Or when Intel, ATI or Nvidia announce upgrades to components that could be in the nnMP.
Although, that's little more than mental masturbation - since Apple ignore Haswell-EP and DDR4 completely.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Is the new Mac Pro a failure?
- Thread starter ZombiePhysicist
- Start date
- Sort by reaction score
- Status
Although, that's little more than mental masturbation - since Apple ignore Haswell-EP and DDR4 completely.
I'll love having support on OSX for Intel's Xeon Phi - Knigts Ferry a couple of these on a cage linked thu thunderbolt 3, would kick some asses (specially if they add some suport on Swift/obj-C) Xeon Phi programmin is not as restricted as GPU programming.
The biggest obstacle for our idea isn't the GPU interface (Apple could release it already and nVidia deliver an nMP compatible Quadro, is not difficult to replace an GPU if you have the part and thermal paste.), the problem is Apple deals with AMD.
Look up "System X" at Virginia Tech. (https://www.eng.vt.edu/news/virgini...d-7th-fastest-supercomputer-world-top500-list)
Unfortunately, it used G5s without ECC, and was unable to be actively used in practice because of the instability due to unnoticed memory errors. It was rebuilt using newer generation G5s with ECC - but by then PowerPC had fallen off the cliff as far as competitive performance.
Big PR stunt for Apple, big failure for Virginia Tech.
Supercomputers use Nvidia GPUs and Intel Xeon Phi accelerators - or phenomenal numbers of x64 cores.
The biggest liability for a Mac super computer, however, is Apple OSX.
Apple creating a "real" computer again would be ideal. I'm not convinced that they ever will.
Fortunately for them, Intel has "fixed" ThunderBolt to be less of a crazy burden. USB-C can be what we always REALLY wanted from TB, it can just carry Data, no longer does it need to have a video feed. So they COULD make a Mac Pro with Xeon that could have the Data abilities of USB-C with no need to have the GPU welded inside. They could even dust off the old cMP plans, slap some new parts in and get back to serious computing.
But as I said, doubtful.
But I would like to posit a question.
What is a GPU that has been modified to be on a smaller card, with no heatsink, and no display outputs? IYou also need to include lower clocks and a new name to jazz up the cheering masses? Well, typically this would be a mobile GPU on an MXM card. Which have more in common with nMP Dx00s than do real GPU for PCIE slots.
So...and this is a big "What if?'...What if Apple or a 3rd party offered an adapter kit to allow standard MXM cards to be mounted in nMP? Obviously this would be tricky, but I'm pretty sure it could be done. Nvidia makes a fully specced GTX980 in MXM form. There are also Quadros of various sorts like this. Of course, since the Apple system EFI curtrently holds the EFI and BIOS for the GPUs, someone would have to write some new EFI BIOS's to work on nMP, but I may know someone who could do that.
Or...what if Apple releases the 7,1 with boards like this built in? So anyone could swap in MXM cards as they are released for other machines? The 7,1 could become the X-Mac with newer GPUs and an i7. This would clear the path for the above mentioned "Real" Mac Pro.
Apple could still pull out of the nosedive that is the 6,1. Hopefully Tim or Jony can find this thread and realize that having the Gibberati endlessly repeat nonsense DOESN'T MAKE IT TRUE.
And before all of the people happy with 6,1 come on the flame attack, I DON"T HATE THE 6,1. It makes a fine 3/4s machine. But there are still lots of people who need a 4/4s machine who have been tossed by the wayside. So a 7,1 X-Mac with MXM GPUs that can be easily upgraded with mobile MXM GPUs, and a "real" Mac Pro in a slightly smaller case similar to cheese grater. Lose the optical bays, perhaps have space for 3.5" drives there. Every machine would have dual CPUs so PCIE lanes aplenty don't have to plan for the "cheapie" model with just one. Add a PCIE slot or two, 10GBE, SATA 3, the whole nine yards. Perhaps MULTIPLE M2 slots?
Again, I don't HATE the 6,1, I MISS what it could have been.
