Pixar's biggest competitive advantage now is its ability to use this math-driven technology not to make better shapes but to tell better stories. DeRose and Pixar aren't sitting on their laurels. "Somewhere out there, a brilliant kid and their friends are working in their garage" using and improving on tools like Blender, DeRose tells the assembled children and adults at MoMath. "They will be the next Pixar."
Interesting that they admit Blender has nearly all the capabilities of Pixar's own software.
I fully appreciate the benefits of Apple's "shed the old, embrace the new" approach to interface design, but in the case of a "pro" machine, I wonder if pushing everything outside the box is a good idea?
A lot of what we've traditionally relied on internal devices to accomplish can now be done with external units, but it's more hassle, more expensive, and in SOME cases, all but impossible (try finding an external solution that's equivalent to a top-of-the-line NVidia GPU/video card). External solutions are usually less capable and more expensive than their PCIe counterparts [...]
If Dan is right, PCIe functionality will be pushed outside the box, via Thunderbolt. Apart from that, it's just the optical drive. Everything else stays inside, obviously. I would hope that Apple would not do this (move PCIe outside the box) unless anything that could formerly be added via PCIe can now, or soon will be, possible via Thunderbolt.
But you've definitely hit upon one limitation that early adopters will likely face. Dan doesn't point out the fact that the current Ivy Bridge with its Cactus Ridge Thunderbolt controllers doesn't have the bandwidth to be useful for high-end external GPUs, nor do the just-announced Redwood Ridge Thunderbolt controllers for Haswell, at least according to this person:
So it won't be until the Falcon Ridge Thunderbolt controllers that high-end external GPUs will be a reality via Thunderbolt. So early adopters will have to accept a limitation that won't be remedied until a major refresh in 2015, when the Falcon Ridge Thunderbolt controller would most likely make its way to the new Mac Pro.
In short, if they do this now, Apple will be way out on a limb for at least two years. But Intel will be out there with them, and maybe it will scare Microsoft into dealing with this problem. This isn't Firewire. I'd love to know what Dan thinks about the whole "Thunderbolt is the new Firewire" (quoting one of the comments on the article linked above) meme that is bound to take hold in the wake of Apple doing this.
You have to wonder if Apple will include two internal GPU slots, thereby removing this limitation, since then you wouldn't need Thunderbolt to handle a second GPU/video card.
Mac Pro's are the only Mac's I buy now as they are hardware configurable. If they drop the Pro I will reluctantly drop OSX (still my favourite OS). I want some control over my machines. Apple can't have it all.
Well you can always build a Hackintosh if that happens, they actually work quite well, you can have faster CPU's, more memory, workstation level GPU cards for a fraction of the cost of a Mac Pro.
Couldn't have said it better. An iMac or Thunderbolt connected setup cannot be taken seriously at the pro level unless your needs are very specific or you don't mind clutter. I just love the comments of people who have never worked in a professional space. Have you ever even seen a pro level graphics card? They are bigger than a Mac Mini, draw large amounts of power, and would cause an iMac to explode if it were possible to shoe horn one in, let alone cool it. That is only one subsystem. For the reasons stated above, I have no desire to get into a system like the one described above, a mess. Please, a Mac Pro tower with Pci 3 and an open selection of graphics cards. Thunderbolt 1 and eventually 2 for peripherals, Firewire for legacy, and plenty of fast RAM. And it wouldn't cost $30k...
And here's something I else: the MBP gets pretty hot at times, even with the fan kicking in. That's the nature of any decently powerful laptop that is your centerpiece. It's already atop a vented stand with its extra fan but that's just venting and blowing on the outside of the case, which helps but doesn't address the circumstance, which is powerful components jammed close together with minimal ventilation and a teeny, cursory fan. Apple has done amazing things in their quest for making things more and more powerful in thinner and thinner cases but at some cost. I've always felt that, during these past few years of lack of modern desktop choices, if they made a MBP that was an inch thick I'd be the first in line. I mean, if all we're being offered is laptops why not make one with a more powerful hotter than low power chip and motherboard, make it thick enough to vent and fan well and there's no reason it would have to be much heavier.
