New Mac Pro this May?

[deconstruct60]

That would require an new motherboard and chipset..

And? A Xeon E5 (v1 or v2 ) would require a new motherboard and chipset. An really upgraded Mac Pro needs a new motherboard and chipset no matter what. Intel is actually withdrawing from retail market the 3500/3600 CPUs in September.

If Apple is drinking Thunderbolt kool-aid by the gallon an E3 makes it easy to deploy a solution that implements it. Sure capped at 32GB and about one (maybe two ) PCI-e slots but it can cover the $2,000-3,000 zone Mac Pro users and a subset of folks grumbling about xMac. They would need to add something significant like dual 10GbE ports to help support a Mac Pro like price point but that is a bit overdue also ( 1GbE has been standard for around a 10 years and frankly wireless is about to catch up to wired 1GbE. If it is going to stick around it needs to be faster. )

Apple could release both E3 and E5 solutions. (E3 at WWDC causing mass confusion and then in the Fall a E5 bump when that upgrade was complete on components ) But it would not be surprising if they just cut off the higher priced and likely slower growing subset of the Mac Pro market. If they cut off the evolutionary paty the E5 I think they should come up with new name. It isn't really the same product at that point.

 

[deconstruct60]

I'm also not convinced that Apple intends to go the other direction with pricing, seeing as the Mac Pro saw price increases at its relative plateaus in 2009, 2010, and 2012.

....The 2012 brought the starting dual configuration up $300 more,....

Three points.

1. That wasn't really a successful pricing direction (over a longer span than just back to 2009). Mac Pro investment stalled if not completely stopped. If that pricing had resulted in "high enough to be interesting" growth Apple never would have done that. So it is a bit perplexing as to why Apple would 'double down' on a strategy that hasn't produced growth.

2. Part of that was just OCD pricing adjustments. The Mac Pro line being linearily increasing across both single and dual package line ups. So lowest dual is higher than highest single. First, if they nuke the dual that OCD compulsion disspears so it wouldn't be present anymore. Second, if Apple punted on E5 1600 that again would eliminate the compulsion (e.g., E3 in $1999-2,599 range means the dual E5 could start at 2,999 )


3. Pricing hasn't moved that much

Year Entry / Mid single
2009 __ $2,499 / $2,999
2010 __ $2,499 / $2,999
2012 __ $2,499 / $2,999


Couple Apple's "must be increasing order" OCD to Intel's pricing drift on dual models (because AMD is comotose and not a treat) and you get the drift on duals. That said, it is highly unlikely the dual Mac Pros are the essential core of the Mac Pro user base.

I think video ram has hindered its mainstream adoption to a degree.

No. Previous, PCI-e ( and previously AGP and PCI) bandwidth speeds have hindered mainstream adoption. PCI-e v3 isn't as restricted and lays ground work for more global memory addressing.

A combination of higher RAM allocations ( > 2GB ) and PCI-e v3.0 make a significant difference.

The bigger root cause blocker at this point is software that does not leverage the resources present more so than the resources not being there. Especially if talking about what hardware is going to be deployed over next 1-3 years. ( the basic Mac Pro design should be oriented to where things are going; not where they have been ).

Beyond that I suspect Apple will market solutions like the 680MX as good enough to fill that gap.

It is 'good enough' for most, but mid-range desktop GPUs are in a different class and not that hard to provision for with a single PCI-e slot. Apple historically hasn't provisioned for > 230W cards but 200W is still enough to put a huge gap between what a 680MX (and its ~100W TDP) is limited to.

 

[GermanyChris]

And? A Xeon E5 (v1 or v2 ) would require a new motherboard and chipset. An really upgraded Mac Pro needs a new motherboard and chipset no matter what. Intel is actually withdrawing from retail market the 3500/3600 CPUs in September.

If Apple is drinking Thunderbolt kool-aid by the gallon an E3 makes it easy to deploy a solution that implements it. Sure capped at 32GB and about one (maybe two ) PCI-e slots but it can cover the $2,000-3,000 zone Mac Pro users and a subset of folks grumbling about xMac. They would need to add something significant like dual 10GbE ports to help support a Mac Pro like price point but that is a bit overdue also ( 1GbE has been standard for around a 10 years and frankly wireless is about to catch up to wired 1GbE. If it is going to stick around it needs to be faster. )

Apple could release both E3 and E5 solutions. (E3 at WWDC causing mass confusion and then in the Fall a E5 bump when that upgrade was complete on components ) But it would not be surprising if they just cut off the higher priced and likely slower growing subset of the Mac Pro market. If they cut off the evolutionary paty the E5 I think they should come up with new name. It isn't really the same product at that point.

What I meant was they need to keep 2011 MP and an 1150 MP..

If there will be TB in the MP there will likely be an 1150 MP

Apple cannot hang on to 1366 at this point no matter

 

[deconstruct60]

What I meant was they need to keep 2011 MP and an 1150 MP..

