Are these the future specs of the new mac pro?

[jbg232]

Cult of Mac claims these are the new specs. Not too much information but it's nice to have something, even if it turns out to be fake.

Here's the original site.

Here are their predictions:
Intel Xeon E5 series processors, six and eight core possible
1600 MHZ memory with 8 channels (25% more physical slot capacity)
SATA III/SAS 6GB/s
Native USB 3.0 and Thunderbolt
PCIE3 native for video cards and others

 

[Neodym]

Must be either some translation error in that article or it is simply fake: Talking about Apple actively deciding against Ivy Bridge in the new MacPro (there is no such thing as Ivy Bridge Xeons currently) and the argument is about "handling voltages, which would negatively impact machine life".

Sounds really strange to me...

 

[Umbongo]

No more information there than you would have found reading about the C600 chipsets and E5 Xeons on Intel's site.

All of micgadget's stuff on Mac Pros is full of made up stuff from "sources". It's to drive traffic and has no basis in reality. Don't give them any credit as it serves to confuse the Mac Pro user base.

 

[goMac]

Obviously fake. The Mac Pro is dead. Apple doesn't care about pros, and those are pro features.

 

[deconstruct60]

... Not too much information but it's nice to have something, even if it turns out to be fake.

Here's the original site.

Here are their predictions:
Intel Xeon E5 series processors, six and eight core possible
1600 MHZ memory with 8 channels (25% more physical slot capacity)

This is only new and significant if they are saying that there will be no usage of the E5 1600 models. There aren't 8 channels on a single E5. If all Mac Pros have 8 then that would mean that there are only dual package models.

That's a bit weird. The entry price of Mac Pros would probably rise to the $3,000 level. I'm sure that will be good for several days of wailing on these forums.


It is doubtful there will be 25% more DIMMs slots. There are 4 now. 25% more is 5. That's a pretty whacked configuration. Before the Mac Pro has "odd ball" DIMM config because one memory channel had 2 DIMM slots on it and the others had just 1. Hence the debates in these forums whether filling just 3 larger DIMMs was better the filling all 4 DIMMs that have flared up from time to time.

Now that Intel has put 4 channels on the E5 to exactly match the 4 DIMMs slots that Mac Pros have had ....... toss that parity down the drain again.... just to keep it uneven.... That's almost neurotic design.

SATA III/SAS 6GB/s

E5 means SATA III support chips. Not particularly incrementally more information. Likewise, E5 means PCI-e v3.0.

Native USB 3.0 and Thunderbolt

not sure what "native" USB 3.0 means. I suppose it means "on the motherboard". Not a huge newsflash because just about every other E5 motherboard out there targeted workstations has USB 3.0.


Thunderbolt. Dual E5's certainly would make embedding a GPU to implement Thunderbolt much easier. Just take a GPU from an iMac (a standard part Apple can scale purchase volume on) and use the glut of PCI-e lanes to hook it and the Thunderbolt controller. Consumes 20 lanes and has no impact on the other 4 PCI-e slots in any way.

Whether that is worth cranking up the entry level price several hundred dollars is another story.

----------

Must be either some translation error in that article or it is simply fake: Talking about Apple actively deciding against Ivy Bridge in the new MacPro (there is no such thing as Ivy Bridge Xeons currently)..

Shhh, don't tell Intel....


http://ark.intel.com/products/family/59137/Intel-Xeon-Processor-E3-Family/server

See all the ones marked V2 ? Those are Ivy Bridge Xeons. Xeon E3's.
They exist. Apple does have the option of re-imagining what a "Mac Pro" is by switch to that line-up. It would be about as weird as going dual package only.

However, if Thunderbolt was the #1 design priority it would make implementation very easy.

 

[xgman]

Not really anything new or unexpected here. Seems like educated guessing.

 

[goMac]

Thunderbolt. Dual E5's certainly would make embedding a GPU to implement Thunderbolt much easier. Just take a GPU from an iMac (a standard part Apple can scale purchase volume on) and use the glut of PCI-e lanes to hook it and the Thunderbolt controller. Consumes 20 lanes and has no impact on the other 4 PCI-e slots in any way.

Still puts them in a bit of an odd spot. I'd guess Apple would still like to keep PCI-E GPUs, but stop shipping MDP displays.

Maybe Apple will keep making MDP displays in since pros don't need all the ports on the display. It just seems very odd to me. If Apple wants people to use Thunderbolt displays, they'd have to put in a high end GPU, but then what kind of displays would people use with the dedicated card...

