I haven't read all the comments, so this might have already been mentioned/discussed, but it makes sense that Apple would "pre-announce" pro models versus their famously super-secret shroud of mystery over their consumer products until they're actually released. We consumers, love the spectacle of an unveiling of the shiny new iPhone X or iMac or whatever. And if we have the money, and want to be the first on the block to own one, we'll gladly wait in line, or lock in the pre-order. On the other hand, if you're running a design studio, you need to make budget decisions for next year, meaning, you need to know what will be available next year. If Apple held all their pro announcements for a special unveiling, then they'd have to wait quite a while for significant sales to come in. But, if Apple announced now, that, say, in June 2018, they'll have new Mac Pros with accompanying Apple-branded 5K displays (and include tech specs, options, and prices), then my hypothetical design firm could plan ahead and allocate money to make purchases next year.
The Mac is a DYING and SHRINKING platform. PERIOD.
Increased Mac sales (doubled over the past 8 years), $23bn revenue/year, Apple in the top 5 PC vendors and is the most profitable PC vendor - equates to a dying platform?
It's a growth platform and Sog35 is if not anything, consistently wrong and myopic in thought.
I haven't read all the comments, so this might have already been mentioned/discussed, but it makes sense that Apple would "pre-announce" pro models versus their famously super-secret shroud of mystery over their consumer products until they're actually released. We consumers, love the spectacle of an unveiling of the shiny new iPhone X or iMac or whatever. And if we have the money, and want to be the first on the block to own one, we'll gladly wait in line, or lock in the pre-order. On the other hand, if you're running a design studio, you need to make budget decisions for next year, meaning, you need to know what will be available next year. If Apple held all their pro announcements for a special unveiling, then they'd have to wait quite a while for significant sales to come in. But, if Apple announced now, that, say, in June 2018, they'll have new Mac Pros with accompanying Apple-branded 5K displays (and include tech specs, options, and prices), then my hypothetical design firm could plan ahead and allocate money to make purchases next year.
I agree. They cornered themselves into a lose/lose situation through silence and lack of product. At least this way they 'lose less' and pick up kudos for laying down a publicly stated (if vague on the dates) roadmap. It does far more good than harm.
Credit where credit is due. This was a difficult thing to do for them.
It opens up a lot of questions regarding management, resources etc but it does mark a change in approach which is overwhelmingly positive. I hope they keep it up.
I clearly see no one in here is deeply invested in the history of backplane technologies from DEC and NeXT moving foward. PCI-E is the bus to rule all bus. It will need to become modular in design for Apple's vision; and it will need to jetison Intel's anchor, Thunderbolt, to get it done.
The backplane design will be key and how the power distribution is across a modular PCI-E bus design will make it possible.
This is unprecedented. I mean, seriously...has this ever happened before that they would announce something this far out from release?
I'm very happy with this news. I'll take the minor speed bumps. However, I think they could have also dropped the prices of the current crop a little.
They did do a $1,000 price drop today on the Mac Pro line. They dropped the prices of the middle config by $1000, and so it is now the low config. A custom 8-core config is now the upper standard config, but at $1000 lower than before.
That hadn't happened at the time of my post. I did read the Daring Fireball article, but saw no mention of price reductions. It's a good move.
thewhitefalcon said: [...] "massive internal storage" means obsolete 3.5" spinning disks on a SATA II bus
The main point of your post notwithstanding, I wouldn't categorize 3.5" spinning disks as obsolete just yet. My projects reside on a 16TB RAID. There's no way I could afford to replace that with SSD storage. I don't think many people or companies can. So any kind of "massive" storage -- whether internal or external -- is going to be based on platters for some time yet.
That's not really true, onLy a small percentage have left the building, and most of those are looking back.
I dunno Mel. Among my peer group, everyone I knew used FCP up until X came out. Now most of them have Adobe subscriptions. Considering the cost of that, and how much they resent that payment model, that's saying something.
Again, this seems to be a matter of opinion. I agree that Apple (eventually) overcame the workflow and feature issues that drove away lots of people in the beginning, but the people I know who use it (because they have to -- that's what the company supplies) don't care for it. They prefer a traditional timeline. I don't use it myself, but the most common complaint I hear from co-workers is that the FCPX/iMovie editing paradigm treats audio like a conjoined twin, and trying to work picture and sound separately is more hassle than it was with FCP7.
