtless. Whole trays discarded at once.
_This is already happening_. As William Gibson said, the future is already here, just unevenly distributed.
Liam, I think we come from different places. Bolting NUCs on the back if monitors is fine although in that case maybe you should have considered all-in-ones (why flog the customer a replacement NUC when you can flog them a replacement monitor at the same time;). Back in the day I was putting together lab kit where being a single integrated box was vastly preferable so that the specialised stuff such as the ADC was inside, plugged into the motherboard. You don't get that with an NUC. Later,I was working in in a situation where big industrial printers were hung onto the PC with SCSI. I suppose the SCSI could be replaced with USB or whatever today but you're still going to want something that can live in a factory environment (IIRC the SCSI was pretty substantial - I think it might have been built as armoured cable just to help it survive).
"Guess what, most people are not fucking gamers."
I think the game industry revenue begs to differ:
"The global gaming industry is significantly larger than the movie industry, generating roughly $184–$188 billion annually as of 2023–2024. This dwarfs the global box office, which brings in roughly $34 billion, and often exceeds the combined revenue of the film and music industries together."
Honestly the current Mac Pro was kind of a joke. Same CPU as some Mac Studio model that was like half the price. RAM is soldered on so that's no advantage of the tower. It's got storage bays, but nothing to do with them since the storage is soldered on and no SATA or M.2 slots. The PCIe slots are there, but almost useless in macOS (since you need macOS ARM drivers and there just aren't that many of them. I will note, if you throw Asahi Linux on one of these, the PCIe is fully useable, since the Linux drivers are portable including to ARM systems. Even apparently the unnatural combinatoin of ARM CPU + Intel ARC GPU.) Apparently the one advantage of the tower is it's SLIGHTLY faster under extended full load operation (due to better thermals), but for the $1,000s in price difference once could have likely just pointed a fan at the Studio to make that up too.
To me this article is pure Apple fanboi'ism... like... Apple makes non-expandable systems, therefore everyone will switch to this.
There's those "Chromebook-based" systems where they've got some wheezy bit of RAM and SSD (sometimes a nice slow eMMC) soldered on, of course. But Intel has now announced "Lunar Lake" (which had RAM integrated on the CPU) is a one-off and they are not doing that any more, so that appeared to be a sign of this kind of trend, but Intels stepped back from doing this.. Anecdotal of course, but my mom got a notebok for $218 new recently (due to RAMpocalypse it now costs about $250). I really assumed (since one of the vaunted reasons for soldering on is to save the cost of the DIMM and M.2 slots) that it'd be a Chromebook-style board with everything soldered on. Nope! There's an access hatch at the bottom of the case, and it's using conventional DIMMs and M.2.
I don't see notebooks going all to "let's just solder it all on"; and desktops I DEFINITELY don't see this even starting to be a trend outside the Apple world other than perhaps the smallest systems (the ones that are basically a laptop motherboard just in a little brick-shaped case with no screen or keyboard rather than a flat 'case' with a keyboard and LCD built on). Any suggesting of SERVERS doing this is kind of a joke IMHO.
I do have a thunderbolt port or two on my desktop... but, I've never used it. Nothing against Thunderbolt, it IS amazing technlology. But, I added additional USB3 connectivity using a PCIe card, GPU via PCIe card. If I filled my M.2 slots I could add a card to add more M.2 storage too (although, to be honest, I have the vast majority of my storage in the form of spinning rust. I mean, 24TB of storage for maybe ~$400 over the last several years, even pre-RAMpocalyipse something like $2000-$3000 in SSD storage and now probably closer to the $4000-$6000 range.). In that vein I could also use those PCIe slots to pop more SATA ports for even more disk storage LOL.
This post has been deleted by its author
Apple have always tried to get a complete hardware lock down, This is just one more step to force their customers to have to buy new hardware every year.
Apple customers as a group are hipsters not geeks, and also aren't hardcore gamers, so Apple can get away with abusing their customers because
they aren't the type of people to mess with the hardware in the first place. They are already the type of person that would buy a whole new machine
rather than just buy a larger RAM or drive and install it themselves. Also most buy a new machine every year or 2 at most anyway because they
are also the type of customer that wastes thousands just sot others notice they have the latest version of hardware.
Just because Apple can do this hardly means that the PC world will or even could follow.
There are far too many significant use cases to justify the continuation of user-expandable hardware, not least are gamers, and far too
many giant manufacturers relying on the fact that people can buy components. AMD, Asus etc literally depend on it.