Some sort of dongle option seems far more likely to me.

 

[deconstruct60]

Still puts them in a bit of an odd spot. I'd guess Apple would still like to keep PCI-E GPUs, but stop shipping MDP displays.

Why would they want to stop selling displays? As long as they can sell sufficient numbers , they are profitable, and there is growth...... where is the down side.

"mini DisplayPort" doesn't have to loose for "Thunderbolt" to win. In fact, the more mini DisplayPort monitors there are on the market the more traction Thunderbolt has. It is a more standard connector which will likely get picked up my more system vendors in their designs.

If Apple has a more plain display at $799 and the fancier TB Display (docking station) at $999 then they really wouldn't conflict with one another much. Especially if they are the same LCD panel.

Likewise at miniDP connector Display with high gamut 10-bit color that $1299 versus $999 TB offering are equally as separated.

In addition, their are many Mac Pro users that don't use Apple displays. Why would Apple want to make it more difficult to hook up? Getting standard reference design GPU PCI-e cards working in the Mac Pro is only a plus.

Apple wants people to use Thunderbolt displays, they'd have to put in a high end GPU,

this is a dubious notion. Apple wants folks to leverage Thunderbolt where appropriate. They didn't help create Thunderbolt solely as a gimmick to sell proprietary displays. The Mac Pro doesn't need a TB display in any way.

Some sort of dongle option seems far more likely to me.

I don't think so. I don't think having users screw around with the TB controller inputs is very "Apple like" in design. A design where the TB controller inputs are fixed in place is more their style.

 

[fox10078]

Why would they want to stop selling displays? As long as they can sell sufficient numbers , they are profitable, and there is growth...... where is the down side.

"mini DisplayPort" doesn't have to loose for "Thunderbolt" to win. In fact, the more mini DisplayPort monitors there are on the market the more traction Thunderbolt has. It is a more standard connector which will likely get picked up my more system vendors in their designs.

If Apple has a more plain display at $799 and the fancier TB Display (docking station) at $999 then they really wouldn't conflict with one another much. Especially if they are the same LCD panel.

Likewise at miniDP connector Display with high gamut 10-bit color that $1299 versus $999 TB offering are equally as separated.

In addition, their are many Mac Pro users that don't use Apple displays. Why would Apple want to make it more difficult to hook up? Getting standard reference design GPU PCI-e cards working in the Mac Pro is only a plus.




this is a dubious notion. Apple wants folks to leverage Thunderbolt where appropriate. They didn't help create Thunderbolt solely as a gimmick to sell proprietary displays. The Mac Pro doesn't need a TB display in any way.




I don't think so. I don't think having users screw around with the TB controller inputs is very "Apple like" in design. A design where the TB controller inputs are fixed in place is more their style.

Thunderbolt can drive MDP monitors, and DVI ones, and HDMI. Seems like exactly what apple would do. Go all thunderbolt and sell adapters.

Apple not having thunderbolt working exactly the same way as it does on iMacs and MBP would be very "unApple like". I imagine that they have figured out a way to make it work, though I'm not sure how and I could be quite wrong.

 

[deconstruct60]

Thunderbolt can drive MDP monitors, and DVI ones, and HDMI. Seems like exactly what apple would do. Go all thunderbolt and sell adapters.

If it takes a kludge to implement Thunderbolt it is gratuitous. Adding gratuitous stuff is not typically the Apple approach to system design.

It isn't a matter if the port is backward compatible. A TB monitor has a dangling magsafe power cable. Apple designs something so there is extra cables lying around?????

Apple not having thunderbolt working exactly the same way as it does on iMacs and MBP would be very "unApple like".

If the Mac Pro needed Thunderbolt you might have a point. It doesn't need it. This is much more consistency purely for consistency sake. (sure there are some corner cases but there is no inherient need across a broad range of the Mac Pro target audience). There must be Thunderbolt because Thunderbolt must be everywhere. That's kool-aid drinking talk, not technology meets liberal arts.

Do iMacs/Mini's have MagSafe connectors (like most Macs sold ) ? No.
Do iMacs have ExpressCard sockets (like some MBPs) ? No.
Do iMacs have two DVDs (like Mac Pros) ? No.
Do MBA's have FW ports ( like most other Macs) ? No.


Thunderbolt isn't naturally aligned with the other Mac models because Thunderbolt presumes that a Display Port signal will be present on the motherboard. The Mac Pro is different because this is not naturally the case. With all other Macs it is.