If there will be TB in the MP there will likely be an 1150 MP

Sockets don't have to make much of a difference it Apple keeps the CPU/RAM daughterboard. Most if not all of the socket differences are on the daughterboard. Same commonality that drove the single versus dual previous generations to share the same motherboard.

If there is a slot differences perhaps the motherboards are also different. However, yet again if that difference is highly contained to just number of physical PCI-e slots then still a large overlap in R&D expediture. Design the board for 4 PCI-e slots and then just chop off 3 ( or 2 ) of them. Since electronic design that is basically a select/cut/delete to get the smaller board ( if need a much smaller board ... could be one with just 2-3 missing connectors .... like initial run of iMac 21" missing the custom mSATA connector. )

What you are not taking into account in the increasing integration. There used to be CPU + Northbridge + Southbridge packages. Now there is just "CPU" ( really CPU+Northbrige and , in 1150 case, GPU ) and IOHub (Southbridge). Both of those could fit on daughter card (or some subset designed area of a common logic board backbone design).

There are some mobile v3 ( Haswell ) offerings that have all three plus the GPU on in one package. that is the trendline that all CPU packages (except some very high end special cases in the purely server area) are moving toward over next 3-6 years.

 

[GermanyChris]

Sockets don't have to make much of a difference it Apple keeps the CPU/RAM daughterboard. Most if not all of the socket differences are on the daughterboard. Same commonality that drove the single versus dual previous generations to share the same motherboard.

If there is a slot differences perhaps the motherboards are also different. However, yet again if that difference is highly contained to just number of physical PCI-e slots then still a large overlap in R&D expediture. Design the board for 4 PCI-e slots and then just chop off 3 ( or 2 ) of them. Since electronic design that is basically a select/cut/delete to get the smaller board ( if need a much smaller board ... could be one with just 2-3 missing connectors .... like initial run of iMac 21" missing the custom mSATA connector. )

What you are not taking into account in the increasing integration. There used to be CPU + Northbridge + Southbridge packages. Now there is just "CPU" ( really CPU+Northbrige and , in 1150 case, GPU ) and IOHub (Southbridge). Both of those could fit on daughter card (or some subset designed area of a common logic board backbone design).

There are some mobile v3 ( Haswell ) offerings that have all three plus the GPU on in one package. that is the trendline that all CPU packages (except some very high end special cases in the purely server area) are moving toward over next 3-6 years.

Provided they could match pins between the two then it would work but I don't think they can..

The flip side is that's exactly what they've done and thats why it's talking so long..

 

[deconstruct60]

Provided they could match pins between the two then it would work but I don't think they can..

They don't have to exactly match. A Superset that covers both would work. Largely what is being exported from the daughterboard is USB, SATA , and PCI-e signals. The 1150 has a DisplayPort output that would be 'dead' on the E5 board but the much bigger (if has more PCI-e slots board can place its own embedded GPU to backfill in the extended board area. if the hocus-pocus about TB requiring some Intel specific CPU magic to work has any merit. I'm not sure it does. just leave the input to Thunderbolt empty and leave TB off of board (empty slot and empty sockets )

The E5 1600 , 2600 , 4600 series all share the same pins out but have functionality because the pin out is a superset. It is not that hard of a problem.

Even more demonstrative by the upcoming Haswell E5 that have the same number of pins but a different electrical assignment (e.g., to support DDR4 )

The flip side is that's exactly what they've done and thats why it's talking so long..

If Apple is moving to adding Xeon E3 to the line up one of the major 'problems' was that E3 v3 slid backwards from Intel's previous roadmaps. It isn't moving out till June. Which happens to be when WWDC is, so that would make some sense timing wise. (throw in updated (speed bumped) GPUs from AMD and Nvidia rolling over next month or so and it is all the more about the right time).


Likewise for Xeon E5. If Apple could have finished the system design by Jan-Feb 2013 it would have made sense to launch with Xeon E5 v1 (Sandy Bridge). That would have saved them the embarrassment of dropping the EU Market because the product design was so far behind the times.

However at this point it makes no sense. They are already exposed in the EU for months. So this late into 2013 make as well wait for v2. All the more so if they can shuffle something else out the door earlier that will satiate a subset of those waiting on the sidelines.


The length of the board design if trying to use a higher intersection of components between the two models wouldn't drive up time all that much. Frankly, not starting until 2012 (or late 2011 ) would be an even bigger casual factor. Apple releasing an update into the second half of 2013 is far more indicative that they were doing anything at all in 2011. That's the primarily factor in the timing.

There was no good reason at all to purposely leave a hole in the EU Market with a 2010-2011 plan that was trying to ship product in 2013. It isn't complexity that is probably the issue here. It is resource allocation.

 

[MyMac1976]

You're my hero.

I must have missed my estimate

 

[GermanyChris]

I must have missed my estimate

Maybe a bit

 

[MyMac1976]

Maybe a bit

Hey the turnip truck misses you

 

[benjancewicz]

I must have missed my estimate

Not too badly!