So is the tone of your comment. It's fine to disagree, but the insults and arrogant swagger are unbecoming a teenager in a parking lot. I'd like to think we can do better here. To whit, I'd like to respectfully address one of your points:
FCPX is a vastly improved version of final cut, if you actually ask the pros who fucking use it, and it has MORE features than the older one, while being more intuitive to use.
I work in a facility that supplies MacBook Pros with FCPX to editors in the field (in house we use a combination of Avid and an awful Velocity system). Of the dozen or so using FCPX, the number who like it is zero. None. Not one.
It's not the feature set they dislike, it's the user interface. They all say they prefer a traditional timeline like MC and Premiere. They say the FCPX/iMovie approach slows them down (they're on tight deadlines) and makes it harder to work on audio and video independently of each other.
That's based on actually asking "the pros who fucking use it," who apparently don't fucking like it. That doesn't mean it's not a good product or that there's any reason someone else shouldn't like it, just that there are no absolutes when it comes to defining what is and isn't "pro."
I don't understand the hate towards the current Mac Pro.
My previous one was HUGE and weighed grunt pounds. The small size of the current one reduces the placement restrictions the old one suffered. (Rack-mount would be even better, but that seems to have gone out of style for some reason.)
Whether storage is internal or external seems utterly irrelevant. Having it external is probably even slightly better because it can be moved to other machines if desired, and is easier to service.
No PCIe slots is only an issue because some manufacturers refuse to simply provide stand-alone versions of their products. Blackmagic makes video capture and output devices in card form, but they also provide stand-alone units that connect via Thunderbolt. I've used mine on three different computers now because I can just unplug it from one and plug it into another. Much more convenient than a card encased inside a computer chassis. Why can't Nvidia and AMD make stand-alone Thunderbolt GPUs?
The onboard storage in the current model is fast. Really fast. MUCH faster than a SATA SSD.
What I like most about the current model is that it's quiet. Really quiet. I wish the Ovation computer was as quiet as it is (it's ironic that Merging Technology bundles a noisy dedicated computer with their AUDIO product).
As an audio engineer I would have preferred that instead of two GPU boards and one CPU board it could be configured with one GPU board and two CPU boards, but it's now apparent that Apple thought the GPUs would do some of the CPU work so I guess the decision made sense at the time.
I agree with those who think it was bizarre to ignore it for three years, but also have to admit that it still holds up very well even today. The only obvious upgrade paths now are the new, even faster storage introduced with the new MacBook Pro and USB-C/TB3. Even without those though, at $1000 off they now present a pretty good value.
Let me tell you what this shows me. Apple has completely forgotten about the iMac, Mac Mini and Mac Pro lines for years. When I read that something like this will not be ready until next year what else can one think? We may get a new iMac in the Fall to Winter of this year. Seeing is believing. And announcements like this that are given just to try to hold the market. But that has been it. What has Apple been doing for the last 3 now 4 years? It certainly hasn't been doing desktop computers. Tim? what have you been doing? All we've heard from you is words when it comes to desktop computing. You don't "love" the Mac. The only reason you've started to do something now is because of competitive pressure and product linkage. You've never really cared about the Mac users. You're great at opening markets and distributor pipelines but you've failed miserable when it comes to computers. The iPhone and things that hook to the iPhone are the only places I've seen an innovation in for years. But who cares, I'm only one voice. The only reason many of us are still here is the operating system. With all the billions and billions in the bank and you can't find any money for engineers and product people to really push the Mac to where it should be is a crying shame.
And what do you wanna bet, after waiting all these years for a solution, the new monitor will not be compatible with the trash can. As a 30 year customer, my loyalty to Apple fades every time I go looking for a monitor, and realize Apple doesn't give a toss about folks who bought the trash can.
PCI-E is the bus to rule all bus. It will need to become modular in design for Apple's vision; and it will need to jetison Intel's anchor, Thunderbolt, to get it done.
Mdriftmeyer, have you noticed the recent shift to PCIe SSDs? I remember when they were a $10,000+ oddity, but now I'm seeing "gaming PC" companies advertise them. Will SATA see 2025, do you think? Without optical drives to plug in, and if storage goes PCIe, what use would it still have?
I clearly see no one in here is deeply invested in the history of backplane technologies from DEC and NeXT moving foward. PCI-E is the bus to rule all bus. It will need to become modular in design for Apple's vision; and it will need to jetison Intel's anchor, Thunderbolt, to get it done.
The backplane design will be key and how the power distribution is across a modular PCI-E bus design will make it possible.
