macplusplus said: You exposed well the dilemma PCIe slot vs Thunderbolt. Now further digging is needed to understand why Apple may have chosen Thunderbolt 2 over PCIe slot as the expansion option in trash can MP.
Because Phil wanted to get up on stage and utter his now infamous little quip? Apple thought they were innovating, and that won out over practicality. In a way, they were innovating, I suppose, just a half-decade+ before the technology was ready to pull it off. Flying cars are innovative, but mostly useless, too. Innovation and practicality need to meet.
It's not a bad machine. It's nearly idea (original cost aside) for someone like me. It just wasn't appropriate for a good segment of the target market. If Apple were committed to eGPU on TB2, I'd probably buy one despite the performance loss. But, as others pointed out above, that doesn't cut it, necessarily, for the real heavy use pros.
mattinoz said: Then one product is a booster for iMacs, a dock for Laptops and allows the Grunt Mac Pro to be up what 100m away with optic fibre Thunderbolt.
Which then looks neat and tidy to the user, and quiet.
Sounds good in theory, but I don't think the software/hardware is really there on the whole. Maybe once TB reaches it's full potential.
You don't need the GPU to be "in the monitor" to run fibre and put the computer in a secure room. It is already done today. You can run fibre from the computer to a monitor (DisplayPort and USB or Thunderbolt). CPU / GPU / Memory is constantly moving huge amounts of bandwidth and any latency or bottlenecks in the transfer of that data will affect performance. It does not make much more sense to move the GPU down a thunderbolt cable to another location any more than putting the CPU and memory at opposite ends of Thunderbolt (the only caveat is when there are no better options -- i.e. laptop). The GPU is used much more than just graphics. "Metal 2" was designed to provide near direct access to the GPU for performance sake -- and is used for more than just graphics (GPU based computational engine; GPU-accelerated machine learning). You could also install both a GPU card (or two) and a TPU (tensors) card - all of these cards work very tightly with the CPU/Memory. I don't know what the obsession is by some people on here trying to convince the masses that it is a good idea to hamper performance by running everything across thunderbolt. Thunderbolt is great for things like large storage subsystems, but not for anything related to processors and memory. Is not this Mac Pro suppose to be for heavy "professional" use -- not as a cute toy. The only reason to obsess over having everything over Thunderbolt is to somehow justify the mistake that Apple made with the 2013 Mac Pro (IMHO). The trash can design is not a bad design, but not for professional "Mac Pro" -- more as a "Mac" (i.e. consumer-oriented headless machine - sort of a Mac Mini Pro).
PCIe 4.0 is coming out this year and 5.0 next year, I don't think Thunderbolt will ever catch up.
And then there's the fact that every Thunderbolt attached device needs it's own case and power supply, which is worse for the environment than a single case with a large efficient power supply. Why don't they like PCIe again? Seems like there more reasons to like it than not.
Is it just that the installation of cards is quite technical? Well then design a terribly clever new installation method, rather than throw out the whole system.
macplusplus said: You exposed well the dilemma PCIe slot vs Thunderbolt. Now further digging is needed to understand why Apple may have chosen Thunderbolt 2 over PCIe slot as the expansion option in trash can MP.
Because Phil wanted to get up on stage and utter his now infamous little quip? Apple thought they were innovating, and that won out over practicality. In a way, they were innovating, I suppose, just a half-decade+ before the technology was ready to pull it off. Flying cars are innovative, but mostly useless, too. Innovation and practicality need to meet.
It's not a bad machine. It's nearly idea (original cost aside) for someone like me. It just wasn't appropriate for a good segment of the target market. If Apple were committed to eGPU on TB2, I'd probably buy one despite the performance loss. But, as others pointed out above, that doesn't cut it, necessarily, for the real heavy use pros.
mattinoz said: Then one product is a booster for iMacs, a dock for Laptops and allows the Grunt Mac Pro to be up what 100m away with optic fibre Thunderbolt.
Which then looks neat and tidy to the user, and quiet.
Sounds good in theory, but I don't think the software/hardware is really there on the whole. Maybe once TB reaches it's full potential.