Well, then there's that clutter thing, yeah... : )
If Dan is right, PCIe functionality will be pushed outside the box, via Thunderbolt. Apart from that, it's just the optical drive. Everything else stays inside, obviously. I would hope that Apple would not do this (move PCIe outside the box) unless anything that could formerly be added via PCIe can now, or soon will be, possible via Thunderbolt.
But you've definitely hit upon one limitation that early adopters will likely face. Dan doesn't point out the fact that the current Ivy Bridge with its Cactus Ridge Thunderbolt controllers doesn't have the bandwidth to be useful for high-end external GPUs, nor do the just-announced Redwood Ridge Thunderbolt controllers for Haswell, at least according to this person:
So it won't be until the Falcon Ridge Thunderbolt controllers that high-end external GPUs will be a reality via Thunderbolt. So early adopters will have to accept a limitation that won't be remedied until a major refresh in 2015, when the Falcon Ridge Thunderbolt controller would most likely make its way to the new Mac Pro.
In short, if they do this now, Apple will be way out on a limb for at least two years. But Intel will be out there with them, and maybe it will scare Microsoft into dealing with this problem. This isn't Firewire. I'd love to know what Dan thinks about the whole "Thunderbolt is the new Firewire" (quoting one of the comments on the article linked above) meme that is bound to take hold in the wake of Apple doing this.
You have to wonder if Apple will include two internal GPU slots, thereby removing this limitation, since then you wouldn't need Thunderbolt to handle a second GPU/video card.
Nor the fact that an external GPGPU System only adds to the complexity of the entire system which will require a separate cooling solution and power solution thus driving the price up by about $1k.
Saying they calibrate is partly marketing speech. I'm sure they have a good system in place, but the statement doesn't deliver much information. Dell calibrates their displays too, as does every other brand. Displays are not stable devices and saying they were calibrated and double checked still doesn't tell you everything. They take time to warm up in general, so when you wake your computer, you have 15 to 30 minutes prior to seeing things precisely as they are intended. It's a gradual change, so most people will never notice it. Some displays seem to warm up faster than others. It doesn't tell you how many regions they measured or their pass/fail tolerance on any metric. It doesn't mention their pass or fail standards for uniformity. When people see a bad one, they assume QA, but they really don't know what is considered passable at the factory. Even after all of that, displays are unstable devices. A year from now it will look different even if you don't notice it due to the gradual nature of its changes. Some offer tools to help minimize this. With Apple you are limited to profiling software, which helps. It's just really really limited in what it can do.
None of this really tells you how well they are done. I mainly wanted to point out that these things lack some of the tools that other brands have used for years to maintain some consistency over time, but at the same time Apple doesn't really aim directly at specialized markets. It's more frequently good enough to work with than ideal when it comes to specialized use cases.
Well yeah trying to place additional links in there with asynchronous frequencies and a greatly expanded distance doesn't sound like good system design. Beyond that intel developed the technology for thunderbolt. They wouldn't develop a solution to undercut their own higher margin chips. Much of the other hardware is available in external add-on form factors with the exception of gpus and a couple other pieces of dedicated specialty hardware. GPUs seem like the most likely candidate, but I still don't think it's that likely. Mobile gpus and IGPs keep getting better. Intel needs better drivers and Apple needs to offer better OpenGL support, but in terms of weighing options, how many people are going to pay easily $600-700 for a mid range gpu shoved into a case with a $50 cable just to retrofit the machine? I remember the Quadro 4000 Mac had severe driver issues for months after its release, and that wasn't taxed with being hot pluggable.
I'm not sure about multi million dollar. It might not be that high per unit, and they likely use more than one. When people accused Dell's U2410 of having an uneven pink cast a few years ago right when it debuted, it came out that they did calibrate them. They just only measured the center. Factory calibration is just part of assembly, and they delivered just enough information to sound good. If they went into the real detail, their audience would lose interest.
