Tailosive bruh here is mentioning FOUR GPU stacks....Let's say TB4 doubles bandwidth as every TB generation has, that still seems like a stretch, unless you're manually connecting each stack with a direct cable. Otherwise stack 1 has to carry all the bandwidth for the next three GPUs and all other components...
Just from a bandwidth flow standpoint, this seems nuts.
One thing I could think of is not use TB4 at all - just go native PCI-E with as many lanes as it needs to make this work.
I don't understand all the doubting Thomas's on this thread. Seems like a great idea.
As long as the connectors have enough bandwidth I don't see how it wouldn't work. Thermal management can be customized to each module, pulling in cool air from the sides and shooting it out the back. I'm willing to bet that multiple, small fans, running at partial strength, is much quieter than one or two giant fans pumping air through a giant rectangle; and in this case fans would only be used when needed, not the case with a large rectangle like the cheesegrater. Also, a thick rectangle doesn't make a lot of sense given that the hardware components are largely 2 dimentional; an array of two dimentional modules does make a lot of sense, however.
And of course the other obvious benefits.. highly customizable in function and cost, high performance capability, and it will lookamazing.
Large fans are quieter. They spin slower and move more air.
The holy grail of unnecessary bespoke chassis and cooling nightmares.
The convenience of the stackable modules is wholly out of whack with how often you'll need to be modifying the stack, and would cause huge downstream problems that are way, way more detrimental than convenience.
Form over function to the nth degree.
Therefore the design is confirmed, no?
Obviously not, I already said I'll believe it when I see it. I have faith that Apple are smarter than this, the rumour is a flight of fancy from someone without a clue.
I believe you may have missed the winkyface...
I saw it, but don’t understand what it’s meant to mean.
This is all based on an old patent of Apple's, where they considered a stackable design. People want to be able to put in their own full size PCIe 3.0 video card, and memory and storage and this plan won't allow that. It learns little from Apple's failures with the current Mac Pro design.
Is that a fact or assumption? Even if this rumor is true, how do we know that the modules wouldn't be big enough to accommodate full size cards? There is no doubt that could contain user-swappable RAM and storage if Apple wanted to go that way.
Nothing in this is inside information. We’ll see more of these attempts to guess the architecture as bits of information filter out in the coming months leading up to WWDC — this one was prompted by the 6K3K news, which makes perfect sense and is likely accurate.
I don't buy this report for the simple reason of the thermal issues they cornered themselves into last time.
A modular system can control for thermal better than an internally expandable one that could have no GPU or multiple beefy GPUs.
A mini only has to worry about the thermals for the CPU. The eGPU chassis has its own power and thermal design and limits.
Wouldn't it make more sense to build in a thermal "chimney" so that a single fan module could control the airflow through the entire stack? That is, if someone were actually designing something like this. Otherwise I don't see how a tight stack of hot-running modules would make any sense. Each one would have to be blowing air horizontally through their own chassis.
You mean like the trash can MacPro? I have one and it is awesome but...
Just because the MacPro was a failure doesn't mean that every design element of it was.
This has got to be a joke. A recent GPU is already larger than a Mac Mini. But I guess we won't be allowed to use those anyway.
Not shipping till late 2020 is probably not a joke.
Something's not right here. Apple traditionally hates cable mess (the main advertising point of the original iMac). A modular design has SO much duplication in chassis materials, fixtures, fittings, power and cooling. And assembly. These things would be very costly to make.
It would be SO simple to make a modern cheese cutter. With SSD and no optical it would be smaller too, or leave room for even more expansion. There must be a reason why its taking so long.
Could the delay be that they're waiting to go Arm? These enclosure designs last for years. The cheese grater lasted 10 years, G3/G4 7 years etc...
Plus they had their fingers severely burned with the embarrassment of the current model which is in a sense 'modular'. It sounds like almost everyone on here wants an updated cheese grater and it seems so obvious that I can't imagine why Apple would make another, potentially humiliating, 'too clever' mistake. Just give the people what they want and need.
The cheese grater designs were expensive but I remember speccing PC's to similar levels and they very often came out cheaper on a like-for-like basis,. And I'll bet there aren't many 10 year old PC workstations still working away.
*sigh* Are PCIe slots such a hard concept for Apple to grasp? I don't want a stack of pancakes, I want a Mac with slots so industry standard components can be used and vendors are encouraged to provide Mac support for their, in many cases, existing hardware.