Fortunately for them, Intel has "fixed" ThunderBolt to be less of a crazy burden. USB-C can be what we always REALLY wanted from TB, it can just carry Data, no longer does it need to have a video feed. So they COULD make a Mac Pro with Xeon that could have the Data abilities of USB-C with no need to have the GPU welded inside. They could even dust off the old cMP plans, slap some new parts in and get back to serious computing.
But as I said, doubtful.
But I would like to posit a question.
What is a GPU that has been modified to be on a smaller card, with no heatsink, and no display outputs? IYou also need to include lower clocks and a new name to jazz up the cheering masses? Well, typically this would be a mobile GPU on an MXM card. Which have more in common with nMP Dx00s than do real GPU for PCIE slots.
So...and this is a big "What if?'...What if Apple or a 3rd party offered an adapter kit to allow standard MXM cards to be mounted in nMP? Obviously this would be tricky, but I'm pretty sure it could be done. Nvidia makes a fully specced GTX980 in MXM form. There are also Quadros of various sorts like this. Of course, since the Apple system EFI curtrently holds the EFI and BIOS for the GPUs, someone would have to write some new EFI BIOS's to work on nMP, but I may know someone who could do that.
Or...what if Apple releases the 7,1 with boards like this built in? So anyone could swap in MXM cards as they are released for other machines? The 7,1 could become the X-Mac with newer GPUs and an i7. This would clear the path for the above mentioned "Real" Mac Pro.
Apple could still pull out of the nosedive that is the 6,1. Hopefully Tim or Jony can find this thread and realize that having the Gibberati endlessly repeat nonsense DOESN'T MAKE IT TRUE.
And before all of the people happy with 6,1 come on the flame attack, I DON"T HATE THE 6,1. It makes a fine 3/4s machine. But there are still lots of people who need a 4/4s machine who have been tossed by the wayside. So a 7,1 X-Mac with MXM GPUs that can be easily upgraded with mobile MXM GPUs, and a "real" Mac Pro in a slightly smaller case similar to cheese grater. Lose the optical bays, perhaps have space for 3.5" drives there. Every machine would have dual CPUs so PCIE lanes aplenty don't have to plan for the "cheapie" model with just one. Add a PCIE slot or two, 10GBE, SATA 3, the whole nine yards. Perhaps MULTIPLE M2 slots?
Again, I don't HATE the 6,1, I MISS what it could have been.
Apple creating a "real" computer again would be ideal. I'm not convinced that they ever will.
Fortunately for them, Intel has "fixed" ThunderBolt to be less of a crazy burden. USB-C can be what we always REALLY wanted from TB, it can just carry Data, no longer does it need to have a video feed. So they COULD make a Mac Pro with Xeon that could have the Data abilities of USB-C with no need to have the GPU welded inside. They could even dust off the old cMP plans, slap some new parts in and get back to serious computing.
But as I said, doubtful.
But I would like to posit a question.
What is a GPU that has been modified to be on a smaller card, with no heatsink, and no display outputs? IYou also need to include lower clocks and a new name to jazz up the cheering masses? Well, typically this would be a mobile GPU on an MXM card. Which have more in common with nMP Dx00s than do real GPU for PCIE slots.
So...and this is a big "What if?'...What if Apple or a 3rd party offered an adapter kit to allow standard MXM cards to be mounted in nMP? Obviously this would be tricky, but I'm pretty sure it could be done. Nvidia makes a fully specced GTX980 in MXM form. There are also Quadros of various sorts like this. Of course, since the Apple system EFI curtrently holds the EFI and BIOS for the GPUs, someone would have to write some new EFI BIOS's to work on nMP, but I may know someone who could do that.
Or...what if Apple releases the 7,1 with boards like this built in? So anyone could swap in MXM cards as they are released for other machines? The 7,1 could become the X-Mac with newer GPUs and an i7. This would clear the path for the above mentioned "Real" Mac Pro.
Apple could still pull out of the nosedive that is the 6,1. Hopefully Tim or Jony can find this thread and realize that having the Gibberati endlessly repeat nonsense DOESN'T MAKE IT TRUE.
