Let's rev up some excitement for the 2013 Mac Pro!

[kitkat99333]

There is a very good reason a 'new' Mac Pro wasn't released in 2012, because the Xeon Chipsets can't use thunderbolt.

Thunderbolt hasn't been integrated natively into the Xeon chipset yet. It multiplexes a digital video signal and a PCIe signal into one stream which is then demultiplexed at each drop. This requires both a source for the digital video, and 4 second generation PCIe 2.0 lanes. Chipsets which have Intel's FDI can provide the video straight from the chipset (hence why the chipsets with integrated graphics can support thunderbolt and why the Xeon and the rest of the socket 2011 lineup without integrated graphics can't support thunderbolt) and theoretically it could be pulled out from almost any video card that has a DisplayPort or HDMI output without much difficulty. The PCIe lanes can be tapped from the chipset itself as most have 4-8 spare lanes which are usually fanned out into 1x and 4x slots. These lanes aren't multiplexed which means they'd have to be fed into an add-in Thunderbolt controller instead of a slot on the motherboard. This cannibalizes the user's ability to use add-in PCIe cards.

As soon as Thunderbolt is added natively it will most likely receive its own dedicated PCIe lanes which are multiplexed on chip. This doesn't cannibalize other PCIe expansion slots and reduces the complexity of the traces on the board.

In simple terms, until thunderbolt is added natively to the chipset ( presumably will be in the ivybridge-E and server grade processors line up) Apple won't release a 'new' generation Mac as they want there entire lineup to carry thunderbolt.

 

[grizfish]

I had a dream that the new Mac Pro came out and the only accessible components were in the drive bays. Also, the reason for the late update was caused by problems with the manufacturing process used to assemble the new apple monitors. All new Macs required this new display and it could not be used in target mode.

Say it aint so!

This waiting is messing with my sleep!

 

[Phrygian]

I was under the illusion for a long time that Apple had a soul. They don't.
As long as they are making the big bucks, they are happy.

Welcome to reality. one less fanboi at a time..

 

[tekboi]

I personally gave up waiting on a MP refreshes and went with the iMac...


I don't regret my decision one bit. Maybe if apple can get it together I'll consider making my next purchase (3-5 years from now) a MacPro.

 

[bearcatrp]

I personally gave up waiting on a MP refreshes and went with the iMac

This is what apple whats you and everyone else wanting a new mac pro to do. Move on to an iMac or mini. Mac Pro is dead folks. Move on.

 

[StephenCampbell]

I'm actually back in the game folks! Some unexpected jobs showed up and I'm able to put into savings enough to drop about 5k on a machine later this year. I would never have been happy with a mini or iMac anyway. I need a normal computer where the parts are accessible and upgradeable.

I'm still very excited about the possibilities! We should keep this thread going until it's released.

 

[OS6-OSX]

Mr. Campbell I tried to keep the post going. As you know it is now 2018 and all desktop and laptops are now illegal. There are only smart phones and tablets!
But it is good to see gas at $1.23 a gallon!

 

[deconstruct60]

There is a very good reason a 'new' Mac Pro wasn't released in 2012, because the Xeon Chipsets can't use thunderbolt.

Thunderbolt hasn't been integrated natively into the Xeon chipset yet.

Thunderbolt is not integrated into the desktop chipset either. Thunderbolt is only available as a discrete controller chip everywhere.

Right now there are 5 discrete controllers available: 5 Flavors all Intel . What is true is that can ONLY buy an Thunderbolt controller from Intel. That's what some of the others ( AMD , Nvidia , and many system vendors chaff at. )

It multiplexes a digital video signal and a PCIe signal into one stream which is then demultiplexed at each drop.

Chuckle, exactly. It takes input signals not from inside an integrated chip but from from what is readily transmitted on many motherboards already. PCI-e input which is used in a variety of contexts on a motherboard. DisplayPort signals which are passed around on every other Mac in the line-up with no problems either; including those with discrete GPUs.

So where is the "integration" in passing around signals that discrete components utilize????????????????

This requires both a source for the digital video, and 4 second generation PCIe 2.0 lanes. Chipsets which have Intel's FDI can provide the video straight from the chipset (hence why the chipsets with integrated graphics can support thunderbolt and why the Xeon and the rest of the socket 2011 lineup without integrated graphics can't support thunderbolt) and theoretically it could be pulled out from almost any video card that has a DisplayPort or HDMI output without much difficulty.