Seriously, why is this so hard? Why does Apple seemingly need to take it all to the Nth degree?
I get it, "pretty and slick", but some people don't care and would like a functional, expandable Mac.
Maybe they see how well the Mac Pro tower is doing, all these years later, and still going because of its modular design and are frightened on the lack of return in dollars to them for such a well designed and elegantly expandable machine.
I am a professional video editor and photographer. Have been using macs since power PC. The software/hardware support has been such a failure for the the pro market since final cut pro 10. Even the manager at the apple store told me recently-while I was complaining about the poor service considering the tens of thousands ive spent over the years-apple doesn't care about pro users any more, it makes luxury products (actual quote verbatim).
Either way, I still own Apple products, but all my professional work has long since transferred to PC where I spend half as much, get three times more and am not constantly dissapointed.
I do not make my living with my Macs, but use them in ways similar to many Pro market customers. I edit photos and video and am picky about it having previously worked as a Pro years ago. I also shoot and edit video as well as compose and record music for my own enjoyment.
This is all based on an old patent of Apple's, where they considered a stackable design. People want to be able to put in their own full size PCIe 3.0 video card, and memory and storage and this plan won't allow that. It learns little from Apple's failures with the current Mac Pro design.
Is that a fact or assumption? Even if this rumor is true, how do we know that the modules wouldn't be big enough to accommodate full size cards? There is no doubt that could contain user-swappable RAM and storage if Apple wanted to go that way.
A full length PCIe card is 12.3 inches long and 4.2 tall including the contacts. Many GPU cards are also double-width, about 1.75 inches. So assuming there is a power bus that can provide 300 W of DC power (and the eGPU doesn’t need its own power cord and power supply), an eGPU box would be something like 13” x 5” x 2”. If it needs an internal power supply, it’ll be about 30% larger, and users will have another cable to stress over.
This is all ok if the interconnect is something x40 PCIe of bandwidth, plus something like 800 W of power, to connect something like 4 boxes. GPUs eat up a lot of I/O and Watts in a hurry. The CPU box could be about the same size, possibly support 2 CPU sockets and 8 RAM slots, but the power supply would be in its own box.
The issue is that a connector with >40 lanes of PCIe, that is external, is rather sporty and likely prone to flakiness. That’s a lot of pins and contacts. Maybe if they do fiber optic, but why spend the effort, when the system isn’t really an improvement over 2 CPU sockets and 3 PCIe slots in a big box, and will only be more expensive? And, Apple is rather choosy about what goes inside its computers and a noisy and power hungry 3rd party GPU is antithetical to their design ethos.
Not only that, the target market doesn’t really care about it. They just want it to be able to fit on top of a desk, have a minimum of cabling, use less power than a 110V 15A outlet can provide, have about 1000 W of computing in it, and be flexible so that it can be adapted to have the most computing perf for the power on a year by year basis.
All this is very strange on Apple’s end. They could have made the 2013 form factor work as long as they updated it. It would have been fine if it was updated in lockstep with contemporary components, especially if its integration made it more affordable, but none of that happened. Dropping the displays is even more vexing as they had an entire MBP market to sell into. Then, the purposeful vagueness of the new Mac Pro is not helpful, and indicative that they really didn’t know what they were going to do with it in 2017. Not knowing the market such that they formed a “Pro” team to figure out what to do is more damning than praiseworthy too.
I don't buy this report for the simple reason of the thermal issues they cornered themselves into last time.
A modular system can control for thermal better than an internally expandable one that could have no GPU or multiple beefy GPUs.
A mini only has to worry about the thermals for the CPU. The eGPU chassis has its own power and thermal design and limits.
Wouldn't it make more sense to build in a thermal "chimney" so that a single fan module could control the airflow through the entire stack? That is, if someone were actually designing something like this. Otherwise I don't see how a tight stack of hot-running modules would make any sense. Each one would have to be blowing air horizontally through their own chassis.
You mean like the trash can MacPro? I have one and it is awesome but...
Just because the MacPro was a failure doesn't mean that every design element of it was.
Maybe not but where they stated they painted themselves into a thermal corner is probably in that failure category...
I don't buy this report for the simple reason of the thermal issues they cornered themselves into last time.
A modular system can control for thermal better than an internally expandable one that could have no GPU or multiple beefy GPUs.