PCI-E 4.0?
Is a modular design possible?
I've read that PCI-E 4.0 is due this year and will be first available on IBM Power9 Systems -- also due this year -- and they will outperform High-end Intel. The Power9 System will also support Nvidia's high-end GPUs.
Things start to get interesting in that IBM is encouraging others to use the Power9 IP to develop variants.
Could Apple be planning to build or use a variant of the Power9 for Pro Macs and Apple's Server Farms?
First reaction - April fools?
Second reaction - if real, glad it's happening - better late than never, I suppose.
For those bemoaning both the cheese grater and trash can cases, there is so much in between. Just finished building out a monster hackintosh using a beautiful Corsair windowed case - room for up to 9 drives (HD or SSD), plenty of PCI slots, fast processor and RAM - more work and less elegant than an Apple-authorized design? Yup. But I have it now. Now. Not "not in 2017". Less expensive, with a 4K monitor than anything we are likely to see from Apple.
Could it be rendered obsolete by Apple with some clever coding? Yup. But the hackintosh market is such a fringe group, why would they even take the time to alienate them?
Anyway - YMMV - it is a pleasant enough surprise to see Apple actually owning what has largely been acknowledged as a hugely ignored segment and making positive changes.
Now if we could just get them to resurrect Aperture!
True. I misread that. Mea Culpa. However, as Sog pointed out. 500,000 mac Pros is a product worth dropping when you're a company who sells millions of other Macs and ios devices, so my point is the same.
Honest questions:
if Apple sells 500,000 Mac Pros, that's a $1.5-3.5 Billion business -- is that a business worth dropping?
Apple would still need to develop and maintain macOS and Mac peripherals -- would dropping the Mac Pro really save any costs?
what do Apple employees use internally for their high-performance needs?
what computer hardware does Apple use for R&D to design and develop Apple products?
Federighi said: "I think if you use Xcode downloads as a metric, it’s possible software developers are actually our largest pro audience. It’s growing very quickly, it’s been fantastic.” -- isn't software development important to Apple and Apple's customers? *
doesn't Apple need to have some skin in the game for maintain their ability to do high performance Mac and iDevice development?
* As to software development, try generating some of the IBM/Swift packages using CLI or Xcode -- requires a bit of time on a loaded iMac 5K
My point wasn't that apple SHOULD drop Mac Pro. I don't think they should. There was a business argument to be made that they should, however because of the niche market. My point was that today's announcement proves that the hand wringing was incorrect (specifically regarding Tim Cook placing profits above all else) and they ARE making the arguments you just made (that I agree with) internally, and those arguments carried the day.
Well, this is it. The pro machines are a tiny fraction of Mac sales. Most of their pro customers are using the iMac, so I don't think they're expecting to make a great deal of money doing this. As I mentioned before, all the divisions in Apple operate their own finances, so what the Mac unit has available for product development depends on how much money they bring into the company. This new spurt is probably because they now have the money available to develop an entirely new class of machine.
Wow, the announcement of Apple coming back into the standalone monitor business is a real slap in the face to LG. I wonder if LG didn't make a piece of junk in their 5K display if Apple would be coming back.
Indeed. Or perhaps LG was always meant to be a stop-gap, or maybe Apple was dipping a toe in the water and when LG completely screwed it up (and I do mean 'completely') they decided that they would pull the whole thing back in house. I suspect you are right though.
My question is, what changed? They've known they were doing a lousy job in the pro market for the last three years? Why decide to say something NOW? Why only start investing in the pro market again now? Was someone cooking the books in terms of stats? Did they finally realize that so long as the iPad stays in line with what it is now it'll never replace a MacBook let alone a Mac Pro? Did Tim finally have a heart to heart with someone who told him that pricing the MacBook Pro out of the range of customers who they should be targeting was a bad idea?
Seriously Apple has been putting all their eggs into iOS and ignoring the Mac aside from making it thinner and thinner and more and more boring. Give me a 1 inch design and powerful hardware and software to take advantage of it, over these needlessly thin designs!
That is a good question: why now?
I can only guess, but I think they would have realised they had a problem with thermals when they tested the existing case with the next chipset they received from Intel. This would have been quite some time ago.
This announcement has come, I think, because they've had some sort of breakthrough, and they want to use it in machines next year. I'm thinking that they are looking at some sort of internal bus that will make it easier for them to build the machine as separate, easily-cooled modules. That's my guess anyway.