Could Apple come up with a solution to put it onto the motherboard? Sure. But if it isn't there, that's a clue that it isn't really aligned with Thunderbolt in the first place. The question to settle is why making this unaligned thing aligned. Hand rationals "well so it will be like the others" are goofy. That isn't a "why". That is a excuse to do it anyway.

 

[goMac]

Maybe they'll have some sort of GPU firmware solution.

I'm just not sure I see them using some onboard GPU for the Thunderbolt ports. That pretty much either means they've basically gimped the GPU system, or that people aren't going to use the Thunderbolt ports for display anyway.

 

[handsome pete]

This is only new and significant if they are saying that there will be no usage of the E5 1600 models. There aren't 8 channels on a single E5. If all Mac Pros have 8 then that would mean that there are only dual package models.

That's a bit weird. The entry price of Mac Pros would probably rise to the $3,000 level. I'm sure that will be good for several days of wailing on these forums.

I don't think they're implying that they will only offer dual processor configurations, but Apple could offer a single 2687w option.

The Ivy Bridge mention doesn't make sense though. Seems pretty obvious to anyone that they'll use Sandy Bridge Xeons.

 

[MatthewAMEL]

Must be either some translation error in that article or it is simply fake: Talking about Apple actively deciding against Ivy Bridge in the new MacPro (there is no such thing as Ivy Bridge Xeons currently) and the argument is about "handling voltages, which would negatively impact machine life".

Sounds really strange to me...

I think what they are referring to is the 'problem' that all steppings of Ivy Bridge have now with current leakage. The Ivy Bridge die runs significantly hotter (up to 20C) than Sandy Bridge at same speeds.

http://en.wikipedia.org/wiki/Ivy_Bridge_%28microarchitecture%29#Heat_issue_when_overclocked

BTW- The Xeon E3s are Ivy Bridge. They are designed for single-processor machines. E5s (Sandy Bridge) are available for up to 4 processor machines.

 

[fox10078]

The Xeon E3s are Ivy Bridge. They are designed for single-processor machines. E5s (Sandy Bridge) are available for up to 4 processor machines.

Just thought of what a mac pro with 4 processors would be like

 

[theSeb]

If it takes a kludge to implement Thunderbolt it is gratuitous. Adding gratuitous stuff is not typically the Apple approach to system design.

It isn't a matter if the port is backward compatible. A TB monitor has a dangling magsafe power cable. Apple designs something so there is extra cables lying around?????



If the Mac Pro needed Thunderbolt you might have a point. It doesn't need it. This is much more consistency purely for consistency sake. (sure there are some corner cases but there is no inherient need across a broad range of the Mac Pro target audience). There must be Thunderbolt because Thunderbolt must be everywhere. That's kool-aid drinking talk, not technology meets liberal arts.

Do iMacs/Mini's have MagSafe connectors (like most Macs sold ) ? No.
Do iMacs have ExpressCard sockets (like some MBPs) ? No.
Do iMacs have two DVDs (like Mac Pros) ? No.
Do MBA's have FW ports ( like most other Macs) ? No.


Thunderbolt isn't naturally aligned with the other Mac models because Thunderbolt presumes that a Display Port signal will be present on the motherboard. The Mac Pro is different because this is not naturally the case. With all other Macs it is.

Could Apple come up with a solution to put it onto the motherboard? Sure. But if it isn't there, that's a clue that it isn't really aligned with Thunderbolt in the first place. The question to settle is why making this unaligned thing aligned. Hand rationals "well so it will be like the others" are goofy. That isn't a "why". That is a excuse to do it anyway.

I cannot imagine the new Mac Pro not having thunderbolt. PC motherboards are coming out with TB and it supports discrete graphic cards via the mDP output. I just cannot find technical details on how it is being done. I assume Apple will do the same.

The Thunderbolt port on both boards can serve as either a Thunderbolt port or a DisplayPort output, similar to the behavior on a Thunderbolt Mac. Discrete GPUs are supported through the on-board Thunderbolt/DisplayPort output, provided you have Lucid's Virtu software installed.

I believe the Thunderbolt controller on the GD80 is this little chip on the board, just behind the analog audio outputs on the IO panel. It also looks like the controller branches off of the Z77 PCH's PCIe 2.0 lanes. Although there are four PCIe x1 slots on the board, you can only use two at a time - likely because of the way the Thunderbolt controller connects to the system as it uses 4 of the 8 available PCIe 2.0 lanes. The remaining lanes aren't enough to support the on-board LAN, Firewire, audio and SATA controllers in addition to the extra PCIe slots and Thunderbolt. As a result there are limitations on what you can use in parallel (e.g. you can't use Firewire and SATA ports 7/8) in order to deal with the PCIe overprovisioning.