You don't need the GPU to be "in the monitor" to run fibre and put the computer in a secure room. It is already done today. You can run fibre from the computer to a monitor (DisplayPort and USB or Thunderbolt). CPU / GPU / Memory is constantly moving huge amounts of bandwidth and any latency or bottlenecks in the transfer of that data will affect performance. It does not make much more sense to move the GPU down a thunderbolt cable to another location any more than putting the CPU and memory at opposite ends of Thunderbolt (the only caveat is when there are no better options -- i.e. laptop). The GPU is used much more than just graphics. "Metal 2" was designed to provide near direct access to the GPU for performance sake -- and is used for more than just graphics (GPU based computational engine; GPU-accelerated machine learning). You could also install both a GPU card (or two) and a TPU (tensors) card - all of these cards work very tightly with the CPU/Memory. I don't know what the obsession is by some people on here trying to convince the masses that it is a good idea to hamper performance by running everything across thunderbolt. Thunderbolt is great for things like large storage subsystems, but not for anything related to processors and memory. Is not this Mac Pro suppose to be for heavy "professional" use -- not as a cute toy. The only reason to obsess over having everything over Thunderbolt is to somehow justify the mistake that Apple made with the 2013 Mac Pro (IMHO). The trash can design is not a bad design, but not for professional "Mac Pro" -- more as a "Mac" (i.e. consumer-oriented headless machine - sort of a Mac Mini Pro).
Apple won't put the GPU on the PCIe slot. Yes it will put it as close to CPU as possible, i.e. on the motherboard. So your effort to prove "how valuable the GPU on the PCIe slot" is redundant.
And once we have the GPU (or maybe two) on the motherboard, whether the second (or third) GPU should be on PCIe slot or on Thunderbolt enclosure is a trivial issue. Some may prefer PCIe slot some other may prefer TB3 over fiber.
bkkcanuck said: I don't know what the obsession is by some people on here trying to convince the masses that it is a good idea to hamper performance by running everything across thunderbolt. Thunderbolt is great for things like large storage subsystems, but not for anything related to processors and memory. Is not this Mac Pro suppose to be for heavy "professional" use -- not as a cute toy. The only reason to obsess over having everything over Thunderbolt is to somehow justify the mistake that Apple made with the 2013 Mac Pro (IMHO). The trash can design is not a bad design, but not for professional "Mac Pro" -- more as a "Mac" (i.e. consumer-oriented headless machine - sort of a Mac Mini Pro).
I think it is compelling for the prosumer. The true pros don't really care about it, as you say (and as I mentioned above from what my friend relayed to me). Space for big hardware and such, or lots of cables, just isn't an issue for the true heavy-use pros. The 2013 Mac Pro was aimed more at video editors and people using Photoshop and such that wanted a nice clean desk in the design agency.
I think the reason there is a the obsession with TB based peripherals, is that *IF* the bandwidth were enough, then you can truly create a one-size-fits-all type platform. You have a main box that is neat and tidy (cf. 2013 Mac Pro) and then add on any kind of modules as needed. That is attractive vs a huge box that you try to stuff everything in side of. Again, that's in theory, because in reality, it just isn't possible yet.
Also, I think Apple had more or less abandoned the true pros, going after the 'big pie slice' of the prosumer-pro oriented market. I'm still guessing the Mac's days are numbered at Apple internally. I think they just made a bad judgement as to how quickly they could move on that, combined with a lack of resources from their insane rate of growth. They've now shifted some resources back to the Mac to add maybe another decade or so. But, I think the plan is to figure out ways to transition everyone to iOS, eventually.
PCIe 4.0 is coming out this year and 5.0 next year, I don't think Thunderbolt will ever catch up.
And then there's the fact that every Thunderbolt attached device needs it's own case and power supply, which is worse for the environment than a single case with a large efficient power supply. Why don't they like PCIe again? Seems like there more reasons to like it than not.
Is it just that the installation of cards is quite technical? Well then design a terribly clever new installation method, rather than throw out the whole system.
Good point. It's not like the internal technology is just going to sit still, either.