What is this "theoretical" stuff ? Intel's video doesn't come "straight from" the chipset. The video is produced by the CPU+GPU combo. This isn't output in any of the standard formats: DisplayPort, HDMI , DVI , etc. That is handled by the I/O Hub chipset. It is convenient layout wise that can get both inputs from the Intel chipset. There is absolutely nothing indicated in distributed material that it is required.

The PCIe lanes can be tapped from the chipset itself as most have 4-8 spare lanes which are usually fanned out into 1x and 4x slots.

Oh golly gee, imagine that. The Intel C600 series has x8 PCI-e v2 lanes.... just like the desktop version that is typically hooked to TB controller in Intel's reference designs.

http://www.intel.com/content/www/us/en/chipsets/server-chipsets/server-chipset-c600.html

A "problem"? Not.

These lanes aren't multiplexed which means they'd have to be fed into an add-in Thunderbolt controller instead of a slot on the motherboard. This cannibalizes the user's ability to use add-in PCIe cards.

WTF???? Not hing is multiplexed until it passed though the TB controller.

The following is a block diagram for a Intel motherboard with TB:

[ from article on some Z77 TB motherboards. http://www.anandtech.com/show/5884/...s-part-2-intels-dz77rek75-asus-p8z77v-premium ]

Note both the x4 PCI-e and DisplayPort output linkages to the discrete Thunderbolt controller. That isn't "integrated" nor multiplexed.


Xeon E5 class CPUs have more PCI-e lanes (40 ) than the much more limited desktop class designs TB is currently deployed with ( 16). As pointed out the chipsets both desktop Z77 and C602 have x8. That means a Xeon E5 system has a minimum 48 PCI-e lanes and the desktop system have 24 and yet somehow that Xeon E5 systems are laboring under some sort of "shortage". LOL. Cannibliized lanes??? .... That is choosing the desktop processor. That has nothing to do with Thunderbolt.

As soon as Thunderbolt is added natively it will most likely receive its own dedicated PCIe lanes which are multiplexed on chip. This doesn't cannibalize other PCIe expansion slots and reduces the complexity of the traces on the board.

Gibberish.

In simple terms, until thunderbolt is added natively to the chipset ( presumably will be in the ivybridge-E and server grade processors line up)

a. Apple is far more likely to use IvyBridge-EP ( Xeon E5) than Ivy Bridge E ( Core i7 49xx )

b. There are not going to be non-discrete (integrated) TB controllers any time son. The same controller is also utilized in the many of the TB perhiperals that have NO x86 CHIP whatsoever in them. So sure as heck don't needs a x86 integrated IO Hub controller there either.

It is total gibberish that these controllers can't operate outside tight coupling to an Intel IO Hub chipset.


c. There are placement issues with thunderbolt.

" ... You'll notice that on all Thunderbolt boards we've tested thus far, the Cactus Ridge controller and Thunderbolt port are very close to one another. The spec for max trace length between the Thunderbolt controller and port is two inches, compared to up to 10 inches for Intel's USB 3.0 controller. ..."
http://www.anandtech.com/show/5884/...s-part-2-intels-dz77rek75-asus-p8z77v-premium

This factor is going to keep Thunderbolt out of core chipset integration for a long time. If Intel folded TB into the chipset than designers would be forced to move the whole, much larger integrated chipset, closer to the edge of the personal computer. It doesn't make any sense at all to force that on folks. When all computers are around the size of a mini and/or laptop perhaps but as a general building block that is more than a little loopy.

Apple won't release a 'new' generation Mac as they want there entire lineup to carry thunderbolt.

It is not a design blocking issue.

 

[KaraH]

I personally gave up waiting on a MP refreshes and went with the iMac...


I don't regret my decision one bit. Maybe if apple can get it together I'll consider making my next purchase (3-5 years from now) a MacPro.

Meanwhile, after 2 decades+ of my macs being separate from my screen I decided to try the all-in-ones. An eMac, and then iMacs for my partner and myself. My 2007 iMac was getting a bit long in the tooth until we saw the 2012 model ... I considered it a slap in the face. Sure the iMac is faster right now but its internal expandability sans major surgery is the ram ... and even with that surgery things are not much better.

 

[WardC]

So let me get this straight...

There is a 2-inch maximum distance requirement for the length between the Thunderbolt ports and the chipset...

The Mac Pro processor is on the board at the bottom of the computer, the ports are up along the backside.

This would be considerably more than two inches...

So, because of this distance problem, Thunderbolt on the current board/port layout of the machine is not possible?