I want 3 full length PCI-E 3.0 x16 slots that if I put in 3 full length AMD 7990x2 they fit and have plenty of power draw to boot. I don't want a triple where we have an x16, x8, x4 configuration. I want an x16, x16, x16 full lanes. Each lane then can have a separate Thunderbolt controller accessing and x16 PCI-E 3.0 bus.
Haven't tried Blender in a while... will give it a go!
I don't have a Mac Pro, but FCPX exploits 8 cores (4 virtual) on my loaded early 2012 iMac 27 (3.4 GHz, 16 GB RAM, Radeon HD 6970 1024MB *)
* I screwed up and didm't max out the VRAM
Anyway, FCPX is wicked fast -- rendering, transcoding and analysis (camera shake, etc.) -- you should give it a try -- it take things like MultiCam and 2K, 3K, 4K video in stride
I'm nearly finished building my Vishera 8350 box in a Corsair 650D with 32GB of RAM and > 1k Power Supply and self-contained radiator cooling for FreeBSD/Linux.
Then it's on to budgeting aside for a Mac Pro and will definitely invest in FCP X.
Dan doesn't point out the fact that the current Ivy Bridge with its Cactus Ridge Thunderbolt controllers doesn't have the bandwidth to be useful for high-end external GPUs, nor do the just-announced Redwood Ridge Thunderbolt controllers for Haswell.
In short, if they do this now, Apple will be way out on a limb for at least two years.
GPUs already perform fine over Thunderbolt:
[VIDEO]
They have their own video memory so the bandwidth isn't such a big deal. It's not an ideal setup though and not necessary. It needs driver support for a proprietary Intel tech that AMD and NVidia won't support for obvious reasons and Apple wouldn't support because they don't like you buying your own GPUs.
Let's say that Apple brought out a new Mac Pro and soldered a GTX 780 chip onto the motherboard and that connected to Thunderbolt, the GPU won't need to be upgraded for another 2-3 years. Intel plans to have a 50Gbps interconnect in 2015:
It's a novel idea for Mac Pro owners but after 2-3 years, paying Apple for a new model is an option and hey, you get a new GPU with it anyway so you still won't need to add another one. Everybody wins.
September 2013
Ivy Bridge EP Xeon
soldered GTX 780
10Gbps Redwood Ridge Thunderbolt (2-4 ports)
...
September 2015
Broadwell EP Xeon
soldered whatever (GTX 980?) GPU (up to 2x the speed of the 780)
50Gbps optical interconnect (2-4 ports)
Intel has plans to go BGA with some of Broadwell and it includes the server chips (albeit optional) so Apple could easily use CPUs soldered onto the boards too so that people can't do their own CPU upgrades and further doing Apple out of an upgrade payment.
I want 3 full length PCI-E 3.0 x16 slots that if I put in 3 full length AMD 7990x2 they fit and have plenty of power draw to boot. I don't want a triple where we have an x16, x8, x4 configuration. I want an x16, x16, x16 full lanes. Each lane then can have a separate Thunderbolt controller accessing and x16 PCI-E 3.0 bus.
It's x16, x4, x4 free. You only get 40 lanes - 16 + 16 + 4 + 4 and only the first one double-wide for the GPU. If you had 3 double wide for x16, x16, x8, you could put 3 7990s in but then you'd need to double the power supply to 2kW, which I doubt Apple would consider. They build machines based on what people are likely to do with them. Giving everyone a 2kW PSU would be overkill.
Yeah, I agree with TS, that thing just makes my stomach churn. I'm not opposed to some modular aspects to the next Mac Pro but that idea just seems so wrong to me.
Here are some related links, I dont think Florida is where the manufacturing facility will be built but it will have two research facilities for chips one for the authentec buy and one for the A series as both articles note that apple is already listing jobs for those facilities.
But the first article is about what Tim Cook said that one mac line will be built in the US and apple is dropping 100 million to make it so this year. Because the Mac Pro is a small line that amount of money for a facility to build the Mac Pro would be an appropriate amount.
You misunderstand. I was asking for a link that it was confirmed that Florida would be the site where the new manufacturing line would be/already is placed.