The inflexible upgrades at purchase also appear to be a trademark. Recently replaced Dad's 2013 era iMac with a new M4 model. (He's been a Mac user for over 20 years and is very much settled in the ecosystem). To be fair apart from a few tweaks, stuff does tend to just work and do what he needs it to.
Did suggest a Mac Mini but he wanted something in the same form factor as before.
His usage doesn't really require a powerhouse, but lots of storage for all the digital photos and music he's go, well over 500GBt. Of course, they don't let you buy the base model with anything above 256GB of storage space....and the 1TB SSD ideally aimed for was accompanied with an insane price increase.
The middle ground of the 512GB model (complete with 16GB RAM that isn't really needed for his usage) and hanging a Samsung SSD off the back alongside a LaCie HDD for backups was the workaround. It's a bit of a mess of cables and adapters to say the least with much raiding of my adapter stash to get everything connected to the thunderbolt ports.
Apple follows the logic of Game Consoles: Build a well known hardware and you get better performance and run time because you know all possible short cuts/tricks and have a well defined hardware without third party induced faults.
All in One systems or SoC have a lot less points of failure and so just work (or die fast and get replaced).
Apple also has the luxury of pruning, changing and evolving its OS versions as they like without many regards to legacy software which is the biggest problem for Microsoft.
And modern systems with all in one are usually better suited for the needs of the users. The big box with expansion approach is versatile but adds overhead that most users never really need.
For years now a standard computer is sufficient for most users doing office like work and media, so they can get smaller/compact and cheaper.
The Studio line from MS was a good idea but I guess they are reluctant to create a too strong competition to their customers/system integrators.
Apple have always tried to get a complete hardware lock down
This is false. The original Apple and Apple ][ computers were expandable and had open specs. Many companies made plug-in cards for the Apple ][, and this versatility was a significant contributor to that system's success. I have seen an Apple ][ with an add-in card automating a numerically-controlled milling machine.
The "new" Apple computer (starting with the Macintosh) went into "lock-down mode", though there have been occasional deviations from that (PowerPC tower units).
The original Apple and Apple ][ computers were expandable and had open specs.
Right on, Old Dog. Jobs and Woz had sharply differing views regarding the desirability of open, user-expandable systems. Jobs’s view eventually came to dominate at Apple during most of the past 42 years (excepting perhaps certain portions of the 1990s), but Woz’s view defined Apple’s original approach and prevailed for many years early on.
I don't think a lot of people who want Mac Pro / Studio type computer want something totally integrated. If anything this is showing that the integration design has failed. All it offered was pcie slots and a very high proce tag.
Can just build something modular with 5090 and whatever the latest Intel or AMD cpu is and be more comfortable long term.
Sure it consumes more power makes a bit of fan noise and pronably not manufacuted on the most cutting edge TSMC node, but it's a desktop not a laptop. A bit of extra power consumption doesn't matter so much here.
I recently got a very cheap no-name Android tablet, under £70, it's fine but a bit slow sometimes. I was amused to see its claimed 10GB RAM is 3GB actual RAM and a 7GB swap partition. Odd numbers, literally and figuratively. What's the betting they're both failed parts, 4GB and 8GB respectively? I'm pleased to see such getting a use, and enabling cheap simple devices, rather than being binned, pollutingly.
A great article, Liam.
There are good physical reasons for centralising everything on one chip.
It is a pity that, metaphorically, the centralisation is also of manufacturer's power over the consumer. I don't think there is a version of Linux than runs well on anything above an Apple M2.
The most recent Mac Pro was dead and irrelevant when it launched. It was an outlier in the desktop tower space.
I use many different form factors depending on the use case:
- AIO (iMac 27" 5K)
- Mini PC (Mac mini)
- Laptops
- Towers (servers, gaming PC, Linux workstation)
As it stands now the high performance gaming space is handled by video cards and x64 CPUs. Both of which put out a great deal of heat. Until that is resolved a large tower case with multiple fans is the only option for me. I'm not a fan of SFF cases for gaming setups as I want excellent airflow and as silent a system as possible even under load. Dual slot GPUs take up two PCIe slots on a desktop motherboard and I need a 3rd PCIe slot for a fast network card (25 Gigabit Ethernet). If one wants to leave space for cooling between PCIe expansion cards then one needs even more available PCIe slots. A SFF motherboard or a mini PC is not going to suffice.