And before all of the people happy with 6,1 come on the flame attack, I DON"T HATE THE 6,1. It makes a fine 3/4s machine. But there are still lots of people who need a 4/4s machine who have been tossed by the wayside. So a 7,1 X-Mac with MXM GPUs that can be easily upgraded with mobile MXM GPUs, and a "real" Mac Pro in a slightly smaller case similar to cheese grater. Lose the optical bays, perhaps have space for 3.5" drives there. Every machine would have dual CPUs so PCIE lanes aplenty don't have to plan for the "cheapie" model with just one. Add a PCIE slot or two, 10GBE, SATA 3, the whole nine yards. Perhaps MULTIPLE M2 slots?
Again, I don't HATE the 6,1, I MISS what it could have been.
It's cool to think about but I think the reality is without you and netkas the MP tower would have only the GPUs that Apple has made available to it - just because other cards work in those towers doesn't mean that was what Apple intended, so to me it's hard to see them add modular style GPUs to any of their Macs.
If it's a 3rd party I guess the restriction is power, heat and SATA, but at that point why not just have each board supply an NVMe interface.
Edit: I can see if another one comes out with the same form factor it may make sense for someone to offer that type of solution if it would work in both versions... that would be cool.
Last edited:
I've posted these before, but:
Would have been very nice. Bold and daring design. Fully modular workstation. Nothing like it in the PC world.
Apple makes profits by selling expansion modules as users' needs grow.
Almost wonder if the delay is because they realized the nMP design wasn't a good thing and they had to go back to the drawing board for 7,1. If only...
I'd find it somewhat acceptable if they enlarged the trashcan a bit but added a double width PCI slot in the center to hold a standard PCI video card.
If Apple management could get it through their heads that most professionals don't care about a machine being small/light/thin then they could make a lot of good compromises with 7,1 and beyond.
I'm pessimistic though... I expect we'll see new Xeons, Thunderbolt3+USB3-C, DDR4 and the current generation of "Firepro" , but all in the same form factor with its limitations. At best we can hope they'll include official support for eGPU expansion chassis since that is part of the Intel TB3 spec.
Would have been very nice. Bold and daring design. Fully modular workstation. Nothing like it in the PC world.
Apple makes profits by selling expansion modules as users' needs grow.
Almost wonder if the delay is because they realized the nMP design wasn't a good thing and they had to go back to the drawing board for 7,1. If only...
I'd find it somewhat acceptable if they enlarged the trashcan a bit but added a double width PCI slot in the center to hold a standard PCI video card.
If Apple management could get it through their heads that most professionals don't care about a machine being small/light/thin then they could make a lot of good compromises with 7,1 and beyond.
I'm pessimistic though... I expect we'll see new Xeons, Thunderbolt3+USB3-C, DDR4 and the current generation of "Firepro" , but all in the same form factor with its limitations. At best we can hope they'll include official support for eGPU expansion chassis since that is part of the Intel TB3 spec.
Wishful thinking, but something tells me the Apple of today will do everything in it's power to engineer a limitation in their implementation which will restrict people from using eGPU's. It appears obvious to me in the post 2012 era that Apple doesn't seem interested in extending the life (or customization past their spec) of 'devices' in favor of complete system replacements at the end of a Apple Care cycle. You could argue if that weren't the case, the precious tower would still be with us in some form today.
At best, we can hope they don't solder the ram into the logic board in the next revision.
--EDIT--
Also, speculating, I think Apple would have some serious issues with allowing eGPU as their official sanctioning of video cards even when the tower rained supreme was pretty light. It's not like Nvidia or ATI/AMD was releasing every flavor of GPU for mac when they had the ability to easily do so.
Last edited:
TB3 Allows for eGPUs - Apple will release one, but it will be a monitor (r5k iMac without CPU), not a dedicated box. Said monitor will let the lowest end TB3 equipped mac drive a 5k display, and have passthrough to allow a computer with dedicated, more powerful GPU to drive it instead.
So sayeth the puddle of water I saw on the footpath.
Whether TB3 eGPU functionality will be whitelisted to only Apple's display, if it'll be open season for anyone who can put a standard pci card in a chassis, or somewhere inbetween - Apple-Display-only for 6 months before OS support is added for non-Apple solutions, or only allowing an MFI-Certified style accessory market, will be the big question.
