nMP cpus

[Jinykim]

Intel has announced they will launch 15 core ivytown processor.
but I heard this is not for nMP......

Does anyone reckon it's possible to see new lineup of cpus that may be compatible with current nMP??

 

[sirio76]

Even if they are compatible(I think they are not) those 15core are a >5.000$ upgrade and offers only slightly better performance than the actual 12core, are you sure you are interested in upgrading your CPU?

 

[thekev]

There is nothing called ivytown. Second you're thinking of E7s. They seem to share a mechanical socket, but they're an entirely different cpu line and won't be compatible. Even if they were, Apple wouldn't implement them. You are likely to see a higher core count when it goes to v2 in a year or possibly longer. The relevant cpus would be Haswell EP, not Ivy Bridge EX.

 

[Jinykim]

There is nothing called ivytown. Second you're thinking of E7s. They seem to share a mechanical socket, but they're an entirely different cpu line and won't be compatible. Even if they were, Apple wouldn't implement them. You are likely to see a higher core count when it goes to v2 in a year or possibly longer. The relevant cpus would be Haswell EP, not Ivy Bridge EX.

you're absolutely right. it's e7 family but i don't know why some of articles saying "Ivy Town".

I've been waiting for my 8-core nMP but I should've ordered 12-core since it is said 8-core cpu doesn't support lrdimm while 12-core cpu does..
feeling like I need to cancel my order than back to the end of queue..

hope there will be higher core cpu for e5 v2.........

 

[thekev]

I've been waiting for my 8-core nMP but I should've ordered 12-core since it is said 8-core cpu doesn't support lrdimm while 12-core cpu does..
feeling like I need to cancel my order than back to the end of queue..

hope there will be higher core cpu for e5 v2.........

Ahh I didn't know about the memory restrictions. Sorry to hear that. I do expect a core count bump in v2, because intel is rumored to drop move the E5s to a baseline of 6 cores with Haswell EP. Perhaps Ivytown refers to a specific chip?

 

[Jinykim]

Ahh I didn't know about the memory restrictions. Sorry to hear that. I do expect a core count bump in v2, because intel is rumored to drop move the E5s to a baseline of 6 cores with Haswell EP. Perhaps Ivytown refers to a specific chip?

I heard that the socket 2011 for Haswell EP will be different from the current 2011 socket...

By the way, i'm wondering if there is 32gb ram module that is not load reduced..

if yes then I think I should stick with the 8 core...

 

[Gav Mack]

Haswell EP 15 core and Broadwell EP 18 core. The Haswell EP powered 7,1 may possibly be more future proof than the 6,1 with the Broadwell EP being pin compatible.

http://vr-zone.com/articles/broadwe...s-reliability-per-core-performance/66845.html

 

[AidenShaw]

I heard that the socket 2011 for Haswell EP will be different from the current 2011 socket...

By the way, i'm wondering if there is 32gb ram module that is not load reduced..

if yes then I think I should stick with the 8 core...

The memory limitations have been discussed at length at https://forums.macrumors.com/threads/1698707/ .

Short synopsis is that Intel says that 32 GiB RDIMMs are not supported. Whether they would work is a question.

 

[Jinykim]

The memory limitations have been discussed at length at https://forums.macrumors.com/threads/1698707/ .

Short synopsis is that Intel says that 32 GiB RDIMMs are not supported. Whether they would work is a question.

I read through the thread before..
but did they say e5-2600 v2 doesn't support the ram module as well???????

I thought that supported 32 gb LRDIMMs ... maybe I misunderstood...

 

[JQuick]

The memory limitations have been discussed at length at https://forums.macrumors.com/threads/1698707/ .

Short synopsis is that Intel says that 32 GiB RDIMMs are not supported. Whether they would work is a question.

Alden is correct.

Additionally, there is another important consideration which I do not think has been discussed: the fact that producing a 32GB RDIMM was a lousy idea from a cost/benefit perspective.

The market for large memory models is highly weighted to multi-socket boards (e.g. E5-26xx). These tend to have 3 DIMM sockets per channel. Furthermore, Xeons have limit of 8 memory ranks per channel. Using the most cost effective parts over the past few years one could produce 4-rank 64GB DIMMs.

So, it is technically possible to make a 64GB RDIMM with those DRAM chips but does it make much financial sense? Using RDIMMs would limit these systems to 2 DIMMs per channel (8 total ranks). LRDIMMS use buffering instead or registers. Part of the buffering logic hides internal ranks, making them show up as 2 rank instead of 4 to the cpu's memory controller.

So LRDIMMs make it possible to fully populate these servers since 3 LRDIMMs per channel appear as 6 ranks rather than the 12 ranks that they are internally.

The bottom line, though, is that the bulk of the market for the past few years (HP proliant, IBM, etc.) built systems which had an additional constraint (8 ranks) which made LRDIMMs more viable than RDIMMs. So even if someone made them few would buy them.

Maybe some low-volume, boutique producer currently makes or is testing 64GB RDIMMs. Maybe the existence of the nMP will spur such production. Until then, it is not surprising that people keep seeing LRDIMMs everywhere but no RDIMMs.

p.s. while editing this, I also vaguely remember reading an erratum for the E5 v2 series which said something about a bug with supporting 4 rank DIMMs. I think it involved s workaround for handling a memory sleep state. That might play a part in this as well.

 

[Jinykim]

Alden is correct.

Additionally, there is another important consideration which I do not think has been discussed: the fact that producing a 32GB RDIMM was a lousy idea from a cost/benefit perspective.