As a person who used to blindly follow the ludicrous norms of the PC industry, and who has developed an absolute hatred for that same industry, I really liked the goals of the 2013 Mac Pro.
If they had only provided a Retina-style display solution, I'd own one today. Instead, I have held off on buying until Apple provided a high-PPI display. They haven't. In three years.
Almost the whole lineup of Apple hardware moved to high-PPI except the professional workhorse computers that you'd have thought would've gotten that technology first. Why do you think pros have moved to iMacs? 5K iMac for photography? "Yes please", they said. "Since you're not making a Mac Pro capable of retina, I might as well buy the 5K iMac," they likely said.
Instead of positioning retina as a photographer's dream tech, it's been used for ... Uh... sexy GUI marketing...??? which was undone by iOS 7 (which utterly wastes the available resolution by using nothing but flat and detail-less bitmaps and right angles).
As a photographer/musician, high-PPI and a quietly fast computer are my priorities. Mac Pro 2013 could've been a moderate success instead of an embarrassment.
I'd love to know what went into prioritizing retina on iMac and utterly ignorin it on the "professional" workstation. I know the whole bandwidth thing was an issue, but the 2013 Mac Pro is close (if you want a PITA third-party's solution) and Apple could've built custom chips and ports like they have plenty of times in the past (like the custom driver chip in the iMac 5K itself).
I think thunderbolt as replacement for internal PCI slots makes sense for plenty of computer users. I hate dealing with internal devices in my PCs. The bulk and noise of the monolithic supercases, the hassle of opening and closing them, dealing with multiple pain-in-the-ass connectors, all the heat, and injuries sustained to my hands working in cases... The 2013 Mac Pro was ideal for certain people. Apple just failed to deliver on the base machine.
Apple should have the following:
• Mac Pro old style (future style) • Mac Pro new style (2013) • Xserve
How to build these out of components they can share among each other is a challenge for their industrial and electrical engineers. I suspect it's possible (less so with 2013 cylinder Macs, unless they make use of a solid block of aluminum as a core heat pipe in the future modular design).
Since I'm mentioning the Xserve, has any leadership at Apple considered the fact that they have all these data centers with no Apple hardware running iCloud? It's all PC hardware/software. Remember the concept of "eating your own dog food" to prove it's good product...?
As for thunderbolt, unfortunately, Intel continues to fail at delivering on thunderbolt's original promises. It should've been an optical/electrical hybrid from day one. Version 3 in the USB 3.1 connector is the closest they've been to delivering on that and it's still shy of the original goal.
I can only guess, but I think they would have realised they had a problem with thermals when they tested the existing case with the next chipset they received from Intel. This would have been quite some time ago.
This announcement has come, I think, because they've had some sort of breakthrough, and they want to use it in machines next year. I'm thinking that they are looking at some sort of internal bus that will make it easier for them to build the machine as separate, easily-cooled modules. That's my guess anyway.
On the "thermals," reading the quotes, we can just take them at their word -- the problem was with the GPUs.
I like your "some sort of breakthrough" optimism, but it seems far more likely that this is simply the result of Schiller and Federighi being sick and tired of being asked about the Mac Pro. With the added bonus of being able to start the hype train for the Late 2017 iMac Pro. Same basic form factor but TB3/USB-C. I'd also bet on integration with iPad Pro tablets being featured in that event. The timing (September?) maybe suggests Coffee Lake, which would also support the idea of a "Pro" iMac.
The design question on "modules" is how they will fit together. People upthread who seem to think "modular" means a return to the box concept are probably mistaken. More likely it is a "rethinking" of the current approach -- they are not abandoning the vision of a core component at the center of a Thunderbolt system. They are just going to "architect" it in a way that allows it to be regularly upgraded BY APPLE: Schiller says, "we want to architect it so that WE can keep it fresh with regular improvements" ... "to do something that can be supported for a long time with customers with updates and upgrades throughout the years." ... "The current Mac Pro, as we’ve said a few times, was constrained thermally and it restricted OUR ability to upgrade it."
The middle quote is the only one with any real ambiguity in this respect, but even that "with customers" clause suggests Apple is going to be the one supplying its customers with "updates and upgrades throughout the years."
I kind of envision three or four elegant, interlocking "modules" -- CPU, GPU, and a Thunderbolt-based PCIe component and/or maybe something to do with Intel's Optane memory/storage.