So sayeth the puddle of water I saw on the footpath.
Whether TB3 eGPU functionality will be whitelisted to only Apple's display, if it'll be open season for anyone who can put a standard pci card in a chassis, or somewhere inbetween - Apple-Display-only for 6 months before OS support is added for non-Apple solutions, or only allowing an MFI-Certified style accessory market, will be the big question.
You would be completely delusional, if you think any product design of any computer manufacturer isn't 100% about maximizing profit.
Cynically, quite true. From a marketing standpoint, Apple has claimed that their goal is to "Delight" the customer, so there's a fundamental conflict between these two objectives. Overall, this probably has the old "pendulum swing" to its balance (like everything else), so the question may very well be where is it and is it continuing to swing too far the one way?
Historically, I'll note that back in the PPC days of the 7500/8500/9500 series that the CPUs were mounted on daughterboards that were a very easy customer DIY ... and which created a quite strong 3rd party market for the same - - this clearly hindered Apple desktop sales during this period because one could easily/cheaply do a pretty nice upgrade by simply unplugging the OEM 120MHz G3 daugtherboard and pop in a 3rd Party 233MHz G4 /etc. Don't think that Apple didn't learn to never leave their hardware design as "open" as that again.
FWIW, I do think that we are still on the "profit" side of the pendulum, because the fingerprints that I'm seeing appear to be all about minimalizing the expense of OS X by streamlining: eliminating features (such as RAID), software applications (such as Aperture), etc. "Doing less with less" reduce its lifecycle costs (Development, QA, Support/Maintenance, etc). Recall also how this sort of "supply chain" thing is reportedly Tim Cook's strength.
Also note how this manifests itself in Apple's "no choice" aspects of GPUs: to deliberately offer fewer GPU permutations means less money spent in OS X/Software development (making it work) and QA (verifying that it works) --> less overhead expenses ultimately returns better profits.
The flaw with this approach is that it is a dangerous dead end. Simply "Choose Your Poison":
* Narrowing down to just one solution assumes that innovation in the technology field has effectively stopped.
The paradigm here is that there's no justification to diversify to explore and/or keep options open for future improvement, because no further improvements are to be forthcoming. Thus, one can simply choose whichever commodity solution has the best terms (eg, lowest cost).
-or-
* Narrowing down to just the one solution picked in-house is the result of an "NIH" (Not Invented Here) echo chamber.
As it says on the NIH page on Wiki:
"False pride often drives an enterprise to use less-than-perfect invention in order to save face by ignoring, boycotting, or otherwise refusing to use or incorporate obviously superior solutions by others."
Recall how Intel's success has been attributed by CEO Andrew Grove in his book: "Only the Paranoid Survive". The question is if Apple is paranoid enough to make the risk of again offering any Insanely Great products. If the answer is no, then they will not survive.
-hh
Apple creating a "real" computer again would be ideal. I'm not convinced that they ever will.
Fortunately for them, Intel has "fixed" ThunderBolt to be less of a crazy burden. USB-C can be what we always REALLY wanted from TB, it can just carry Data, no longer does it need to have a video feed. So they COULD make a Mac Pro with Xeon that could have the Data abilities of USB-C with no need to have the GPU welded inside. They could even dust off the old cMP plans, slap some new parts in and get back to serious computing.
But as I said, doubtful.
But I would like to posit a question.
What is a GPU that has been modified to be on a smaller card, with no heatsink, and no display outputs? IYou also need to include lower clocks and a new name to jazz up the cheering masses? Well, typically this would be a mobile GPU on an MXM card. Which have more in common with nMP Dx00s than do real GPU for PCIE slots.
So...and this is a big "What if?'...What if Apple or a 3rd party offered an adapter kit to allow standard MXM cards to be mounted in nMP? Obviously this would be tricky, but I'm pretty sure it could be done. Nvidia makes a fully specced GTX980 in MXM form. There are also Quadros of various sorts like this. Of course, since the Apple system EFI curtrently holds the EFI and BIOS for the GPUs, someone would have to write some new EFI BIOS's to work on nMP, but I may know someone who could do that.