Oh absolutely! There was a mockup of a modular iMac redesign, and for the life of me I can't find it, but it was gorgeous.
Not that it's a good idea, but it was really well done!
I couldn't find any better modular design. I did find this mock up that is surprising similar to the one I did in Preview as a general concept of how they can make it smaller in size.
I apologize for causing stomach ailments on a Sunday evening for you guys. Was just trying to get the imagination rolling.
Hopefully, someone more artistically inclined than myself will come up with something nice and post it for all of us to see, and to cure the indigestion I have caused. I never graduated from stickfigures...
I couldn't find any better modular design. I did find this mock up that is surprising similar to the one I did in Preview as a general concept of how they can make it smaller in size.
that was posted by Marvin a while back. I just found it and was going to link.
You save the space that the Mac Pro would have taken up to begin with. There's more cabling with external storage + 2-3 peripherals but you can put them in a box if you like.
The size of the card isn't important. The fastest cards are not even 2x the performance of the iMac's GPU but cost $1000 or more each. Not to mention they are outside the safe power limits of the Mac Pro too.
The Radeon 7990 is 375W. I'm not sure how many professionals would put a 375W GPU into slots with a 300W power allocation but I doubt they'd be in business long.
The Mac Pro can obviously hold higher wattage cards than the iMac but they don't perform significantly better in practise. OS X doesn't support the highest-end GPUs on the software side either.
With the high-end GPUs out of the equation, you're just left with expansion cards, which work fine over Thunderbolt.
I've made the point before that games aren't the only benchmark, but for other applications the general consensus isn't typically the hottest gaming gpu. In some cases workstation variants are a better idea. "Expansion cards work fine over thunderbolt" is a pretty bold statement. You are more limited than you would be with PCI cards, not that it matters for everyone. You may have to reboot any time it's plugged in or if the connection is dropped. Even the proof of concept videos admit it's not really perfect. Personally I would stick to the ones where whatever company oems the device makes a usb3 or thunderbolt version rather than relying on card + enclosure. Thunderbolt isn't a plug and play PCIe unless you're dealing with certified devices.
Quote:
Originally Posted by mdriftmeyer
I want 3 full length PCI-E 3.0 x16 slots that if I put in 3 full length AMD 7990x2 they fit and have plenty of power draw to boot. I don't want a triple where we have an x16, x8, x4 configuration. I want an x16, x16, x16 full lanes. Each lane then can have a separate Thunderbolt controller accessing and x16 PCI-E 3.0 bus.
I can't think of any workstations that actually do that. Typically on the Windows variants, they encourage the use of workstation cards, which seem to carry lower tdp than some of the top gaming cards. I don't think Apple was using 200W cards back when they designed the case used for the Mac Pro today. Regardless of oem brand, you would need power connectors on them. The PCI lanes don't have enough direct power to deal with several of those. Without over-subscribing the slots, you would also be limited to the dual cpu machines for such a configuration. The lanes originate on the cpus as of Sandy Bridge, so you would need 2 at x40 each to get the 48 lanes required for your configuration aside from lanes allocated to other on board ports. I'm not sure how they would implement thunderbolt there. It might not be a huge engineering issue. I just don't know how it would be done.
Comments
Interesting quote from the Verge Pixar article quoted above:
Quote:
Now, says DeRose, open-source software like Blender can do almost everything Pixar's software can do. Last summer, Pixar even open-sourced its subdivision surface code library. "We had a competitive advantage for ten years," DeRose says, "but now we get more value by letting everyone contribute."
Pixar's biggest competitive advantage now is its ability to use this math-driven technology not to make better shapes but to tell better stories. DeRose and Pixar aren't sitting on their laurels. "Somewhere out there, a brilliant kid and their friends are working in their garage" using and improving on tools like Blender, DeRose tells the assembled children and adults at MoMath. "They will be the next Pixar."
Interesting that they admit Blender has nearly all the capabilities of Pixar's own software.
Originally Posted by v5v
I fully appreciate the benefits of Apple's "shed the old, embrace the new" approach to interface design, but in the case of a "pro" machine, I wonder if pushing everything outside the box is a good idea?