A mini only has to worry about the thermals for the CPU. The eGPU chassis has its own power and thermal design and limits.
Wouldn't it make more sense to build in a thermal "chimney" so that a single fan module could control the airflow through the entire stack? That is, if someone were actually designing something like this. Otherwise I don't see how a tight stack of hot-running modules would make any sense. Each one would have to be blowing air horizontally through their own chassis.
You mean like the trash can MacPro? I have one and it is awesome but...
Just because the MacPro was a failure doesn't mean that every design element of it was.
Maybe not but where they stated they painted themselves into a thermal corner is probably in that failure category...
Fair enough. Personally I don't expect Apple to adopt this theoretically design anyway. On the other hand, as with everyone else, I can't understand why they haven't just released a big elegant chassis and declared victory already.
I don't buy this report for the simple reason of the thermal issues they cornered themselves into last time.
A modular system can control for thermal better than an internally expandable one that could have no GPU or multiple beefy GPUs.
A mini only has to worry about the thermals for the CPU. The eGPU chassis has its own power and thermal design and limits.
Wouldn't it make more sense to build in a thermal "chimney" so that a single fan module could control the airflow through the entire stack? That is, if someone were actually designing something like this. Otherwise I don't see how a tight stack of hot-running modules would make any sense. Each one would have to be blowing air horizontally through their own chassis.
So you want a single fan module to blow air thru a "thermal chimney" that cools ALL modules...?
I would assume you are envisioning this as a bottom intake, top exhaust airflow...?
But guess what, the modules have stuff in them (mostly horizontal PCBs)...! Your vertical "thermal chimney" has just become an awkward meandering tightly packed maze for the airflow, one which only becomes hotter & more restrictive the further up the module stack the airflow goes......
So, in the case of stacked modules, individual airflow & control (in a horizontal movement) would be preferred...
This is all based on an old patent of Apple's, where they considered a stackable design. People want to be able to put in their own full size PCIe 3.0 video card, and memory and storage and this plan won't allow that. It learns little from Apple's failures with the current Mac Pro design.
Is that a fact or assumption? Even if this rumor is true, how do we know that the modules wouldn't be big enough to accommodate full size cards? There is no doubt that could contain user-swappable RAM and storage if Apple wanted to go that way.
A full length PCIe card is 12.3 inches long and 4.2 tall including the contacts. Many GPU cards are also double-width, about 1.75 inches. So assuming there is a power bus that can provide 300 W of DC power (and the eGPU doesn’t need its own power cord and power supply), an eGPU box would be something like 13” x 5” x 2”. If it needs an internal power supply, it’ll be about 30% larger, and users will have another cable to stress over.
This is all ok if the interconnect is something x40 PCIe of bandwidth, plus something like 800 W of power, to connect something like 4 boxes. GPUs eat up a lot of I/O and Watts in a hurry. The CPU box could be about the same size, possibly support 2 CPU sockets and 8 RAM slots, but the power supply would be in its own box.
The issue is that a connector with >40 lanes of PCIe, that is external, is rather sporty and likely prone to flakiness. That’s a lot of pins and contacts. Maybe if they do fiber optic, but why spend the effort, when the system isn’t really an improvement over 2 CPU sockets and 3 PCIe slots in a big box, and will only be more expensive? And, Apple is rather choosy about what goes inside its computers and a noisy and power hungry 3rd party GPU is antithetical to their design ethos.
Not only that, the target market doesn’t really care about it. They just want it to be able to fit on top of a desk, have a minimum of cabling, use less power than a 110V 15A outlet can provide, have about 1000 W of computing in it, and be flexible so that it can be adapted to have the most computing perf for the power on a year by year basis.
All this is very strange on Apple’s end. They could have made the 2013 form factor work as long as they updated it. It would have been fine if it was updated in lockstep with contemporary components, especially if its integration made it more affordable, but none of that happened. Dropping the displays is even more vexing as they had an entire MBP market to sell into. Then, the purposeful vagueness of the new Mac Pro is not helpful, and indicative that they really didn’t know what they were going to do with it in 2017. Not knowing the market such that they formed a “Pro” team to figure out what to do is more damning than praiseworthy too.
Well put. My thoughts almost exactly, especially re: the market and the strangeness of it all.
I’ve nothing to add, except to say this will probably be a case study for business school someday.