Or...what if Apple releases the 7,1 with boards like this built in? So anyone could swap in MXM cards as they are released for other machines? The 7,1 could become the X-Mac with newer GPUs and an i7. This would clear the path for the above mentioned "Real" Mac Pro.
Apple could still pull out of the nosedive that is the 6,1. Hopefully Tim or Jony can find this thread and realize that having the Gibberati endlessly repeat nonsense DOESN'T MAKE IT TRUE.
And before all of the people happy with 6,1 come on the flame attack, I DON"T HATE THE 6,1. It makes a fine 3/4s machine. But there are still lots of people who need a 4/4s machine who have been tossed by the wayside. So a 7,1 X-Mac with MXM GPUs that can be easily upgraded with mobile MXM GPUs, and a "real" Mac Pro in a slightly smaller case similar to cheese grater. Lose the optical bays, perhaps have space for 3.5" drives there. Every machine would have dual CPUs so PCIE lanes aplenty don't have to plan for the "cheapie" model with just one. Add a PCIE slot or two, 10GBE, SATA 3, the whole nine yards. Perhaps MULTIPLE M2 slots?
Again, I don't HATE the 6,1, I MISS what it could have been.
The 6,1 is exactly what Apple wanted it to be, a closed system. Apple makes money selling the new version of their product, they need the old one to be obsolete.
At this point we'll be lucky if TB3 is included on the 7,1 tho I think the Skylake Xeon's support 48 PCIE lanes rather than the 40 the MacPro 6,1 has, so there should be no reason that TB3 would not be included, other than spite.
Indeed, closing the ecosystem must be Apple's concern. I imagine not just for profit, but for control.
R&D, QA and support costs driven down.
I can even understand that position, although it's limiting for us users of course.
Vader, don't get your hopes up with Skylake, those 48 lanes are only 2S systems and above. Currently nMP is 1S only and probably will stay that way, which is to say that you should get the same lanes when/if it comes.
Still, you can have 2 controllers (4 ports) with the current setup. It's 2 less ports than the current version but with additional bandwidth.
The problem here is that if you use 2 displays then you're out of ports to use additionally. :-(
The HDMI (now 2.0) port, if maintained, can overcome this though.
R&D, QA and support costs driven down.
I can even understand that position, although it's limiting for us users of course.
Vader, don't get your hopes up with Skylake, those 48 lanes are only 2S systems and above. Currently nMP is 1S only and probably will stay that way, which is to say that you should get the same lanes when/if it comes.
Still, you can have 2 controllers (4 ports) with the current setup. It's 2 less ports than the current version but with additional bandwidth.
The problem here is that if you use 2 displays then you're out of ports to use additionally. :-(
The HDMI (now 2.0) port, if maintained, can overcome this though.
Skylake in the E5 Xeon version will bring 48 lanes of PCI-E per CPU. Whilst the nMP is a single CPU system the current E5 1600 in the nMP is limited to 6 Cores so the 8, 10 and 12 are using E5 2600 series still. As such should get 48 lanes when a Skylake launches however that is 1H 17, more likely to see a broadwell-ep nMP before then.
Broadwell-ep however will still only be 40 lanes PCI-E, which unless Apple seriously believe that they can hold out till 1H 2017 then most likely will be a broadwell-ep launch for the nMP v2
Do you really expect Apple to design 2 variants of the nMP when Skylake -EP comes out? Really?
Right now, the nMP uses indeed 2600 from 8 core up, but there are 1600 with higher core count than 6, I imagine you know that. With Broadwell-EP coming out, 1600 should up the core count and maybe again with Skylake-EP.
Something will have to change by then or even EOL for nMP in it's current form.
I don't see Apple having one platform for up to 10 cores or so (max core count for 1600 by then) and another for the 2600 series for more cores. The former will be based on Kaby Lake and the latter on Purley? Don't think so.
Maybe then they revive the cMP, which is more Purley like than the nMP.
Or even better, the nnMP will be entirely based on Purley, starting on 10-12 cores, but the cost, which is already high enough, will skyrocket.