[...] The design question on "modules" is how they will fit together. People upthread who seem to think "modular" means a return to the box concept are probably mistaken. More likely it is a "rethinking" of the current approach -- they are not abandoning the vision of a core component at the center of a Thunderbolt system. They are just going to "architect" it in a way that allows it to be regularly upgraded BY APPLE: Schiller says, "we want to architect it so that WE can keep it fresh with regular improvements" ... "to do something that can be supported for a long time with customers with updates and upgrades throughout the years." ... "The current Mac Pro, as we’ve said a few times, was constrained thermally and it restricted OUR ability to upgrade it."
That's the way I interpreted his remarks, too.
The modular design he mentioned doesn't mean separate components. He was referring to the internal architecture.
It also sounded like he was talking about Apple being able to upgrade the internals over time, not necessarily the user.
Comments
Credit where credit is due. This was a difficult thing to do for them.
It opens up a lot of questions regarding management, resources etc but it does mark a change in approach which is overwhelmingly positive. I hope they keep it up.
The backplane design will be key and how the power distribution is across a modular PCI-E bus design will make it possible.
I dunno Mel. Among my peer group, everyone I knew used FCP up until X came out. Now most of them have Adobe subscriptions. Considering the cost of that, and how much they resent that payment model, that's saying something.
Again, this seems to be a matter of opinion. I agree that Apple (eventually) overcame the workflow and feature issues that drove away lots of people in the beginning, but the people I know who use it (because they have to -- that's what the company supplies) don't care for it. They prefer a traditional timeline. I don't use it myself, but the most common complaint I hear from co-workers is that the FCPX/iMovie editing paradigm treats audio like a conjoined twin, and trying to work picture and sound separately is more hassle than it was with FCP7.
I work in a facility that supplies MacBook Pros with FCPX to editors in the field (in house we use a combination of Avid and an awful Velocity system). Of the dozen or so using FCPX, the number who like it is zero. None. Not one.
It's not the feature set they dislike, it's the user interface. They all say they prefer a traditional timeline like MC and Premiere. They say the FCPX/iMovie approach slows them down (they're on tight deadlines) and makes it harder to work on audio and video independently of each other.
That's based on actually asking "the pros who fucking use it," who apparently don't fucking like it. That doesn't mean it's not a good product or that there's any reason someone else shouldn't like it, just that there are no absolutes when it comes to defining what is and isn't "pro."
My previous one was HUGE and weighed grunt pounds. The small size of the current one reduces the placement restrictions the old one suffered. (Rack-mount would be even better, but that seems to have gone out of style for some reason.)
Whether storage is internal or external seems utterly irrelevant. Having it external is probably even slightly better because it can be moved to other machines if desired, and is easier to service.
No PCIe slots is only an issue because some manufacturers refuse to simply provide stand-alone versions of their products. Blackmagic makes video capture and output devices in card form, but they also provide stand-alone units that connect via Thunderbolt. I've used mine on three different computers now because I can just unplug it from one and plug it into another. Much more convenient than a card encased inside a computer chassis. Why can't Nvidia and AMD make stand-alone Thunderbolt GPUs?
The onboard storage in the current model is fast. Really fast. MUCH faster than a SATA SSD.
What I like most about the current model is that it's quiet. Really quiet. I wish the Ovation computer was as quiet as it is (it's ironic that Merging Technology bundles a noisy dedicated computer with their AUDIO product).
As an audio engineer I would have preferred that instead of two GPU boards and one CPU board it could be configured with one GPU board and two CPU boards, but it's now apparent that Apple thought the GPUs would do some of the CPU work so I guess the decision made sense at the time.
I agree with those who think it was bizarre to ignore it for three years, but also have to admit that it still holds up very well even today. The only obvious upgrade paths now are the new, even faster storage introduced with the new MacBook Pro and USB-C/TB3. Even without those though, at $1000 off they now present a pretty good value.
PCI-E 4.0?
Is a modular design possible?
I've read that PCI-E 4.0 is due this year and will be first available on IBM Power9 Systems -- also due this year -- and they will outperform High-end Intel. The Power9 System will also support Nvidia's high-end GPUs.
Things start to get interesting in that IBM is encouraging others to use the Power9 IP to develop variants.
Could Apple be planning to build or use a variant of the Power9 for Pro Macs and Apple's Server Farms?
Indeed. Or perhaps LG was always meant to be a stop-gap, or maybe Apple was dipping a toe in the water and when LG completely screwed it up (and I do mean 'completely') they decided that they would pull the whole thing back in house. I suspect you are right though.