As for the why... aside from the modularity I described above, you could also argue it's more efficient to design a power supply for each component optimally, rather than including one huge power supply that maybe 80% of the buyers won't ever use fully (but still needs to be that big for the other 20%).
And, yea, I think there is a move away from such technical capabilities. I've also heard anecdotal data that only single-digit percentages of computer buyers ever crack the case open... even in many professional settings. Machines often get handed down rather than upgraded.
macplusplus said: Apple won't put the GPU on the PCIe slot. Yes it will put it as close to CPU as possible, i.e. on the motherboard. So your effort to prove "how valuable the GPU on the PCIe slot" is redundant.
And once we have the GPU (or maybe two) on the motherboard, whether the second (or third) GPU should be on PCIe slot or on Thunderbolt enclosure is a trivial issue. Some may prefer PCIe slot some other may prefer TB3 over fiber.
You're probably right about what Apple will/won't do, though maybe they'll surprise us.
The problem is that GPUs have generally moved at a much faster pace than the rest of the system. You buy that new, latest and greatest machine, and a year later, a new GPU series is out that could save you many hours of time (which equals $$$). Then, you have to hand down that machine and buy new (if it's available, which it hasn't been from Apple). And, that's if you're big enough in scale for the hand-me-down method.
Also, TB3 over fiber gets you nothing, as it's still TB3. We'll need TB4 or TB5 before we get to fibre speeds (where it then might be as fast as current computer busses... but probably not as fast as the bus at that point in time). The only thing fibre gets you currently is location extension.
bkkcanuck said: I don't know what the obsession is by some people on here trying to convince the masses that it is a good idea to hamper performance by running everything across thunderbolt. Thunderbolt is great for things like large storage subsystems, but not for anything related to processors and memory. Is not this Mac Pro suppose to be for heavy "professional" use -- not as a cute toy. The only reason to obsess over having everything over Thunderbolt is to somehow justify the mistake that Apple made with the 2013 Mac Pro (IMHO). The trash can design is not a bad design, but not for professional "Mac Pro" -- more as a "Mac" (i.e. consumer-oriented headless machine - sort of a Mac Mini Pro).
I think it is compelling for the prosumer. The true pros don't really care about it, as you say (and as I mentioned above from what my friend relayed to me). Space for big hardware and such, or lots of cables, just isn't an issue for the true heavy-use pros. The 2013 Mac Pro was aimed more at video editors and people using Photoshop and such that wanted a nice clean desk in the design agency.
I think the reason there is a the obsession with TB based peripherals, is that *IF* the bandwidth were enough, then you can truly create a one-size-fits-all type platform. You have a main box that is neat and tidy (cf. 2013 Mac Pro) and then add on any kind of modules as needed. That is attractive vs a huge box that you try to stuff everything in side of. Again, that's in theory, because in reality, it just isn't possible yet.
Also, I think Apple had more or less abandoned the true pros, going after the 'big pie slice' of the prosumer-pro oriented market. I'm still guessing the Mac's days are numbered at Apple internally. I think they just made a bad judgement as to how quickly they could move on that, combined with a lack of resources from their insane rate of growth. They've now shifted some resources back to the Mac to add maybe another decade or so. But, I think the plan is to figure out ways to transition everyone to iOS, eventually.
PCIe 4.0 is coming out this year and 5.0 next year, I don't think Thunderbolt will ever catch up.
And then there's the fact that every Thunderbolt attached device needs it's own case and power supply, which is worse for the environment than a single case with a large efficient power supply. Why don't they like PCIe again? Seems like there more reasons to like it than not.
Is it just that the installation of cards is quite technical? Well then design a terribly clever new installation method, rather than throw out the whole system.
Good point. It's not like the internal technology is just going to sit still, either.
As for the why... aside from the modularity I described above, you could also argue it's more efficient to design a power supply for each component optimally, rather than including one huge power supply that maybe 80% of the buyers won't ever use fully (but still needs to be that big for the other 20%).
And, yea, I think there is a move away from such technical capabilities. I've also heard anecdotal data that only single-digit percentages of computer buyers ever crack the case open... even in many professional settings. Machines often get handed down rather than upgraded.
macplusplus said: Apple won't put the GPU on the PCIe slot. Yes it will put it as close to CPU as possible, i.e. on the motherboard. So your effort to prove "how valuable the GPU on the PCIe slot" is redundant.