That's my humble vision of possible configs.
Right now, the nMP uses indeed 2600 from 8 core up, but there are 1600 with higher core count than 6, I imagine you know that. With Broadwell-EP coming out, 1600 should up the core count and maybe again with Skylake-EP.
Something will have to change by then or even EOL for nMP in it's current form.
I don't see Apple having one platform for up to 10 cores or so (max core count for 1600 by then) and another for the 2600 series for more cores. The former will be based on Kaby Lake and the latter on Purley? Don't think so.
Maybe then they revive the cMP, which is more Purley like than the nMP.
Or even better, the nnMP will be entirely based on Purley, starting on 10-12 cores, but the cost, which is already high enough, will skyrocket.
That's my humble vision of possible configs.
Why do they need two versions? Nothing in my post says will need two versions! E5 1600 and E5 2600 use the same socket! They don't have two versions to use the E5 2600's and E5 1600's now. Why would they need it later on?
That's a step along the direction that I propose - but it's cutting the MP5.1 into pieces and building new MP5.1 systems from the pieces.I've posted these before, but:
Would have been very nice. Bold and daring design. Fully modular workstation. Nothing like it in the PC world.
Apple makes profits by selling expansion modules as users' needs grow.
Almost wonder if the delay is because they realized the nMP design wasn't a good thing and they had to go back to the drawing board for 7,1. If only...
I'd find it somewhat acceptable if they enlarged the trashcan a bit but added a double width PCI slot in the center to hold a standard PCI video card.
If Apple management could get it through their heads that most professionals don't care about a machine being small/light/thin then they could make a lot of good compromises with 7,1 and beyond.
I'm pessimistic though... I expect we'll see new Xeons, Thunderbolt3+USB3-C, DDR4 and the current generation of "Firepro" , but all in the same form factor with its limitations. At best we can hope they'll include official support for eGPU expansion chassis since that is part of the Intel TB3 spec.
Let's cut the 6,1 into pieces - and build from that.
I didn't realize the Skylake Xeon's with 48 lane PCIE weren't shipping until 2017. This leaves TB3 as very doubtful in my mind for the MacPro 7,1.
As you say, the only CPU's Apple is left with are the 1600 and the 2600 Broadwell-ep with 40 lanes and that still uses the C610 chipset with 8 PCIE 2.0 Lanes, so in order to get Thunderbolt 3 they need to come up with 8 more PCIE lanes.
1. Intel could build Apple a custom chipset with 16 PCIE lanes
2. Apple could cut the link speed of the second GFX card to 8x
3. Apple could use 2 1600 CPU's and only offer TB3 on the high end MP7,1
1 is the best case, 2 is the wisest move, the second GFX card doesn't do anything most of the time and we're still talking about and 8x link at PCIE 3.0 speeds. 3, I'm not even sure if the 1600 supports SMP or Apple could even fit 2 of them in the nMP case even if they did.
None of the above are very likely, what is likely is we'll see a refresh of the MacPro at WWDC 2016 in June /without TB3.
Apple could use a second CPU instead of using a second GPU which pretty much no one uses anyway and drive every component with the maximum PCIe Lanes. But it's the 2016 Prosumer Apple, so won't ever happen.
mcnallym, have you seen the roadmap where Intel details the future platforms?
Purley for 2S and up, Basin Falls for 1S.
Realize that both these use different PCHs (Lewisburg and Kaby Lake) and probably different sockets as well. Remember that 2S+ will come with 6 channel mem, so this automatically should mean new socket, as well as OP fabric. 1600 should become closer to the desktop platform, at least when it comes to the PCH, and similar to what it is now.
2S and up are different beasts altogether.
Vader, 2017 is possibly a good bet but with all these delays who knows. The process is quite common then but Intel might decide to postpone release. I wouldn't count on it before middle to year's end though. Going in the nMP will be even later.
TB3 does not imply the use of Skylake. The extra 8 PCIe3 lanes in Broadwell-EP will be good for 2 TB3 controllers, each with 2 ports, each double the TB2 bandwidth. You get a reduced number of ports but increase in bandwidth.