The only way to have thunderbolt would be to have the chipset super-close to the Thunderbolt ports?

I have never heard this before.

 

[d-m-a-x]

So let me get this straight...

There is a 2-inch maximum distance requirement for the length between the Thunderbolt ports and the chipset...

The Mac Pro processor is on the board at the bottom of the computer, the ports are up along the backside.

This would be considerably more than two inches...

So, because of this distance problem, Thunderbolt on the current board/port layout of the machine is not possible?

The only way to have thunderbolt would be to have the chipset super-close to the Thunderbolt ports?

I have never heard this before.

I'll take a Mac Pro with no thunderbolt over a MP with ancient processors, that is just fine

 

[deconstruct60]

So let me get this straight...

There is a 2-inch maximum distance requirement for the length between the Thunderbolt ports and the chipset...

The Mac Pro processor is on the board at the bottom of the computer, the ports are up along the backside.

Why are those coupled? There is absolutely zero need for a discrete TB controller to be tightly co-located with the CPU package at all.

The TB controller can be put on the main motherboard right next to where the PCI-e card slots are. TB controller needs PCI-e lanes (which are routed off the daughter card in large numbers anyway) and displayPort input (which can easily be embedded onto the mainboard. ).

This would be considerably more than two inches...

And so? it is immaterial.

So, because of this distance problem, Thunderbolt on the current board/port layout of the machine is not possible?

That is your made up story. It is obviously possible since iMacs have it.
With the exact current layout on the Mac Pro, yes it impossible because it isn't there, nor is there large empty space market "filled me in with TB".
However, what is there could be shuffled around a bit.

If kick FW off the rear panel then can move FW controller closer to front of the board. Can slide a USB 3.0 controller in also a bit farther from back panel. Nuke the support chips for the two analogy audio input and output sockets on the rear panel. Together that should open up enough room for TB controller and TB ports to be located close to the back panel.

Higher up on the board past the 4th PCI slot extend the board a bit and slide in a mobile GPU + VRAM + cooling system. Run the two DispayPort outputs back down to TB controller. run a x4 PCI-e lanes from daughter card to TB controller. And done.

The only way to have thunderbolt would be to have the chipset super-close to the Thunderbolt ports?

When you start to integrate the TB controller into larger packages it will have placement constraints. However, Intel is only targeting integrating TB controllers into SoC (system on a chip) solutions. Effectively the whole computer CPU+GPU+IO HUB+TB Controller+kitchen sink are all inside of one package. At that point the whole computer is small (e.g., MBA ) and the SoC package is close to an edge because the whole system is small. But yes, for a MBA you probably wouldn't place the SoC package exactly dead center of the box. ( not hard because it isn't done now either. See step 16 http://www.ifixit.com/Teardown/MacBook+Air+13-Inch+Mid+2012+Teardown/9457/2 )


For the Xeon E5 class though isn't anywhere close to being in SoC status. GPUs haven't been integrated. IO hub not integrated. Not in Ivy Bridge. Probably not in Haswell ( which looks to do some hocus pocus with changing the electrical assignments on the pin out configuration). Not even Broadwell.

In short, there is deep confusion between what Intel is doing for super Ultraslim laptops and the Mac Pro class systems. They aren't the same. Hand waving that SoC solutions are what is holding up the Mac Pro is gibberish.

There are always going to be discrete TB controllers beacuse the perihperals need them. There is no x86 chip so can't put an integrated SoC or even just IOhub into the peripheral. The Mac Pro can continue to use a discrete solution for a long time just as it has for FW all these years.
( Haswell Xeon E5 IOHub chips should merge in USB 3.0 so the 2015 board design would have a bit more freedom in board and/or daughtercard layout. )

 

[sirio76]

Next generation Thunderbolt

http://appleinsider.com/articles/13...olt-with-4k-resolution-support-coming-in-2014

 

[skippermonkey]

I'm vaguely working on the assumption that this could be my last MacPro (I've updated every two or three years since my old Power Macintosh 7200). If so I'll be throwing huge amounts of cash Apple's way to get the very best one I can, with the biggest, most bad-ass GPU they sell. And then hope it lasts a long time...

 

[bsbeamer]

There's been a few rumors at NAB about a new MacPro, but nothing pointing to an announcement or any sort of release coming soon from what I'm gathering. A few contradictory articles that all seem to be based on the same sourced information, which all point to possible delays and then an announcement at WWDC.