And once we have the GPU (or maybe two) on the motherboard, whether the second (or third) GPU should be on PCIe slot or on Thunderbolt enclosure is a trivial issue. Some may prefer PCIe slot some other may prefer TB3 over fiber.
You're probably right about what Apple will/won't do, though maybe they'll surprise us.
The problem is that GPUs have generally moved at a much faster pace than the rest of the system. You buy that new, latest and greatest machine, and a year later, a new GPU series is out that could save you many hours of time (which equals $$$).
What we miss here is that this just a paradigm, not a technical necessity.Apple has certainly sound reasons to defend the new paradigm it introduces with GPU(s) on the motherboard, a further digging may reveal those reasons. The point is, Apple has succesfully proved that GPU on the PCIe slot is not the only architecture option available. It has succesfully proved that not only with Mac Pro, but also with the iMac Pro. They designed a new thermal architecture but they didn’t give up their “tradition” of the GPU on the motherboard.
macplusplus said: What we miss here is that this just a paradigm, not a technical necessity.Apple has certainly sound reasons to defend the new paradigm it introduces with GPU(s) on the motherboard, a further digging may reveal those reasons. The point is, Apple has succesfully proved that GPU on the PCIe slot is not the only architecture option available. It has succesfully proved that not only with Mac Pro, but also with the iMac Pro. They designed a new thermal architecture but they didn’t give up their “tradition” of the GPU on the motherboard.
They proved what? The design was considered a failure (admitted by Apple) by the high-end pros.
I get what you're saying... if the internal GPUs can drive the displays AND the bus is big enough externally to support the other work-horse GPU stuff and get the data back... then it doesn't matter. I don't think that has been proven at all. (Also, as another poster pointed out, GPUs aren't the only data-heavy external thing.)
macplusplus said: What we miss here is that this just a paradigm, not a technical necessity.Apple has certainly sound reasons to defend the new paradigm it introduces with GPU(s) on the motherboard, a further digging may reveal those reasons. The point is, Apple has succesfully proved that GPU on the PCIe slot is not the only architecture option available. It has succesfully proved that not only with Mac Pro, but also with the iMac Pro. They designed a new thermal architecture but they didn’t give up their “tradition” of the GPU on the motherboard.
They proved what? The design was considered a failure (admitted by Apple) by the high-end pros.
I get what you're saying... if the internal GPUs can drive the displays AND the bus is big enough externally to support the other work-horse GPU stuff and get the data back... then it doesn't matter. I don't think that has been proven at all. (Also, as another poster pointed out, GPUs aren't the only data-heavy external thing.)
No the design was never considered a failure by Apple. Apple did not admit anything, “we painted ourselves into a thermal corner” is not the admission of failure. Why would they admit? They have kernel support for dual GPUs, a core set Pro applications, immense control over the production... If that design is a failure, then why Apple repeated the same “failure” with iMac Pro putting again the GPU on the motherboard. If you’ll claim that lMac Pro too is a failure, then this discussion is too loaded.
macplusplus said: What we miss here is that this just a paradigm, not a technical necessity.Apple has certainly sound reasons to defend the new paradigm it introduces with GPU(s) on the motherboard, a further digging may reveal those reasons. The point is, Apple has succesfully proved that GPU on the PCIe slot is not the only architecture option available. It has succesfully proved that not only with Mac Pro, but also with the iMac Pro. They designed a new thermal architecture but they didn’t give up their “tradition” of the GPU on the motherboard.
They proved what? The design was considered a failure (admitted by Apple) by the high-end pros.
I get what you're saying... if the internal GPUs can drive the displays AND the bus is big enough externally to support the other work-horse GPU stuff and get the data back... then it doesn't matter. I don't think that has been proven at all. (Also, as another poster pointed out, GPUs aren't the only data-heavy external thing.)
You can buy 128GBps external pci expansion chassis today. That TB3 is 40GBps is just what that standard supports.
