Is the Mac Pro dual processor?

Already been discussed elsewhere. There is a 12 core Xeon CPU that has not been officially announced by Intel

 

It must be a high end CPU that perhaps apple gets exclusively for a while (intel has done this for them before). If it was dual CPU they would be touting 16 cores.

 

LIke the Seb says - The new Mac Pros will use up to 12-core E5 Ivy Bridge Xeons that haven't been released yet. Moreover, Apple has probably been and will be using Sandy Bridge Xeon E5's for testing, until the Ivy Bridge versions are released.

 

Like the Mac Pro itself the processor is a bit of sneak peak.

"... Chipzilla is said to also have a 12-core processor in the pipeline as well. ... "
http://www.engadget.com/2012/10/17/intel-roadmap-reveals-10-core-xeon-e5-2600-v2-cpu/

in the presentation I think they explicitly said "next generation Intel processor" , but they are not trying to be specific because they aren't suppose to be talking about it.

Apple's website is a bit premature not so accurate characterizing it as

" .. starting with the new-generation Intel Xeon E5 ... "

At this point technically it is 'the unreleased generation Intel Xeon E5 ..." Very similar to how the Thunderbolt 2 is also unreleased technology not due till substantively later in the year. It wouldn't be surprising if the GPUs are also unreleased speed bump versions.

It would not be surprising if the 12 core Xeon E5 v2 versions are not part of the initial release of E5 v2.

 

Actually it is quite possible Apple has some early samples from Intel of the Ivy Bridge processors to test with.

Apple won't get the production 12 core processors much earlier than anyone else. Intel just hasn't released it yet.

 

They've already done 10-core versions of the earlier Westmere-based model. All other things being equal, more cores are better, but, they aren't usually equal. I would rather have 4C/8T of 3.7/4.1 GHz, than 12C/24T of 2.4/2.8 GHz. But, the caches on the big chips do make a huge difference. I think the current E5-2687W is about at the sweet spot right now.

 

Most likely. Now lets hope it isn't clocked at 1.8GHz.

 

No with an E5. There are E7 Westmere

http://www.anandtech.com/show/4259/westmereex-intels-flagship-improves

But nothing in E5 series has been at 10 so far. Either in the 1600 and 2600 series. ( these are the only two that are the natural evolutionary path for Mac Pro. The others below a huge mismatches for varying reasons. )


http://www.cpu-world.com/news_2012/2012030701_Intel_rolls_out_Xeon_E5-1600_and_E5-2600_CPUs.html



Or the not going to leveraged 4600 ( or 1400 r 2400 ) series.

http://www.cpu-world.com/news_2012/2012040901_Prices_of_Xeon_E5-4600_microprocessors.html


That is why the jump to 12 is a bit unexpected. The jump from 8-> 10 was for sure (it is the progression Intel has been following for several iterations) . The 12 seems to be "well because we can and perhaps several folks asked for it...." that's why I don't expect the latter to be in the initial release set.


The E7 are on track to jump from 10 to 15 (and had to skip v1 "sandy bridge" to wait on the process shrink. )


http://www.cpu-world.com/news_2013/...upcoming_Intel_Xeon_E5_v2_and_E7_v2_CPUs.html

Can't have everything. There is a fixed amount of memory bandwidth. More cores going faster increases memory access pressure to point of diminishing returns. Twelve is probably stretching it for just 4 memory controllers.

Sweet spot except for price. It costs around $1885.

I have doubts the 12 core model is going to upclocked at that high of a base rate.

----------

No on the E7 part there. Wrong product line. There is somewhat little need for a 2 socket capable processor in a Mac pro and certainly a bit less value add. A 4 way capable one is alot of money for a capabilit to quickly flush down the drain.

but yes on E5 v2.

 

15? I thought it was a law of nature that it had to be an even number. Aren't some functional units shared across two cores?

Couldn't agree more.

I was hoping for an entry-level E3 Mac Pro model myself. The E3's are Intel's best kept secret apparently.

 

It isn't a law of nature. Folks who think in binary like to have power of 2 numbers of elements or at least even numbers. ( 2 , 4 , 8 , 16 .... )

You are thinking of AMD in sharing. The Intel Xeon designs tend to be a layer cake design that looks like..

[ x86 core ] [ L3 Cache block ] [ x86 core ]

where is a ring bus that goes all the way around the inside between the elements in two concentric squarish loops.. The Memory controller. PCI-e switch etc are the top and bottom layers. You just keep stacking these core-cache-core layers to build up numbers ( that's why they jump by 2 ).

The 15 core is may have three causes.

One possible case is that Intel just turns just one of the cores off. That is repetitively common in big iron boxes. IBM will sell a mainframe with just a subset of activated cores/CPUs. If a core fails you just switch on the spare and just keep going. I suspect this is something similar in that have a spare core in there as opposed to goosing the yields with cores that failed testing.

E7 is suppose to have more RAS features (Reliability , Availability, Serviceability) over time as it gradually puts Itanium out to pasture ( or they merge ... http://vr-zone.com/articles/intel-to-merge-xeon-and-itanium-in-2015-2017/17781.html ). A spare core would be a RAS feature.



Second, it also allows them to bump the cache a bit but they come in pairs. It may be want the cache more than both cores.

Third related to that is that there are more QPI links and die space needs. The "un core" ( non x86 core) logic may just needed more space and they ran out of room ( remember this product had to wait to on a shrink to have any possibility of being small enough to ship. )
[that is one reason I think the "x86 core count war" will end in another tick-tock iteration or two. It has become a little to easy to just copy paste in another layer of cores. They are just getting huge now. ]

If Apple had shipped a new Mac Pro at WWDC it probably would have been E3s. But they are still keeping the Mac Pro up in the now classic price zone and continuing to follow the core count wars wars as it transitions over to GPGPUs.

 

In addition to what deconstruct60 said, it is possible that the 16 core dies has a very high chance of one bad core, so they decided to just call them 15 core units and disable whichever core fails testing.

 

This is exactly what I assumed when we first heard about it. Intel just making sure they can deliver with the yields. Though I'm always reminded of Paul Otellini saying "We see a distinct advantage in having all the cores on our die work," in response to AMD's triple core CPU.