What's clear is that the technology isn't 100% ready for what people want as a new, fully refreshed and updated machine. There could be another "update" like the one from last year, but that just ended up angering the masses more than anything, so I doubt we'll see that again. Though USB3 and PCIe3 would be more than enough for most, add in SATA6G and I'd buy one without question...

My deadline was NAB to figure out my roadmap for upgrades/purchases moving forward. Like many others I've spoken with, the PC solutions are readily available and cannot be ignored. I've considered an expander chassis "upgrade" (like the ones from Cubix), but the price and setup with them really makes me question why I'm not building a hack-pro instead. Personally, giving it until the end of the week before I make any rash decisions either way.

The new software that's being discussed from Avid seems to have a June-ish timeframe for release. Adobe's new offerings will probably fall closely to that as well. There are no technical requirements for either, so it'll be interesting to see what's recommended/required when they do come out and what Apple offerings are listed.

 

[goMac]

Why are those coupled? There is absolutely zero need for a discrete TB controller to be tightly co-located with the CPU package at all.

Yes there is. On the i series Core chips, the PCI controller is on CPU. The further you move the Thunderbolt controller from the PCI controller, the more flakey it will behave. Same reason why i series Core chips and older high performance Xeons have moved the RAM extremely close to the GPU.

Physically longer lanes introduce all sorts of odd latency and possibility for data degradation, so two inches doesn't sound unreasonable, even if that could possibly be stretched a bit.

 

[calaverasgrande]

I'll take a Mac Pro with no thunderbolt over a MP with ancient processors, that is just fine

Yeah, but will Apple be willing to release a new Mac Pro model in 2013 with no thunderbolt?
Right now they are carrying two different displays (TB display and LED display) pretty much just to accommodate the Mac Pro.

 

[deconstruct60]

Yes there is. On the i series Core chips, the PCI controller is on CPU.

The isn't the only one. And that one on the CPU is typically grossly oversubscribed. There is a block diagram up above that clearly demonstrates there is zero necessity. None of the Z77 boards with TB co-locate with the CPU.

The further you move the Thunderbolt controller from the PCI controller, the more flakey it will behave.

Not true. In part because the data is split along the 4 PCI-e lanes. Part of the problem is reducing data traffic that was meant for 4 lanes of bandwidth down to 2 ( or any number smaller than 4.)

Same reason why i series Core chips and older high performance Xeons have moved the RAM extremely close to the GPU.

And the number of RAM lanes are far more limited on the desktop versions than with the 4 channell server models.

Physically longer lanes introduce all sorts of odd latency and possibility for data degradation,

If it presents completely insurmountable problems then how to PCI-e slots work?????

 

[goMac]

If it presents completely insurmountable problems then how to PCI-e slots work?????

PCI-e slots are very sensitive to trace lengths. This is pretty basic EE stuff. I'm not even EE and I learned it. Traces to the same component should be as short as possible and all the same length.

It's possible Intel's guidance might be more around a video signal coming directly from the CPU. It's hard to tell why they would have a number of two inches. But that they have a number at all doesn't sound unreasonable.

Trace lengths are the obvious reason a lot of the Mac Pro has such an odd design. Putting the RAM onto daughter cards next to the CPUs would enhance performance by reducing latency and interference. The PCI slots (especially the high performance ones) are probably within 2-3 inches away from the CPU as well. There could also be secondary controllers on the board bridging the connection (which seems likely given the daughtercard design.)

EE isn't my thing, but I know enough that Intel having some sort of requirement on trace distance doesn't sound crazy.

 

[deconstruct60]

Yeah, but will Apple be willing to release a new Mac Pro model in 2013 with no thunderbolt?

Given Apple's tendencies to OCD, growing part component volume with inclusion in multiple Mac offerings, and that the 2013 motherboard is somewhat likely be the 2014 motherboard board ( Sandy Bridge 2012 , Ivy Bridge 2013 ) .... pretty high it will be there.

The simplest move would be to just take one of the discrete GPU options from the iMac and include it in all Mac Pros. Consistent parts across multiple Mac models is entirely within their modus operandi. So GPU+VRAM+TB controller all get bumped up in volume.

Right now they are carrying two different displays (TB display and LED display) pretty much just to accommodate the Mac Pro.

Tail wagging the dog. Those displays aren't primarily targeted at the Mac Pro. Examine the MagSafe connectors and permantetly attached fixed length display cable, that is clearly evidence that they are not.

Frankly, what the Mac Pro is missing is something that isn't trying to be a docking station and is much more focused on being a display.