A lot of what we've traditionally relied on internal devices to accomplish can now be done with external units, but it's more hassle, more expensive, and in SOME cases, all but impossible (try finding an external solution that's equivalent to a top-of-the-line NVidia GPU/video card). External solutions are usually less capable and more expensive than their PCIe counterparts [...]
If Dan is right, PCIe functionality will be pushed outside the box, via Thunderbolt. Apart from that, it's just the optical drive. Everything else stays inside, obviously. I would hope that Apple would not do this (move PCIe outside the box) unless anything that could formerly be added via PCIe can now, or soon will be, possible via Thunderbolt.
But you've definitely hit upon one limitation that early adopters will likely face. Dan doesn't point out the fact that the current Ivy Bridge with its Cactus Ridge Thunderbolt controllers doesn't have the bandwidth to be useful for high-end external GPUs, nor do the just-announced Redwood Ridge Thunderbolt controllers for Haswell, at least according to this person:
http://www.anandtech.com/show/6885/new-thunderbolt-controllers-dsl45104410-and-future-20gbps-falcon-ridge-tb-controller-announced
So it won't be until the Falcon Ridge Thunderbolt controllers that high-end external GPUs will be a reality via Thunderbolt. So early adopters will have to accept a limitation that won't be remedied until a major refresh in 2015, when the Falcon Ridge Thunderbolt controller would most likely make its way to the new Mac Pro.
In short, if they do this now, Apple will be way out on a limb for at least two years. But Intel will be out there with them, and maybe it will scare Microsoft into dealing with this problem. This isn't Firewire. I'd love to know what Dan thinks about the whole "Thunderbolt is the new Firewire" (quoting one of the comments on the article linked above) meme that is bound to take hold in the wake of Apple doing this.
You have to wonder if Apple will include two internal GPU slots, thereby removing this limitation, since then you wouldn't need Thunderbolt to handle a second GPU/video card.
Quote:
Originally Posted by Bloodshotrollin'red
Mac Pro's are the only Mac's I buy now as they are hardware configurable. If they drop the Pro I will reluctantly drop OSX (still my favourite OS). I want some control over my machines. Apple can't have it all.
Well you can always build a Hackintosh if that happens, they actually work quite well, you can have faster CPU's, more memory, workstation level GPU cards for a fraction of the cost of a Mac Pro.
Quote:
Originally Posted by Jim W
Couldn't have said it better. An iMac or Thunderbolt connected setup cannot be taken seriously at the pro level unless your needs are very specific or you don't mind clutter. I just love the comments of people who have never worked in a professional space. Have you ever even seen a pro level graphics card? They are bigger than a Mac Mini, draw large amounts of power, and would cause an iMac to explode if it were possible to shoe horn one in, let alone cool it. That is only one subsystem. For the reasons stated above, I have no desire to get into a system like the one described above, a mess. Please, a Mac Pro tower with Pci 3 and an open selection of graphics cards. Thunderbolt 1 and eventually 2 for peripherals, Firewire for legacy, and plenty of fast RAM. And it wouldn't cost $30k...
And here's something I else: the MBP gets pretty hot at times, even with the fan kicking in. That's the nature of any decently powerful laptop that is your centerpiece. It's already atop a vented stand with its extra fan but that's just venting and blowing on the outside of the case, which helps but doesn't address the circumstance, which is powerful components jammed close together with minimal ventilation and a teeny, cursory fan. Apple has done amazing things in their quest for making things more and more powerful in thinner and thinner cases but at some cost. I've always felt that, during these past few years of lack of modern desktop choices, if they made a MBP that was an inch thick I'd be the first in line. I mean, if all we're being offered is laptops why not make one with a more powerful hotter than low power chip and motherboard, make it thick enough to vent and fan well and there's no reason it would have to be much heavier.
Well, then there's that clutter thing, yeah... : )
Quote:
Originally Posted by TenThousandThings
If Dan is right, PCIe functionality will be pushed outside the box, via Thunderbolt. Apart from that, it's just the optical drive. Everything else stays inside, obviously. I would hope that Apple would not do this (move PCIe outside the box) unless anything that could formerly be added via PCIe can now, or soon will be, possible via Thunderbolt.