PCIe 4.0 is coming out this year and 5.0 next year, I don't think Thunderbolt will ever catch up.
And then there's the fact that every Thunderbolt attached device needs it's own case and power supply, which is worse for the environment than a single case with a large efficient power supply. Why don't they like PCIe again? Seems like there more reasons to like it than not.
Is it just that the installation of cards is quite technical? Well then design a terribly clever new installation method, rather than throw out the whole system.
Good point. It's not like the internal technology is just going to sit still, either.
As for the why... aside from the modularity I described above, you could also argue it's more efficient to design a power supply for each component optimally, rather than including one huge power supply that maybe 80% of the buyers won't ever use fully (but still needs to be that big for the other 20%).
And, yea, I think there is a move away from such technical capabilities. I've also heard anecdotal data that only single-digit percentages of computer buyers ever crack the case open... even in many professional settings. Machines often get handed down rather than upgraded.
Are you sure about the power supply thing? I believe it's more efficient (in terms of energy usage) to have a big power supply only running at 50% of it's capacity that several smaller ones with exactly matched capacities. You would have to over-provision all the smaller power supplies to get the same overall efficiency which doesn't seem economical.
Good point about the hand me downs, handed thought of that.
Don't get me wrong about Thunderbolt, I think it's a good thing and Apple and Intel did a good thing for the industry by developing it. I just think it is "PCIe for laptops" that they have created, and it's a misapplication to use it on desktops.
Yep, I suspect a lot of people will just skip 4.0. I also wonder why we were stuck on 3.0 so long.
Thanks to the Great GPU Shortage of 2017-18, I wonder if anything will even be made to use PCIe 4. As slowly as ATI and nVidia come out with products these days, who says they won’t just skip it, too?
And speaking of 3, why are we still on SATA 3? Did they look at M.2 and say, “Well, we better not even try”? SATA 3.2 has 16 Gb/s, but… Oh, I see that M.2 is a derivative of the SATA standard. Okay; guess they’re just moving to that entirely. Now to wait another decade for products that use it to cost less than a kidney.
macplusplus said: No the design was never considered a failure by Apple. Apple did not admit anything, “we painted ourselves into a thermal corner” is not the admission of failure. Why would they admit? They have kernel support for dual GPUs, a core set Pro applications, immense control over the production... If that design is a failure, then why Apple repeated the same “failure” with iMac Pro putting again the GPU on the motherboard. If you’ll claim that lMac Pro too is a failure, then this discussion is too loaded.
Well no, they aren't just going to come out and say it in those words. But, having a special meeting, admitting they went a different direction than the market/needs of the users, promise of a complete re-design, special focus groups to be sure they nail it next time, etc. kind of speaks something.
Again, it wasn't a fail for all markets, just the high-end pros (which should be the target market). I'm considering buying one... I'm just wavering a bit now due to eGPU changes regarding TB2, and that I wonder how long future macOS updates will be supported on it. It's still the only thing Apple makes that fits my needs aside from possibly a $6000+ iMac Pro, now, but I'd rather not have a single-purpose, all-in-one as my display (even if I wanted to spend that much).
nht said: You can buy 128GBps external pci expansion chassis today. That TB3 is 40GBps is just what that standard supports.
But, how would you connect it? Can you combine a several TB3 ports?
ascii said: Are you sure about the power supply thing? I believe it's more efficient (in terms of energy usage) to have a big power supply only running at 50% of it's capacity that several smaller ones with exactly matched capacities. You would have to over-provision all the smaller power supplies to get the same overall efficiency which doesn't seem economical.
You might be correct in terms of energy use or power-efficiency, but there is also space efficiency, cooling systems, etc. That's why I said 'you could also argue' as it's a discussion to be had. At least space-wise, it would depend on the situations and person. Some could care less if it were twice the size of a 'cheese grater' while others would just love the little cylinder on their desk and a rack full of equipment in another room, etc.
tallest skil said: Thanks to the Great GPU Shortage of 2017-18...
No doubt... grrr! I love that with all the complaining going on over the environment and energy use, we've got people wasting tons of energy creating the next bubble.