Intel could have by now upgrade the C610 into the C620 (or whatever) so as to at least make the PCIe2 lanes into PCIe3, but that is a major overhaul and will come only with Skylake-EP, and more. USB 3.1 seems to be AOL though.
For dual CPU setup they would have to go with 2600, which would further increase the price, that we all already find (and agree) high enough
Purley for 2S and up, Basin Falls for 1S.
Realize that both these use different PCHs (Lewisburg and Kaby Lake) and probably different sockets as well. Remember that 2S+ will come with 6 channel mem, so this automatically should mean new socket, as well as OP fabric. 1600 should become closer to the desktop platform, at least when it comes to the PCH, and similar to what it is now.
2S and up are different beasts altogether.
Vader, 2017 is possibly a good bet but with all these delays who knows. The process is quite common then but Intel might decide to postpone release. I wouldn't count on it before middle to year's end though. Going in the nMP will be even later.
TB3 does not imply the use of Skylake. The extra 8 PCIe3 lanes in Broadwell-EP will be good for 2 TB3 controllers, each with 2 ports, each double the TB2 bandwidth. You get a reduced number of ports but increase in bandwidth.
Intel could have by now upgrade the C610 into the C620 (or whatever) so as to at least make the PCIe2 lanes into PCIe3, but that is a major overhaul and will come only with Skylake-EP, and more. USB 3.1 seems to be AOL though.
For dual CPU setup they would have to go with 2600, which would further increase the price, that we all already find (and agree) high enough
4. Apple could add a second PCIe switch
And 1600s don't do SMP - that's what 2600s do.
You know that right now they use the 2600 CPUs?For dual CPU setup they would have to go with 2600, which would further increase the price, that we all already find (and agree) high enough
I didn't realize the Skylake Xeon's with 48 lane PCIE weren't shipping until 2017. This leaves TB3 as very doubtful in my mind for the MacPro 7,1.
As you say, the only CPU's Apple is left with are the 1600 and the 2600 Broadwell-ep with 40 lanesand that still uses the C610 chipset with 8 PCIE 2.0 Lanes, so in order to get Thunderbolt 3 they need to come up with 8 more PCIE lanes.
1. Intel could build Apple a custom chipset with 16 PCIE lanes
2. Apple could cut the link speed of the second GFX card to 8x
3. Apple could use 2 1600 CPU's and only offer TB3 on the high end MP7,1
1 is the best case, 2 is the wisest move, the second GFX card doesn't do anything most of the time and we're still talking about and 8x link at PCIE 3.0 speeds. 3, I'm not even sure if the 1600 supports SMP or Apple could even fit 2 of them in the nMP case even if they did.
None of the above are very likely, what is likely is we'll see a refresh of the MacPro at WWDC 2016
mcnallym, have you seen the roadmap where Intel details the future platforms?
Purley for 2S and up, Basin Falls for 1S.
Realize that both these use different PCHs (Lewisburg and Kaby Lake) and probably different sockets as well. Remember that 2S+ will come with 6 channel mem, so this automatically should mean new socket, as well as OP fabric. 1600 should become closer to the desktop platform, at least when it comes to the PCH, and similar to what it is now.
2S and up are different beasts altogether.
Vader, 2017 is possibly a good bet but with all these delays who knows. The process is quite common then but Intel might decide to postpone release. I wouldn't count on it before middle to year's end though. Going in the nMP will be even later.
TB3 does not imply the use of Skylake. The extra 8 PCIe3 lanes in Broadwell-EP will be good for 2 TB3 controllers, each with 2 ports, each double the TB2 bandwidth. You get a reduced number of ports but increase in bandwidth.
Intel could have by now upgrade the C610 into the C620 (or whatever) so as to at least make the PCIe2 lanes into PCIe3, but that is a major overhaul and will come only with Skylake-EP, and more. USB 3.1 seems to be AOL though.
For dual CPU setup they would have to go with 2600, which would further increase the price, that we all already find (and agree) high enough
Broadwell-EP only supports 40 PCIE 3.0 lanes per CPU.
Last edited:
- Status