From a software perspective, Thunderbolt is treated just like another display output driven by Intel's processor graphics. I installed a GeForce GTX 680 along with Lucid's Virtu GPU virtualization software to see if I could use the 680 for gaming but drive the display using Intel's processor graphics and the Thunderbolt port. The setup worked flawlessly.

Virtu recognized the configuration immediately once I had NVIDIA's drivers installed, and I was able to run the 680 headless - using only the Thunderbolt port to drive the external display. Intel's HD 4000 powered things in Windows, while the 680 kicked in for games.

 

[initialsBB]

No more information there than you would have found reading about the C600 chipsets and E5 Xeons on Intel's site.

Am I missing something ? When I look at the C600 series on Intel's site they make no mention of USB3 or PCI Express 3 ?

 

[24Frames]

deconstruct60, is there any reason why they couldn't implement Thunderbolt as a high speed data transfer solution without the ability to connect a display?

That would seem to be the most sensible solution, as the mini DisplayPort / DisplayPort connectivity on the GPU cards seems adequate for anything a Mac Pro user is likely to want to do. Also I note that Apple continue to sell the 27-inch CINEMA Display alongside the TB display. In the past there have been periods when Apple didn't sell any displays that would work with a Mac Pro, so the fact that they continue to sell the 27-inch CINEMA Display may be because they intend to supply it with 2012 Mac Pros.

USB 3.0 doesn't need to be in the chipset for Apple to implement it. I am sure that 2012 Mac Pros will have USB 3.0, as all the other Macs will be getting it (Intel implemented it in the chipsets for Ivy Bridge that will be used in MBPs, iMacs...) and it would seem odd for the Mac Pro not to have it.

You can pretty much work out the specs from a close look at Intel's, AMD's and NVidias new CPU and GPU offerings and choosing the ones with similar price point to those used in the 2010 Mac Pro model components. The linked articles don't seem to make a very good job of it. Other threads here have done a lot better!

It would be interesting if the single CPU models used dual CPU boards allowing the user to add a second CPU when funds allowed. This would be great, worth putting up with a week or two of moaning about the high price points. I think it is unlikely that Apple will do this due to the price points of the CPUs.

 

[nanofrog]

...is there any reason why they couldn't implement Thunderbolt as a high speed data transfer solution without the ability to connect a display?

This is not only possible, but easier to do, as there's no need to make provisions for display data.

But the question is whether or not Apple would take such an approach or not.

 

[jbg232]

For anyone interested in more in-depth reading, Tom's hardware has a really good review (better than wikipedia) of the new xeons and great explanations of the new chipset features here.

 

[xgman]

Could Apple come up with a solution to put it onto the motherboard? Sure. But if it isn't there, that's a clue that it isn't really aligned with Thunderbolt in the first place.

I just saw yesterday the first motherboards for PC's with TB on-board are selling.

 

[dbit]

So I'm assuming all the mac pro's will use xeons, including the single socket?

Most of the tests I've seen show the i7 3930k basically offering the same performance as a single xeon for about half the price. It seems the xeon e5's value really kicks in with multiple processors. Also a lot of their benefits seemed to be in the efficiency department for machines working as 24/7 servers.

I was looking at some pc workstation configurations for 3D and video production and the xeon configurations (especially single socket) just didn't make sense price/performance wise. Over on the adobe premiere hardware forums many video pro's and machine builders have opted for 3930k machines over xeons for their day to day working machines unless they needed dual processor. Sure a lot of them send out to render farms or nodes as well, but they preferred the faster speeds of the 3930k for daily work as well as for the value. I couldn't justify an extra $1000-$1500 for the xeon configurations based on the performance increases I was seeing in their real world tests as well as benchmarks.

 

[danwellsvt]

I interpreted the memory change to be 25% more memory CHANNELS, not slots - getting rid of the piggybacked DIMM, but still providing four slots. I'm assuming that the "25% increased capacity" is just bad translation from Chinese to English, although it does make the optimum RAM sizes all 25% larger if you make the piggybacked DIMM into a real channel. 32 GB (perhaps the most popular Mac Pro RAM size) is no longer a compromised configuration.