They started thinking about making a new Mac Pro, that is modular, shortly before April 2017. The timeline has not changed. They will be bringing out a new Mac Pro in 2019, after working hard on designing any and every possible bottleneck out of the thing.
It all sounds quite coherent… except for the fact that, between those start and end points, they have fully designed… and built… and sold from retail outlets, a new Mac Pro, such that, as we speak, it is on people’s desks doing work. … In early 20*18*.
Are we to believe that the iMac Pro was designed and built over a couple of weekends, using high-end parts thrown together, randomly and without even a hint of optimisation… and built, and sold?
The bottom line seems to be that it takes Apple *** 2 years *** !!! to work out how to make a machine modular, even though they themselves have already built more than one. …As in, over and above all the other work involved in designing and building an [i]Mac Pro.
So I guess is the new joke is ‘how many years does it take Apple to change a lightbulb?”
The other takeaway is that the meeting where they leaked the Black iMac was a ruse to quell discontent among the screwed over Mac Pro user base. Apparently they were hoping the hopped up iMac would shut people up- they had nothing if they are just now hiring.
Other than EFI, there is not a whole lot of difference between any whitebox PC and a Mac. We do not need Jony’s Stylings or some modular thing with proprietary connectivity that locks us into a dungeon of Apple’s capricious and fickle product plans. An updated Cheesegrater would be just fine- something capable of using standard cards, memory, etc.
l would be willing to bet that if Apple asked H-P to market a Mac version of it’s workstations they could have product ready to ship before WWDC. Apple needs to decide if they want to make computers or just be a lifestyle brand selling phones.
Man you love dreaming up nonsense conspiracies. So you think the iMP just popped out of a clamshell, fully formed? Yeah no. It’s a killer workstation with an all new internal thermal design and it too took time to produce.
But but...you can make a DIY PC in about 30 minutes on PC Part Picker for half the price and just as powerful, if not more! So why does it take Apple 2yrs to design a Mac? /s
The only real joke is the content of his posts...
Actually it is a very rational question! Designing Xeon based PCs is a well established practice. If apple keeps slipping and falling there must be something wrong at Apple. It strikes me as a project that has restarted many times due to the lack of direction and no solud design parameters. Either that or they are working with totally new non Xeon hardware.
I wonder… maybe they are having the problem that we have all had at one time or another — two opposed governing criteria. • We have to make a standard, modular, boxxy machine. • We can’t just make a standard, modular, boxxy machine. (Horrors!; this allows the possibility of a cheap version!)
Maybe they have not yet realised that they have a problem with having a coherent goal, and they think the problem is with the design work. (Releasing the iMac Pro fits this neatly; they have already convinced themselves that making AIO machines with a thin bezel is brilliant and exceptional.)
PCIe 4.0 is coming out this year and 5.0 next year, I don't think Thunderbolt will ever catch up.
And then there's the fact that every Thunderbolt attached device needs it's own case and power supply, which is worse for the environment than a single case with a large efficient power supply. Why don't they like PCIe again? Seems like there more reasons to like it than not.
Is it just that the installation of cards is quite technical? Well then design a terribly clever new installation method, rather than throw out the whole system.
What's stopping a future iteration of Thunderbolt from using PCIe 4 or 5?
The bottom line seems to be that it takes Apple *** 2 years *** !!! to work out how to make a machine modular, even though they themselves have already built more than one.
I’m growing more afraid that the new Mac Pro will simply be proprietary boxes that you plug into each other (through a proprietary stacking mechanism with a proprietary port) to make it “modular.” Modularity and expandability are different concepts.
nht said: You can buy 128GBps external pci expansion chassis today. That TB3 is 40GBps is just what that standard supports.
But, how would you connect it? Can you combine a several TB3 ports?
No, each port is typically on a TB switch so if you grab two ports on the same switch you're still bandwidth limited. Bonding two TB switches isn't doable today.
But TB4 or beyond could or it may get replaced by some other interconnect standard that's just an external 16x PCIe 5 connection with expensive and short cables...
nht said: You can buy 128GBps external pci expansion chassis today. That TB3 is 40GBps is just what that standard supports.
But, how would you connect it? Can you combine a several TB3 ports?