But you've definitely hit upon one limitation that early adopters will likely face. Dan doesn't point out the fact that the current Ivy Bridge with its Cactus Ridge Thunderbolt controllers doesn't have the bandwidth to be useful for high-end external GPUs, nor do the just-announced Redwood Ridge Thunderbolt controllers for Haswell, at least according to this person:
http://www.anandtech.com/show/6885/new-thunderbolt-controllers-dsl45104410-and-future-20gbps-falcon-ridge-tb-controller-announced
So it won't be until the Falcon Ridge Thunderbolt controllers that high-end external GPUs will be a reality via Thunderbolt. So early adopters will have to accept a limitation that won't be remedied until a major refresh in 2015, when the Falcon Ridge Thunderbolt controller would most likely make its way to the new Mac Pro.
In short, if they do this now, Apple will be way out on a limb for at least two years. But Intel will be out there with them, and maybe it will scare Microsoft into dealing with this problem. This isn't Firewire. I'd love to know what Dan thinks about the whole "Thunderbolt is the new Firewire" (quoting one of the comments on the article linked above) meme that is bound to take hold in the wake of Apple doing this.
You have to wonder if Apple will include two internal GPU slots, thereby removing this limitation, since then you wouldn't need Thunderbolt to handle a second GPU/video card.
Nor the fact that an external GPGPU System only adds to the complexity of the entire system which will require a separate cooling solution and power solution thus driving the price up by about $1k.
Quote:
Originally Posted by hmm
Saying they calibrate is partly marketing speech. I'm sure they have a good system in place, but the statement doesn't deliver much information. Dell calibrates their displays too, as does every other brand. Displays are not stable devices and saying they were calibrated and double checked still doesn't tell you everything. They take time to warm up in general, so when you wake your computer, you have 15 to 30 minutes prior to seeing things precisely as they are intended. It's a gradual change, so most people will never notice it. Some displays seem to warm up faster than others. It doesn't tell you how many regions they measured or their pass/fail tolerance on any metric. It doesn't mention their pass or fail standards for uniformity. When people see a bad one, they assume QA, but they really don't know what is considered passable at the factory. Even after all of that, displays are unstable devices. A year from now it will look different even if you don't notice it due to the gradual nature of its changes. Some offer tools to help minimize this. With Apple you are limited to profiling software, which helps. It's just really really limited in what it can do.
None of this really tells you how well they are done. I mainly wanted to point out that these things lack some of the tools that other brands have used for years to maintain some consistency over time, but at the same time Apple doesn't really aim directly at specialized markets. It's more frequently good enough to work with than ideal when it comes to specialized use cases.
Well yeah trying to place additional links in there with asynchronous frequencies and a greatly expanded distance doesn't sound like good system design. Beyond that intel developed the technology for thunderbolt. They wouldn't develop a solution to undercut their own higher margin chips. Much of the other hardware is available in external add-on form factors with the exception of gpus and a couple other pieces of dedicated specialty hardware. GPUs seem like the most likely candidate, but I still don't think it's that likely. Mobile gpus and IGPs keep getting better. Intel needs better drivers and Apple needs to offer better OpenGL support, but in terms of weighing options, how many people are going to pay easily $600-700 for a mid range gpu shoved into a case with a $50 cable just to retrofit the machine? I remember the Quadro 4000 Mac had severe driver issues for months after its release, and that wasn't taxed with being hot pluggable.
I'm not sure about multi million dollar. It might not be that high per unit, and they likely use more than one. When people accused Dell's U2410 of having an uneven pink cast a few years ago right when it debuted, it came out that they did calibrate them. They just only measured the center. Factory calibration is just part of assembly, and they delivered just enough information to sound good. If they went into the real detail, their audience would lose interest.
I want 3 full length PCI-E 3.0 x16 slots that if I put in 3 full length AMD 7990x2 they fit and have plenty of power draw to boot. I don't want a triple where we have an x16, x8, x4 configuration. I want an x16, x16, x16 full lanes. Each lane then can have a separate Thunderbolt controller accessing and x16 PCI-E 3.0 bus.