I wouldn't think the Thunderbolt video question is too hard - Apple's never been shy about enabling full functionality only with certain GPUs. I bet we get customized GPUs with an internal Mini DisplayPort on the top of the card instead of the back panel (in addition to any back panel connections) - it's easy for Apple to run a lead from the internal Mini DP to the Thunderbolt circuitry on the motherboard. If you upgrade with an Apple-blessed card, it's just one more wire to connect, no different from the power lead. If you upgrade to a non-Apple card, it'll either be no video over Thunderbolt or you have to snake a cable into the case. Apple (or more likely a third party) may also release an external Y adapter allowing Thunderbolt displays to connect to the Thunderbolt port for data plus a Mini DisplayPort for video, but this won't be the default solution - it would be included with third-party video cards, while the stock cards connected internally.
How do we know (or suspect) that there will be no E5-1600 processors? The E5-1650 looks like a very attractive processor for the entry level, and it matches the specs we've seen (it's an E5, plus it's got 6 cores). It's 30% cheaper than an E5-2643 with a similar clock rate and only 4 cores, and a third the price of a 6 core E5-2667 that is only almost as fast). It is basically a rebadged Core i7-3930k, but it is also almost exactly the same price as a 3930k. Apple would probably use the Xeon version, just to say "Mac Pros are Xeons". If I were Tim Cook, I'd think something like this:

Single E5-1650 in the entry level (about the same clock as the top iMac, but 50% more cores). I probably wouldn't bother with the E5-1660 ($400 for 100 mHz speed bump and extra cache), unless some applications REALLY want the extra cache...
Unfortunately, it's against Apple's DNA to make this machine a more reasonably priced minitower! It would be a great xMac, clearly differentiated from the iMacs below and the dual-proc Mac Pros above.

Next step up is either DUAL E5-2643s (total of 8 cores, each core a little faster, but another $1100 in processors), or dual E5-2640s (takes a significant clock speed hit, but 12 cores for the cost of 8 faster ones). If the 6-core is an xMac, start the Mac Pro at 12 cores with dual E5-2640s.

Top end is dual E5-2670s for 16 cores, possibly with an option for dual E5-2687w's. A dual E5-2687w system contains a ton of money in processors, and would sell for $6500 or so, BUT it is as powerful as FOUR top iMacs chained together...

 

[deconstruct60]

deconstruct60, is there any reason why they couldn't implement Thunderbolt as a high speed data transfer solution without the ability to connect a display?

Since Intel is likely following the K.I.S.S. philosophy yes. It is much simpler , and yes consistent, when something that looks and feels like a mini Display Port on a computer acts like a mini Display Port on a computer.

So for any computer that is has a Thunderbolt socket there should be video that comes out. It does not[/B make sense to make the universally consistent for all devices with TB sockets. You would almost be guarantted to get a number of product support calls that went

" I plugged in my monitor. The plug fits in there but there is no view. My PC is broke. You need to fit it."

Telling the the user "Well you have the special case were it isn't connected" PC isn't simple. If it was simple it would "just work".


Additionally, to get better leverage TB needs a standard socket to roll out on . If Thunderbolt showed up with "yet another incompatible" socket the adoption rate would be much slower. They tried to take a stab at usurping USB. The USB folks shot that down (for excellent reasons not the least of which is the "war" that regularly breaks out where on either side claim the other is doom. "USB 3.0 dead long live TB " "death to TB, long live USB 3.0". )

Intel (and Apple I guess) probably worked out a deal with the Display Port folks. The Display Port folks let TB somewhat usurp the Display Port socket and Intel will advocate that every general PC they certify TB on pumps out Display Port signals. It is a win/win. DP gets more widespread distribution ( so more Display Port monitors pop up on the market). TB gets a socket and frankly more DP monitors at the end of TB chains gives the socket much larger value proposition for the first 2-3 years of its lifetime.

That would seem to be the most sensible solution, as the mini DisplayPort / DisplayPort connectivity on the GPU cards seems adequate for anything a Mac Pro user is likely to want to do.

Right! That why there is no desperate need for TB. The GPU cards already have multiple video outputs that hook to hundreds of monitors.

But the same is true for the other stuff also. There are PCI-e SATA/SAS cards that hook to hundreds of SATA devices. Pro Audio card that hook to external Pro audio boxes. Pro Video cards that hook to external pro video monitors/devices. etc. etc. etc.

Also I note that Apple continue to sell the 27-inch CINEMA Display alongside the TB display. In the past there have been periods when Apple didn't sell any displays that would work with a Mac Pro, so the fact that they continue to sell the 27-inch CINEMA Display may be because they intend to supply it with 2012 Mac Pros.