No, each port is typically on a TB switch so if you grab two ports on the same switch you're still bandwidth limited. Bonding two TB switches isn't doable today.
But TB4 or beyond could or it may get replaced by some other interconnect standard that's just an external 16x PCIe 5 connection with expensive and short cables...
Well, there is OCuLInk 2 (basically external PCIe 4.0) but that's pretty rare to see, even in the server space.
nht said: You can buy 128GBps external pci expansion chassis today. That TB3 is 40GBps is just what that standard supports.
But, how would you connect it? Can you combine a several TB3 ports?
No, each port is typically on a TB switch so if you grab two ports on the same switch you're still bandwidth limited. Bonding two TB switches isn't doable today.
But TB4 or beyond could or it may get replaced by some other interconnect standard that's just an external 16x PCIe 5 connection with expensive and short cables...
It is possible, but the cables would have to be fibre only -- and my guess is bloody expensive at this point due to the digital signal processing speed needed to reach speeds like that - and having more cost than benefit for something like a Mac Pro.
Morphroth said: Are we to believe that the iMac Pro was designed and built over a couple of weekends, using high-end parts thrown together, randomly and without even a hint of optimisation… and built, and sold?
No, I think the iMac Pro was intended to be the end of the line Mac to appease video editors and developer types for the next few years... in other words, they didn't intend to make another Mac Pro.
ZestyMordant said: What's stopping a future iteration of Thunderbolt from using PCIe 4 or 5?
True, but the question is if it will always lag behind, just due to technology and physics limitations (or budget). I'm not sure why we can't have the optical version of TB right now that Intel originally sold the concept on. Is there some physical thing stopping that? Would it just cost too much? Or, are they just rolling it out slowly, version by version, to maximize the money they are making on it?
Or, is it a problem like desktop/mobile performance, where the mobile keeps getting faster, but will never catch the desktop because it does too, and doesn't have the power-limitations.
nht said: No, each port is typically on a TB switch so if you grab two ports on the same switch you're still bandwidth limited. Bonding two TB switches isn't doable today.
But TB4 or beyond could or it may get replaced by some other interconnect standard that's just an external 16x PCIe 5 connection with expensive and short cables...
I meant that in regards to the mentioned "128GBps external pci expansion chassis"... that isn't possible for the Mac, unless you could bond multiple TB ports. You're just meaning it's being done in other industries, right?
Comments
And once we have the GPU (or maybe two) on the motherboard, whether the second (or third) GPU should be on PCIe slot or on Thunderbolt enclosure is a trivial issue. Some may prefer PCIe slot some other may prefer TB3 over fiber.
I think the reason there is a the obsession with TB based peripherals, is that *IF* the bandwidth were enough, then you can truly create a one-size-fits-all type platform. You have a main box that is neat and tidy (cf. 2013 Mac Pro) and then add on any kind of modules as needed. That is attractive vs a huge box that you try to stuff everything in side of. Again, that's in theory, because in reality, it just isn't possible yet.
Also, I think Apple had more or less abandoned the true pros, going after the 'big pie slice' of the prosumer-pro oriented market. I'm still guessing the Mac's days are numbered at Apple internally. I think they just made a bad judgement as to how quickly they could move on that, combined with a lack of resources from their insane rate of growth. They've now shifted some resources back to the Mac to add maybe another decade or so. But, I think the plan is to figure out ways to transition everyone to iOS, eventually.
Good point. It's not like the internal technology is just going to sit still, either.
As for the why... aside from the modularity I described above, you could also argue it's more efficient to design a power supply for each component optimally, rather than including one huge power supply that maybe 80% of the buyers won't ever use fully (but still needs to be that big for the other 20%).
And, yea, I think there is a move away from such technical capabilities. I've also heard anecdotal data that only single-digit percentages of computer buyers ever crack the case open... even in many professional settings. Machines often get handed down rather than upgraded.
You're probably right about what Apple will/won't do, though maybe they'll surprise us.