Quote:
Originally Posted by Dick Applebaum
Haven't tried Blender in a while... will give it a go!
I don't have a Mac Pro, but FCPX exploits 8 cores (4 virtual) on my loaded early 2012 iMac 27 (3.4 GHz, 16 GB RAM, Radeon HD 6970 1024MB *)
* I screwed up and didm't max out the VRAM
Anyway, FCPX is wicked fast -- rendering, transcoding and analysis (camera shake, etc.) -- you should give it a try -- it take things like MultiCam and 2K, 3K, 4K video in stride
I'm nearly finished building my Vishera 8350 box in a Corsair 650D with 32GB of RAM and > 1k Power Supply and self-contained radiator cooling for FreeBSD/Linux.
Then it's on to budgeting aside for a Mac Pro and will definitely invest in FCP X.
GPUs already perform fine over Thunderbolt:
[VIDEO]
They have their own video memory so the bandwidth isn't such a big deal. It's not an ideal setup though and not necessary. It needs driver support for a proprietary Intel tech that AMD and NVidia won't support for obvious reasons and Apple wouldn't support because they don't like you buying your own GPUs.
Let's say that Apple brought out a new Mac Pro and soldered a GTX 780 chip onto the motherboard and that connected to Thunderbolt, the GPU won't need to be upgraded for another 2-3 years. Intel plans to have a 50Gbps interconnect in 2015:
http://www.tomshardware.com/news/Thunderbolt-silicon-photonics-optical-networking-50-gigabits-per-second-Jef-Demain,12668.html
It's a novel idea for Mac Pro owners but after 2-3 years, paying Apple for a new model is an option and hey, you get a new GPU with it anyway so you still won't need to add another one. Everybody wins.
September 2013
Ivy Bridge EP Xeon
soldered GTX 780
10Gbps Redwood Ridge Thunderbolt (2-4 ports)
...
September 2015
Broadwell EP Xeon
soldered whatever (GTX 980?) GPU (up to 2x the speed of the 780)
50Gbps optical interconnect (2-4 ports)
Intel has plans to go BGA with some of Broadwell and it includes the server chips (albeit optional) so Apple could easily use CPUs soldered onto the boards too so that people can't do their own CPU upgrades and further doing Apple out of an upgrade payment.
It's x16, x4, x4 free. You only get 40 lanes - 16 + 16 + 4 + 4 and only the first one double-wide for the GPU. If you had 3 double wide for x16, x16, x8, you could put 3 7990s in but then you'd need to double the power supply to 2kW, which I doubt Apple would consider. They build machines based on what people are likely to do with them. Giving everyone a 2kW PSU would be overkill.
As some here have suggested a modular design, there is a concept drawing out there for one:
http://ipack3d.com/page8/index.php?id=2959698825183484676
There was a long thread back a while that discussed the MacPro and suggested the modular form might not work, but the concept is cool.
http://forums.appleinsider.com/t/150659/tim-cook-confirms-updated-mac-pro-coming-in-2013
Originally Posted by Bergermeister
As some here have suggested a modular design, there is a concept drawing out there for one:
http://ipack3d.com/page8/index.php?id=2959698825183484676
I hate that thing so much. It's so opposite of everything Apple does.
Quote:
Originally Posted by Tallest Skil
I hate that thing so much. It's so opposite of everything Apple does.
I'd love to see other mockups of what people are suggesting here, but this is the only one I have seen.
Originally Posted by Bergermeister
I'd love to see other mockups of what people are suggesting here, but this is the only one I have seen.
Oh absolutely! There was a mockup of a modular iMac redesign, and for the life of me I can't find it, but it was gorgeous.
Not that it's a good idea, but it was really well done!
Yeah, I agree with TS, that thing just makes my stomach churn. I'm not opposed to some modular aspects to the next Mac Pro but that idea just seems so wrong to me.