For my home VM server I use a tower case (many big fans that spin slowly and quietly). I also need a number of PCIe slots for GPU transcoding, fast network cards (25 Gigabit Ethernet), and a RAID controller.
Video cards are also being used for local AI compute so a desktop motherboard packed with video cards requires a tower case. Having said that something like the Mac Ultra or a Nvidia DGX Spark is better suited to local AI work if one has the budget for those systems.
I started in computing January 1 1985. Throughout rthe years I have seen many come and go. I was one of the fist users of Microsoft Word and Excel in Windows format on a MacPlus in 1987. But consistently across the decades I have seen Microsoft used and acused by criminals time and again. I seen the Apple Haters harangue Apple over and over. But I have yet to see an Apple product unless with a Ransomware Message on the screen! SO all you Apple haters please tell me. You knock Apple, yopu slur, some libel and slander, but still you run off and have to pay exit for protection of your OS where Apple users sit and quietly laugh at you.And it seems you wonder why? Apple have consistently broken with then boring to much ahead. Just look at how many Asian Smartphone builders are out there whilstApple retain the iPhone. Of course there will always be the cheaper offerings, but! Do they last?
So this is a "Macs don't get viruses" again? I assume that's what you're trying to say, although technically, "I have yet to see an Apple product unless with a Ransomware Message on the screen!" means the exact opposite. And in fact Mac OS is as possible to install ransomware on as any other operating system. Both LockBit (because they were big and targeted lots of things) and North Korea (because they focus on cryptocurrency researchers and operators with whom Mac OS is popular) are well-known for building Mac OS versions of their ransomware.
Like “yopu slur”?
.
“Microsoft Word and Excel in Windows format on a MacPlus in 1987”
In Windows format, but on a Mac and in the Macintosh version of the program? In context, what do you mean? Just that you told your Mac to save spreadsheets in BIFF2 with the 1900 instead of the 1904 date system, and documents in RTF?
On a Mac Plus? In 1987? Even though the first Windows version of Word was not released until November of 1989?
.
“I have seen Microsoft used and acused” (sic)
???
Accused of what by criminals?
The problem here goes beyond spelling.
.
“I have yet to see an Apple product unless with a Ransomware Message on the screen!”
I have never seen an Apple product with a ransomware message. Not my own Mac Plus (delivered in early September of 1986), nor either of my iPhones, nor any other Mac I have ever used, nor the iPad I am typing on now.
But I know of a certain famous politician who also capitalizes random nouns.
.
“Apple have consistently broken with then boring to much ahead.”
All your base are belong to us.
The number of people expanding Computers is low. Most buy a System kitted out, and at the time for a big GPU upgrade usually the whole thing gets replaced.
Building and upgrading PCs becomes more of a hobby and is for a small group of people with expertise. One issue is that the market cycles are so fast that if you do the work, read the tests and select your ideal components some of them are already hard to get
There is of course a strong second hand market that might profit from the RAM price rise that supplies the low cost market but that is shrinking and very capable Laptops are cheap.
And we of course still have expansion: USB-C
That works for most expansion stuff including special hardware connections to industry or medical devices etc.
The old Mac Pro design was a necessary evil of the x86 era. That architecture demanded an oversized power supply, massive GPUs, and a huge cooling system just to keep the whole thing from melting down. Looking back, it was absurd.
The transition to the far more efficient ARM architecture changed everything. Now an entire high‑performance workstation can fit into something smaller than a shoebox, and it still outperforms most high‑end PCs.
If you miss the antiquated tower aesthetic, you’re free to stick with x86. Just be prepared to hook it up to a nuclear power plant and the radiator from a truck to keep it cool.
There is another thing too, happening. Until 10-15 years ago, the only way to have fast access to data was to put it in RAM.
With NVMe, things can be loaded from file on demand, instead of pre-loaded into RAM.
8GB RAM can be quite fine on the new MacBook Neo for this reason.
The Intel MacPro supported 1.5TB of RAM. Apple Silicon now only comes with up to 256GB (according to the article).
But a smart application can often keep the rest on NVME.
Not a fan of apple products, or rather the companies designed in constraints and planned obsolescence, so it makes sense they've killed this restricted to what they'll allow expansion tower. My ancient iBook laptop, bought used only for the FireWire facilty to update an equally ancient Kodak Pro cameras firmware is a prime example.