The problem is that GPUs have generally moved at a much faster pace than the rest of the system. You buy that new, latest and greatest machine, and a year later, a new GPU series is out that could save you many hours of time (which equals $$$). Then, you have to hand down that machine and buy new (if it's available, which it hasn't been from Apple). And, that's if you're big enough in scale for the hand-me-down method.
Also, TB3 over fiber gets you nothing, as it's still TB3. We'll need TB4 or TB5 before we get to fibre speeds (where it then might be as fast as current computer busses... but probably not as fast as the bus at that point in time). The only thing fibre gets you currently is location extension.
I get what you're saying... if the internal GPUs can drive the displays AND the bus is big enough externally to support the other work-horse GPU stuff and get the data back... then it doesn't matter. I don't think that has been proven at all. (Also, as another poster pointed out, GPUs aren't the only data-heavy external thing.)
Good point about the hand me downs, handed thought of that.
Don't get me wrong about Thunderbolt, I think it's a good thing and Apple and Intel did a good thing for the industry by developing it. I just think it is "PCIe for laptops" that they have created, and it's a misapplication to use it on desktops.
Yep, I suspect a lot of people will just skip 4.0. I also wonder why we were stuck on 3.0 so long.
And speaking of 3, why are we still on SATA 3? Did they look at M.2 and say, “Well, we better not even try”? SATA 3.2 has 16 Gb/s, but… Oh, I see that M.2 is a derivative of the SATA standard. Okay; guess they’re just moving to that entirely. Now to wait another decade for products that use it to cost less than a kidney.
Again, it wasn't a fail for all markets, just the high-end pros (which should be the target market). I'm considering buying one... I'm just wavering a bit now due to eGPU changes regarding TB2, and that I wonder how long future macOS updates will be supported on it. It's still the only thing Apple makes that fits my needs aside from possibly a $6000+ iMac Pro, now, but I'd rather not have a single-purpose, all-in-one as my display (even if I wanted to spend that much).
But, how would you connect it? Can you combine a several TB3 ports?
You might be correct in terms of energy use or power-efficiency, but there is also space efficiency, cooling systems, etc. That's why I said 'you could also argue' as it's a discussion to be had. At least space-wise, it would depend on the situations and person. Some could care less if it were twice the size of a 'cheese grater' while others would just love the little cylinder on their desk and a rack full of equipment in another room, etc.
No doubt... grrr!
I love that with all the complaining going on over the environment and energy use, we've got people wasting tons of energy creating the next bubble.
They started thinking about making a new Mac Pro, that is modular, shortly before April 2017. The timeline has not changed. They will be bringing out a new Mac Pro in 2019, after working hard on designing any and every possible bottleneck out of the thing.
It all sounds quite coherent… except for the fact that, between those start and end points, they have fully designed… and built… and sold from retail outlets, a new Mac Pro, such that, as we speak, it is on people’s desks doing work. … In early 20*18*.
Are we to believe that the iMac Pro was designed and built over a couple of weekends, using high-end parts thrown together, randomly and without even a hint of optimisation… and built, and sold?
The bottom line seems to be that it takes Apple *** 2 years *** !!! to work out how to make a machine modular, even though they themselves have already built more than one. …As in, over and above all the other work involved in designing and building an [i]Mac Pro.
Maybe they have not yet realised that they have a problem with having a coherent goal, and they think the problem is with the design work. (Releasing the iMac Pro fits this neatly; they have already convinced themselves that making AIO machines with a thin bezel is brilliant and exceptional.)
But TB4 or beyond could or it may get replaced by some other interconnect standard that's just an external 16x PCIe 5 connection with expensive and short cables...
True, but the question is if it will always lag behind, just due to technology and physics limitations (or budget). I'm not sure why we can't have the optical version of TB right now that Intel originally sold the concept on. Is there some physical thing stopping that? Would it just cost too much? Or, are they just rolling it out slowly, version by version, to maximize the money they are making on it?
Or, is it a problem like desktop/mobile performance, where the mobile keeps getting faster, but will never catch the desktop because it does too, and doesn't have the power-limitations.
I meant that in regards to the mentioned "128GBps external pci expansion chassis"... that isn't possible for the Mac, unless you could bond multiple TB ports. You're just meaning it's being done in other industries, right?