It is not any harder to have variant mDP monitor as it is to construct the TB ( and mDP ) "Displays" as being a variant of the iMac. What is likely more difficult is to continue to justify the same price for both. That is pretty much a farce to be Thunderbolt-less and the same price.

The mDP "display" actually works with both TB and non-TB Macs.... which is actually more coverage than the TB display has.

This whole TB Displays drive TB on Mac Pro is a colossal tail wags dog inference. It makes no sense at all.

Need to use external drives/IO capture tool/"expensive TB device" across Mac laptop and stationary Mac Pro... maybe. It nice but not always necessary. But TB display? Not.

Logo fixated folks will happen to buy more TB displays than the laptops can sell by themselves. That is just dubious margin increase. Apple just isn't that competitive in the display market; it won't be a large marginal increase. An up-to-date Mac Pro would generate more revenue and growth.

----------

PCI-e slots are very sensitive to trace lengths.

Of course they are, but not to the same extent. Everything is sensitive to trace lengths. You can easily split the distance with "half" being PCI-e lanes and the other "half" being controller-socket. More likely it will be 3/4 PCI-e and 1/4 controller-socket. Namely you can get a longer distance by using both.

I don't think you even bothered to look at the iFixit presentation of the MBA board. the CPU and TB controller are on opposite sides of the MBA. This whole close proximity to CPU thing is smoke.

 

[goMac]

I don't think you even bothered to look at the iFixit presentation of the MBA board. the CPU and TB controller are on opposite sides of the MBA. This whole close proximity to CPU thing is smoke.

The TB is near the platform controller, which controls the PCI lanes and graphics in the Macbook Air.

I've also heard rumblings that Thunderbolt does require support in the CPU. That one doesn't make much sense to me, but the sources I've heard that from would normally be reliable and in the know. Something about how even though Thunderbolt just uses PCI lanes, the CPU has to be aware of which PCI lanes the Thunderbolt controller is on.

 

[deconstruct60]

The TB is near the platform controller, which controls the PCI lanes and graphics in the Macbook Air.

Not sure why you used that picture. Here is the more direct one.

http://www.ifixit.com/Teardown/MacBook+Air+13-Inch+Mid+2012+Teardown/9457/2

Thunderbolt controller in yellow on the left. The CPU in red on the right.
Yes the orange IO Hub /Platform controller has the source of the PCI-e lanes the TB controller is "consuming". For this generation the video is routed out of the CPU and into the IO Hub. Future generations the video will likely come straight out of the CPU+GPU package.

If the TB controller , IO Hub, CPU , and GPU were all combined in one package it would "land" where the TB controller is now. That other stuff would either disappear inside the new package or get moved into the now far more empty spaces left behind.

I've also heard rumblings that Thunderbolt does require support in the CPU. That one doesn't make much sense to me, but the sources I've heard that from would normally be reliable and in the know. Something about how even though Thunderbolt just uses PCI lanes, the CPU has to be aware of which PCI lanes the Thunderbolt controller is on.

Of course it does to some extent. The TB controller appears to the rest of the computer as a PCI-e switch. Since there has to be one hierarchy of PCI-e switches ( so locations and addresses can be coherently handed out ) the PCI-e controller in the CPU needs to be aware of both the controller in the IO and any downstream PCI-e switches from there.

PCI-e is a packet based network. The network's switches need to know where to send packets for a given address. Or that operating multiple PCI-e networks (cpu's and the IOhubs ).

There are some firmware issues that deal with setting up networks upon boot ( and some hot-plug changes to the network), but core support for that should already be in the CPU. The number of switches present kind of explodes with a fully populated TB daisy chain, so could be some corner cases where someone cut corners in CPU but seems doubtful. More likely it is about someone getting in and just doing some work, not the hardware.

 

[d-m-a-x]

Yeah, but will Apple be willing to release a new Mac Pro model in 2013 with no thunderbolt?
Right now they are carrying two different displays (TB display and LED display) pretty much just to accommodate the Mac Pro.

I'm sure they will. The last WWC was a real bummer though, it seemed to go from bad to worse. They should have released a sandy bridge MP

 

[Lil Chillbil]

I'm sure they will. The last WWC was a real bummer though, it seemed to go from bad to worse. They should have released a sandy bridge MP

I know, and where was I when this was going on... trying to watch the livestream from a g4 tower 400mhz

 

[Umbongo]

E5-2600V2s will have up to 12 cores each. 24-core Mac Pro is possible.