Quote:
Originally Posted by Mechanic
http://appleinsider.com/articles/12/12/06/apple-will-invest-100m-to-bring-some-production-of-macs-to-us-in-2013
http://www.electronista.com/articles/13/04/16/florida.set.to.become.important.apple.hub/
http://appleinsider.com/articles/13/04/08/apples-us-based-chip-development-expanding-in-florida-could-be-related-to-fingerprint-tech
Here are some related links, I dont think Florida is where the manufacturing facility will be built but it will have two research facilities for chips one for the authentec buy and one for the A series as both articles note that apple is already listing jobs for those facilities.
But the first article is about what Tim Cook said that one mac line will be built in the US and apple is dropping 100 million to make it so this year. Because the Mac Pro is a small line that amount of money for a facility to build the Mac Pro would be an appropriate amount.
You misunderstand. I was asking for a link that it was confirmed that Florida would be the site where the new manufacturing line would be/already is placed.
I couldn't find any better modular design. I did find this mock up that is surprising similar to the one I did in Preview as a general concept of how they can make it smaller in size.
I apologize for causing stomach ailments on a Sunday evening for you guys. Was just trying to get the imagination rolling.
Hopefully, someone more artistically inclined than myself will come up with something nice and post it for all of us to see, and to cure the indigestion I have caused. I never graduated from stickfigures...
Quote:
Originally Posted by SolipsismX
I couldn't find any better modular design. I did find this mock up that is surprising similar to the one I did in Preview as a general concept of how they can make it smaller in size.
that was posted by Marvin a while back. I just found it and was going to link.
post #131
Just noticed that this thread is filed under Software > MacOS
But we are discussing hardware...
Quote:
Originally Posted by Marvin
You save the space that the Mac Pro would have taken up to begin with. There's more cabling with external storage + 2-3 peripherals but you can put them in a box if you like.
The size of the card isn't important. The fastest cards are not even 2x the performance of the iMac's GPU but cost $1000 or more each. Not to mention they are outside the safe power limits of the Mac Pro too.
The Radeon 7990 is 375W. I'm not sure how many professionals would put a 375W GPU into slots with a 300W power allocation but I doubt they'd be in business long.
The Mac Pro can obviously hold higher wattage cards than the iMac but they don't perform significantly better in practise. OS X doesn't support the highest-end GPUs on the software side either.
With the high-end GPUs out of the equation, you're just left with expansion cards, which work fine over Thunderbolt.
I've made the point before that games aren't the only benchmark, but for other applications the general consensus isn't typically the hottest gaming gpu. In some cases workstation variants are a better idea. "Expansion cards work fine over thunderbolt" is a pretty bold statement. You are more limited than you would be with PCI cards, not that it matters for everyone. You may have to reboot any time it's plugged in or if the connection is dropped. Even the proof of concept videos admit it's not really perfect. Personally I would stick to the ones where whatever company oems the device makes a usb3 or thunderbolt version rather than relying on card + enclosure. Thunderbolt isn't a plug and play PCIe unless you're dealing with certified devices.
Quote:
Originally Posted by mdriftmeyer
I want 3 full length PCI-E 3.0 x16 slots that if I put in 3 full length AMD 7990x2 they fit and have plenty of power draw to boot. I don't want a triple where we have an x16, x8, x4 configuration. I want an x16, x16, x16 full lanes. Each lane then can have a separate Thunderbolt controller accessing and x16 PCI-E 3.0 bus.
I can't think of any workstations that actually do that. Typically on the Windows variants, they encourage the use of workstation cards, which seem to carry lower tdp than some of the top gaming cards. I don't think Apple was using 200W cards back when they designed the case used for the Mac Pro today. Regardless of oem brand, you would need power connectors on them. The PCI lanes don't have enough direct power to deal with several of those. Without over-subscribing the slots, you would also be limited to the dual cpu machines for such a configuration. The lanes originate on the cpus as of Sandy Bridge, so you would need 2 at x40 each to get the 48 lanes required for your configuration aside from lanes allocated to other on board ports. I'm not sure how they would implement thunderbolt there. It might not be a huge engineering issue. I just don't know how it would be done.