The market for large memory models is highly weighted to multi-socket boards (e.g. E5-26xx). These tend to have 3 DIMM sockets per channel. Furthermore, Xeons have limit of 8 memory ranks per channel. Using the most cost effective parts over the past few years one could produce 4-rank 64GB DIMMs.

So, it is technically possible to make a 64GB RDIMM with those DRAM chips but does it make much financial sense? Using RDIMMs would limit these systems to 2 DIMMs per channel (8 total ranks). LRDIMMS use buffering instead or registers. Part of the buffering logic hides internal ranks, making them show up as 2 rank instead of 4 to the cpu's memory controller.

So LRDIMMs make it possible to fully populate these servers since 3 LRDIMMs per channel appear as 6 ranks rather than the 12 ranks that they are internally.

Thanks! now I get it why there is no 32GB RDIMM.

and Thank you AidenShaw the spec document you posted on the thread was extremely helpful.

I just cancelled my order for 8-core and placed a new order for 12-core.

hope to hear a great news from owc and other vendors soon.

 

[deconstruct60]

....
The market for large memory models is highly weighted to multi-socket boards (e.g. E5-26xx). These tend to have 3 DIMM sockets per channel.

Not really. Depends upon which server subsegment you are in.
http://www.intel.com/content/www/us/en/motherboards/server-motherboards/compare-server-boards.html

Select a sampling of Intel's dual socket cloud/datacenter motherboard offerings


S2400BB4 E5 2400 12 DIMMs 6 per socket ==> 2 per channel (2400 series only has 3 channels )

S2400LP E5 2400 12 DIMMs ...... => 2 per channel

S2600GL E5 2600 16 DIMMs 8 per socket ==> 2 per channel

S2600GZ E5 2600 24 DIMMs 12 per socket ==> 3 per channel

HNS2600JFQ E5 2600 8 DIMMs 4 per socket ==> 1 per channel

HNS2600JF E5 2600 8 DIMMs 4 per socket ==> 1 per channel

Flip the selector over to dual and HPC server boards.....


no 24 DIMM models at all. And SIX boards with just 8 DIMM slots.


SuperMicro Blades

http://www.supermicro.com/products/SuperBlade/

No 24 DIMMs models.

http://www.supermicro.com/products/SuperBlade/TwinBlade/

No 24 DIMMs models ( not even 16)


The cloud/blade market isn't that small. The Mac Pro isn't out in some highly abnormal zone here on DIMM count.

The bottom line, though, is that the bulk of the market for the past few years (HP proliant, IBM, etc.)

A very substantial quantity of server boards these days are being bought by data center operators directly. Facebook, Google, Amazon , etc. a large chunk of their servers don't come from the catalog pages of Dell, HP, IBM/Lenovo, etc.

Maybe some low-volume, boutique producer currently makes or is testing 64GB RDIMMs. Maybe the existence of the nMP will spur such production. Until then, it is not surprising that people keep seeing LRDIMMs everywhere but no RDIMMs.

There are lots of blade servers out there (both ad hoc and components for large system vendor blade containers. ). They tend to be about just as limited on DIMMs space as the Mac Pro was and is.

 

[deconstruct60]

Intel has announced they will launch 15 core ivytown processor.
but I heard this is not for nMP......

The E7 aren't going to work in a Mac Pro. It is extremely simple.
In the Xeon naming scheme the second numerical digit is the socket type.

https://communities.intel.com/commu...n-processor-numbering-more-than-just-a-number

Intel might get sloppy but the E7 line up that is leaked are E7 48xx numbers. They are socket '8' not socket '6'. There is nothing in the name that suggest these are compatible at all. Different product line "E7" and different socket number '8'.

Does anyone reckon it's possible to see new lineup of cpus that may be compatible with current nMP??

No. There will be later E5 1600/2600 equivalent class CPUs, but they will have a different chipset. While the CPU is plausibly removable, the core chipset isn't (and that isn't a 'Apple' thing either. Other system vendors same restriction). Meaning stuck with CPUs that are compatible with the C602 chipset. The future ones aren't going to be. The current Mac Pro is using the processor on the tail end of an Intel tick-tock cycle. There isn't another one coming with these socket design parameters.

There is nothing called ivytown.

Actually there is in the E7 zone.

".. it would seem that Intel are preparing a stack of ‘Ivytown’ processors along this design ..."

http://www.anandtech.com/show/7753/intel-readying-15core-xeon-e7-v2

Second you're thinking of E7s. They seem to share a mechanical socket,

They share a pin count; 2011. They are not likely mechanically the same. Neither are the E5 v3 likely to be the same mechanically.

It appears that Intel is driving new upper end Xeon CPU designs by re-assigning pins to different electrical duties rather than changing the pin count.

E7 ( shrink the PCI-e lane count to increase QPI links )
E5 v3 ( tweak DDR3 pins to being DDR4 pins )

Just because the pin count is the same doesn't necessarily mean the socket is mechanically the same. There are notch/groove adjustments that can be made so customer do not mismatch electrically incompatible CPUs with the chipset and/or board routing that is present on the board.

. You are likely to see a higher core count when it goes to v2 in a year or possibly longer. The relevant cpus would be Haswell EP, not Ivy Bridge EX.

We are at E5 v2 now. E5 v3 ( Haswell EP ) is perhaps late 2014 increasing likely early 2015 ( with desktop Broadwell reportedly sliding due to market malaise... server market likely in the same state as 2014 progresses. )