I posted this about a year ago: "'Modular' means that it will be rackable or stackable in some way. Like a couple of MacMinis with a GPU, SuperDrive, a multi-Terabyte drive or two and an Airport Extreme one on top of the other, but with the same form factor and a clever way of connecting them. Notwithstanding the Xserve, a rack would be very non-Apple, so I expect something much more elegant than that. Better too than the RackMac MacMini product. Also expect the CPU module to be water or oil cooled. I'll wait 12-15 months for that!." Please send grateful contributions via AppleInsider :-)
I'll believe it when I see it. Has the distinct whiff of bullshit.
I have big, big concerns about bandwidth on this entire concept.
I have big concerns about what the hell is going on here. It’s getting ridiculous.
Not really. They explained in detail in the Techcrunch/media event why they had to start over. That was about two years ago. Two years for a new pro from scratch isn’t ridiculous.
Yeah, it is. If you read the comments, not only here, but on other Mac centric sites, as well as the site for FCP, and others, you’ll see that the large majority of us don’t particularly want a new Mac Pro from scratch.
what we want is a more conventional machine that’s a continuation of the older cheese grater design. It doesn’t have to be the exact same case outside. The dimensions can be a bit different for the sake of new technologies, if necessary. But the old case was so well designed in the first place, and as a result, so versatile, that they could, and did, rearrange the interior in major ways several times, starting from the original G5, for which it was first designed.
apple could have, which is something we wanted, come up with a modernized version, and they could have done that three years ago, well after it should have been very clear to them that the trash can model was a failure. A failure, as they even acknowledged, because they failed to understand, during its design, that they walled themselves in (literally) with that overly small case, which didn’t have enough thermal capacity to redo their incorrect understanding of where pros were going, and how they used their machines. This is what they’ve admitted publically.
instead of waiting 5 years before responding, they should have responded a couple of years after the introduction, when the problems should have become obvious to them, as it was to everyone else. In addition, while they cogitated over it, they could have upgraded that model, yearly. Newer CPUs and GPUs were advancing, and using the same amount of power, and throwing off the same amount of heat. There’s been no response from them as to why they left this machine, without upgrades, even now. It’s almost as though they were so startled over this flub, that they descended into a fugue, where they couldn’t do anything.
even now, we have little idea as to what they’re doing. We see these concept designs which, frankly, mean nothing. I’m seeing designs that are so small as to be useless. Do we really want a stack of Mac Mini-like components that plug into each other? It’s different, as the trash can model is. But will it corral them into another dead end design that can’t be “properly” upgraded? Are we going to have to buy an Apple module for a GPU? Or anything else? If there’s no PCI bus, as some have conjectured, how do we connect other devices that need to attach directly to that bus? Is Apple, again, reinventing the wheel? To what purpose? Is this really go8ng to help productivity, as they’re claiming as their goal?
and talking about that - productivity isn’t related to the physical design at all. It’s related to performance and connectivity. How they get there isn’t important. That they get there is. That they get there quickly matters, because if they don’t, there may not be enough high level Mac users left to care.
If the definition of New Mac Pro is 'modular' and they expect me to buy only their Apple hardware, they just lost a 40 year mac user.. My 2015 mac pro, while cute, is less than desirable. Faulty thunderbolt lanes, It's video cards, KILLED two 4K LG monitors. All connectors jammed into one tiny area. You need to have children's hands to get in there, with out knocking 2 or more connectors out. Time for 'Hackintosh'. My 2015 has been nothing but a head ache, video cards, harnesses, logic boards all replaced. Quality control - sell the units, then give users a hard time about getting it repaired.
Comments
Just from a bandwidth flow standpoint, this seems nuts.
One thing I could think of is not use TB4 at all - just go native PCI-E with as many lanes as it needs to make this work.
It would be SO simple to make a modern cheese cutter. With SSD and no optical it would be smaller too, or leave room for even more expansion. There must be a reason why its taking so long.
Could the delay be that they're waiting to go Arm? These enclosure designs last for years. The cheese grater lasted 10 years, G3/G4 7 years etc...
Plus they had their fingers severely burned with the embarrassment of the current model which is in a sense 'modular'. It sounds like almost everyone on here wants an updated cheese grater and it seems so obvious that I can't imagine why Apple would make another, potentially humiliating, 'too clever' mistake. Just give the people what they want and need.
The cheese grater designs were expensive but I remember speccing PC's to similar levels and they very often came out cheaper on a like-for-like basis,. And I'll bet there aren't many 10 year old PC workstations still working away.