As long as people buy them large enough (by Apple's def) they will sell them. There are still non TB macs out there they are quite useful on that are not Mac Pros. There probably a higher attachment rate to the sale of new Mac Pros since those other Macs aren't sold anymore. But there are a substantially larger number of those other devices.

USB 3.0 doesn't need to be in the chipset for Apple to implement it.

It depends upon how lazy one thinks Apple is trying to be. The discrete USB 3.0 controllers are from different vendors. If they are trying to be Scrooge Mc Duck and do the absolute minimum work and still try to give lip service to providing robust standard USB drivers then there is a problem.

While the mini and laptops are space constrained, Apple's commitment is kind of questionable when the 2011 iMacs turned up and there was no USB 3.0 . I suppose they were shooting at one "major new component at a time" with the design updates ( insert TB 2011 , insert USB 3.0 2012). I also wouldn't be surprised if USB 3.0 was only really supported in 10.8 (Mountain Lion). It made sense to wait for the Intel core chipset integrated USB 3.0 controller for the iMac.

If Apple is only going to support the Intel controller then it be Haswell Xeon E5 before the support chipset got USB 3.0 . [ there likely will not be an Ivy Bridge E5 chipset update. ]

I think Apple will do it ( perhaps Mountain Lion only), because it is kind of ridiculous to "too poor to do the work" on a box that is 3-4 times the average PC unit cost.

It would be interesting if the single CPU models used dual CPU boards allowing the user to add a second CPU when funds allowed.

Price/Performance wise it wouldn't be. Those offerings tend to deficient along that metric. They are largely a gimmick for those who either wistfully plan to some day upgrade (but never will) or have some short term purchasing problem ( can't buy the two that they actually really need right away).

 

[deconstruct60]

I cannot imagine the new Mac Pro not having thunderbolt.

None of the newly announced HP, Dell, or any of the other E5 based workstation vendors have it. None of the not so newly announced i7 39xx based workstation vendors have it. None of the standard motheboards from typical vendors SuperMicro, MSI , Asus ,etc. for these SB-E/SB-EP processors have it.

Nobody. So it is quite imaginable.

PC motherboards are coming out with TB and it supports discrete graphic cards via the mDP output. I just cannot find technical details on how it is being done. I assume Apple will do the same.

The mDP output from the iGPU. Sure. A couple were announced months ago at CeBit.

I just saw yesterday the first motherboards for PC's with TB on-board are selling.

And the universally true fact among all of those I have seen is that those motherboards all host a Core i CPU package with a iGPU in it.

You folks completely ignore the most signficant factor. If Display Port singals are naturally on the motheboard ( either there is an embedded mobile GPU card and/or there is a integrated GPU unit in the CPU package ) those are naturally aligned with the Thunderbolt objectives. It is a 'no brainer' to add it.

TB on board highly enhances the GPU being on board. It only moves Intel's ( and AMD's ) merged CPU/GU agenda forward that much more faster.

Decoupling TB from the onboard GPU doesn't. That in addition to the "expectation mismatches" set up by TB sometimes does and sometimes not, makes for completely in Intels non interest to push for "data only" PCs at this time.

When TB is more mature and widely adopted perhaps there will be a push for an alternative "data only" socket. We'll see, but that is not likely to come for at least a couple of years, if ever.

I think is more likely that Xeon E5's will pick up a iGPU unit around the Haswell update. At that point the E5's will be just like the other Core i models that Apple uses. They'll have GPUs and can use practically the same exact methodologies to hook up to the TB controller on Mac Pro as on the other Macs.

It is a matter getting the server folks to fess up to the fact the GPU can actually get more significant computational workload done than the generic x86 cores. In two years or so that will be more clearly evident.

 

[Umbongo]

Am I missing something ? When I look at the C600 series on Intel's site they make no mention of USB3 or PCI Express 3 ?

What I should have said is more along the lines "Micgadget aren't saying anything that isn't already known about the platform yet are saying it in a way that suggests they have deep moles inside Apple that tell them unknown facts." They use this method of reporting to get credibility as a reliable rumour source when they are not. They gamble on the likelihood of things by saying they are certainties and they know because they have the inside scoop. They usually fail.

PCI-Exprss 3.0 is covered under Xeon documentation as it is on-board the CPU. USB 3.0 will be provided by a separate controller as is done on pretty much every other X79 and 5520 LGA 2011 board. Apple aren't going to implement USB 3.0 across all their other Macs and not put it on the Mac Pro.