That is a good question: why now?
I can only guess, but I think they would have realised they had a problem with thermals when they tested the existing case with the next chipset they received from Intel. This would have been quite some time ago.
This announcement has come, I think, because they've had some sort of breakthrough, and they want to use it in machines next year. I'm thinking that they are looking at some sort of internal bus that will make it easier for them to build the machine as separate, easily-cooled modules. That's my guess anyway.
If they had only provided a Retina-style display solution, I'd own one today. Instead, I have held off on buying until Apple provided a high-PPI display. They haven't. In three years.
Almost the whole lineup of Apple hardware moved to high-PPI except the professional workhorse computers that you'd have thought would've gotten that technology first. Why do you think pros have moved to iMacs? 5K iMac for photography? "Yes please", they said. "Since you're not making a Mac Pro capable of retina, I might as well buy the 5K iMac," they likely said.
Instead of positioning retina as a photographer's dream tech, it's been used for ... Uh... sexy GUI marketing...??? which was undone by iOS 7 (which utterly wastes the available resolution by using nothing but flat and detail-less bitmaps and right angles).
As a photographer/musician, high-PPI and a quietly fast computer are my priorities. Mac Pro 2013 could've been a moderate success instead of an embarrassment.
I'd love to know what went into prioritizing retina on iMac and utterly ignorin it on the "professional" workstation. I know the whole bandwidth thing was an issue, but the 2013 Mac Pro is close (if you want a PITA third-party's solution) and Apple could've built custom chips and ports like they have plenty of times in the past (like the custom driver chip in the iMac 5K itself).
I think thunderbolt as replacement for internal PCI slots makes sense for plenty of computer users. I hate dealing with internal devices in my PCs. The bulk and noise of the monolithic supercases, the hassle of opening and closing them, dealing with multiple pain-in-the-ass connectors, all the heat, and injuries sustained to my hands working in cases... The 2013 Mac Pro was ideal for certain people. Apple just failed to deliver on the base machine.
Apple should have the following:
• Mac Pro old style (future style)
• Mac Pro new style (2013)
• Xserve
How to build these out of components they can share among each other is a challenge for their industrial and electrical engineers. I suspect it's possible (less so with 2013 cylinder Macs, unless they make use of a solid block of aluminum as a core heat pipe in the future modular design).
Since I'm mentioning the Xserve, has any leadership at Apple considered the fact that they have all these data centers with no Apple hardware running iCloud? It's all PC hardware/software. Remember the concept of "eating your own dog food" to prove it's good product...?
As for thunderbolt, unfortunately, Intel continues to fail at delivering on thunderbolt's original promises. It should've been an optical/electrical hybrid from day one. Version 3 in the USB 3.1 connector is the closest they've been to delivering on that and it's still shy of the original goal.
I like your "some sort of breakthrough" optimism, but it seems far more likely that this is simply the result of Schiller and Federighi being sick and tired of being asked about the Mac Pro. With the added bonus of being able to start the hype train for the Late 2017 iMac Pro. Same basic form factor but TB3/USB-C. I'd also bet on integration with iPad Pro tablets being featured in that event. The timing (September?) maybe suggests Coffee Lake, which would also support the idea of a "Pro" iMac.
The design question on "modules" is how they will fit together. People upthread who seem to think "modular" means a return to the box concept are probably mistaken. More likely it is a "rethinking" of the current approach -- they are not abandoning the vision of a core component at the center of a Thunderbolt system. They are just going to "architect" it in a way that allows it to be regularly upgraded BY APPLE: Schiller says, "we want to architect it so that WE can keep it fresh with regular improvements" ... "to do something that can be supported for a long time with customers with updates and upgrades throughout the years." ... "The current Mac Pro, as we’ve said a few times, was constrained thermally and it restricted OUR ability to upgrade it."
The middle quote is the only one with any real ambiguity in this respect, but even that "with customers" clause suggests Apple is going to be the one supplying its customers with "updates and upgrades throughout the years."
I kind of envision three or four elegant, interlocking "modules" -- CPU, GPU, and a Thunderbolt-based PCIe component and/or maybe something to do with Intel's Optane memory/storage.
The modular design he mentioned doesn't mean separate components. He was referring to the internal architecture.
It also sounded like he was talking about Apple being able to upgrade the internals over time, not necessarily the user.