Seriously, why is this so hard? Why does Apple seemingly need to take it all to the Nth degree?
I get it, "pretty and slick", but some people don't care and would like a functional, expandable Mac.
Maybe they see how well the Mac Pro tower is doing, all these years later, and still going because of its modular design and are frightened on the lack of return in dollars to them for such a well designed and elegantly expandable machine.
This is all ok if the interconnect is something x40 PCIe of bandwidth, plus something like 800 W of power, to connect something like 4 boxes. GPUs eat up a lot of I/O and Watts in a hurry. The CPU box could be about the same size, possibly support 2 CPU sockets and 8 RAM slots, but the power supply would be in its own box.
The issue is that a connector with >40 lanes of PCIe, that is external, is rather sporty and likely prone to flakiness. That’s a lot of pins and contacts. Maybe if they do fiber optic, but why spend the effort, when the system isn’t really an improvement over 2 CPU sockets and 3 PCIe slots in a big box, and will only be more expensive? And, Apple is rather choosy about what goes inside its computers and a noisy and power hungry 3rd party GPU is antithetical to their design ethos.
Not only that, the target market doesn’t really care about it. They just want it to be able to fit on top of a desk, have a minimum of cabling, use less power than a 110V 15A outlet can provide, have about 1000 W of computing in it, and be flexible so that it can be adapted to have the most computing perf for the power on a year by year basis.
All this is very strange on Apple’s end. They could have made the 2013 form factor work as long as they updated it. It would have been fine if it was updated in lockstep with contemporary components, especially if its integration made it more affordable, but none of that happened. Dropping the displays is even more vexing as they had an entire MBP market to sell into. Then, the purposeful vagueness of the new Mac Pro is not helpful, and indicative that they really didn’t know what they were going to do with it in 2017. Not knowing the market such that they formed a “Pro” team to figure out what to do is more damning than praiseworthy too.
I would assume you are envisioning this as a bottom intake, top exhaust airflow...?
But guess what, the modules have stuff in them (mostly horizontal PCBs)...! Your vertical "thermal chimney" has just become an awkward meandering tightly packed maze for the airflow, one which only becomes hotter & more restrictive the further up the module stack the airflow goes......
So, in the case of stacked modules, individual airflow & control (in a horizontal movement) would be preferred...
I’ve nothing to add, except to say this will probably be a case study for business school someday.
what we want is a more conventional machine that’s a continuation of the older cheese grater design. It doesn’t have to be the exact same case outside. The dimensions can be a bit different for the sake of new technologies, if necessary. But the old case was so well designed in the first place, and as a result, so versatile, that they could, and did, rearrange the interior in major ways several times, starting from the original G5, for which it was first designed.
apple could have, which is something we wanted, come up with a modernized version, and they could have done that three years ago, well after it should have been very clear to them that the trash can model was a failure. A failure, as they even acknowledged, because they failed to understand, during its design, that they walled themselves in (literally) with that overly small case, which didn’t have enough thermal capacity to redo their incorrect understanding of where pros were going, and how they used their machines. This is what they’ve admitted publically.
instead of waiting 5 years before responding, they should have responded a couple of years after the introduction, when the problems should have become obvious to them, as it was to everyone else. In addition, while they cogitated over it, they could have upgraded that model, yearly. Newer CPUs and GPUs were advancing, and using the same amount of power, and throwing off the same amount of heat. There’s been no response from them as to why they left this machine, without upgrades, even now. It’s almost as though they were so startled over this flub, that they descended into a fugue, where they couldn’t do anything.
even now, we have little idea as to what they’re doing. We see these concept designs which, frankly, mean nothing. I’m seeing designs that are so small as to be useless. Do we really want a stack of Mac Mini-like components that plug into each other? It’s different, as the trash can model is. But will it corral them into another dead end design that can’t be “properly” upgraded? Are we going to have to buy an Apple module for a GPU? Or anything else? If there’s no PCI bus, as some have conjectured, how do we connect other devices that need to attach directly to that bus? Is Apple, again, reinventing the wheel? To what purpose? Is this really go8ng to help productivity, as they’re claiming as their goal?
and talking about that - productivity isn’t related to the physical design at all. It’s related to performance and connectivity. How they get there isn’t important. That they get there is. That they get there quickly matters, because if they don’t, there may not be enough high level Mac users left to care.