Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Robert123123

macrumors newbie
Original poster
Jan 3, 2014
11
0
Does anyone know if it is possible to replace the cpu in a nmp by a xeon e7 cpu? I am eyeing the soon to be released 15 cores.
 
This is incorrect. That socket has been discontinued.

it is indeed the LGA 2011 socket

If it will work or not, that I couldn't tell you. But I sure hope so. The processors that are coming in the Mac Pro are going to be outdated in a few months.
 
it is indeed the LGA 2011 socket

If it will work or not, that I couldn't tell you. But I sure hope so. The processors that are coming in the Mac Pro are going to be outdated in a few months.

Sigh... it's a troll thread. I figured it could be ignorance at first. That he claimed LGA 2011 was discontinued suggests it's a troll thread, especially considering the OP registered just to create this. It will probably be wastelanded soon enough.

Well on the off chance it's not a troll thread, they don't seem to run that hot assuming this article is valid.


http://www.cpu-world.com/news_2013/2013100101_Some_details_of_Ivy_Bridge-EX_processors.html

Intel has either used different sockets or chipsets in the past. Even if they don't, I highly doubt this will be a cost effective route, and it seems like a risky test. They're probably more than 6k.
 
Last edited:
Sigh... it's a troll thread. I figured it could be ignorance at first. That he claimed LGA 2011 was discontinued suggests it's a troll thread, especially considering the OP registered just to create this. It will probably be wastelanded soon enough.

Actually the OP's question is perfectly valid. The new E7s coming out are Ivy Bridge based using the LGA2011 socket. LGA1567 was Nehalem based and is indeed discontinued. It is pretty obvious the OP is referring to the upcoming Ivy Bridge E7s as they are the only ones with 15 cores...
 
What's the advantage of the e7 vs the e5

e7 is for quad CPU set ups. It has an extra QPI link to talk to more processors basically. They also support more RAM, but in the MP there really isn't any reason to use one.
 
Sigh... it's a troll thread. I figured it could be ignorance at first. That he claimed LGA 2011 was discontinued suggests it's a troll thread, especially considering the OP registered just to create this. It will probably be wastelanded soon enough.

Well on the off chance it's not a troll thread, they don't seem to run that hot assuming this article is valid.


http://www.cpu-world.com/news_2013/2013100101_Some_details_of_Ivy_Bridge-EX_processors.html

Intel has either used different sockets or chipsets in the past. Even if they don't, I highly doubt this will be a cost effective route, and it seems like a risky test. They're probably more than 6k.

Not sure where these idiotic, negative and above all unhelpful messages, that clearly ignore what is written, come from.
But thanks lixuelai, I indeed referred to the E7 v2, that's why I wrote that in the title. And indeed mentioned that lga 1567 is discontinued, which it is. And finally, all indications (thx 808?), short of the cpu in our hands, point to e7v2 using lga 2011.

I am also aware that their tdp makes then fine for the nmp, the question is whether Apple has a history of tightly controlling which cpu is allowed ( it would be quite easy to refuse to boot if an e7 is detected). On the other hand, I believe apple would want to have room to grow beyond 12cores and maybe 15 is already a planned update for next year - it brings a good perf. Boost at little effort. So I had the same thought as dalupus.

----------

e7 is for quad CPU set ups. It has an extra QPI link to talk to more processors basically. They also support more RAM, but in the MP there really isn't any reason to use one.

You do realize that the xeon e5 currently used in the nmp are also meant for dual cpu setups, yet Apple uses them in single cpu mode? Can you see why they do it? If not, read the below!

----------

What's the advantage of the e7 vs the e5

They will have some CPUs packing up to 15 cores and more cache, that's about the only relevant point for the nmp, since only single cpu is allowed.
 
Last edited:
These are the interesting ones:

Xeon E7-2870 v2 2.3GHz 15 cores
Xeon E7-2880 v2 2.5GHz 15 cores
Xeon E7-8857 v2 3.0GHz 12 cores

The will all be over $4,000, and probably closer to $5,000. I've seen much higher prices suggested for some of the really top ones, the 155W TDP models and even the E7-2880 V2.

eit: So they won't work anyway as they aren't using the same socket, deconstruct60 points out here: https://forums.macrumors.com/posts/18591327/
 
Last edited:
I wonder if anyone wants to take the chance at testing a 155W TDP in a Mac Pro and see if it can sustain the power and heat
 
I wonder if anyone wants to take the chance at testing a 155W TDP in a Mac Pro and see if it can sustain the power and heat

better buy with a CC with a good warranty and return policy (i.e. amex plat)
 
They will probably work. These are the interesting ones:

Xeon E7-2870 v2 2.3GHz 15 cores
Xeon E7-2880 v2 2.5GHz 15 cores
Xeon E7-8857 v2 3.0GHz 12 cores

The challenges will be two-fold:
1. Will the EFI recognize the family of Xeon? Right now, Apple is using E5 1600 and 2600 series chips. Will the E7 2800 and 8800 series be recognized by the hardware?

2. How about the OS? OS X usually has families of processors hard-coded into it. Look at Mavericks 10.9.1 as an example: it can't speed step the same chips Apple is using in the Mac Pros without a patch. Guys building Hacks with those Xeons have had no luck getting them to work 100% yet.

It's very unlikely Apple would ever consider adding processor support for the E7s in OS X, so you're left with either hacking the kernel (not impossible), or just running Windows. But if you're going to run Windows: build a PC instead.

All in all I think this is a wasteful pursuit, personally.
 
Xeon E7-48xx v2 (quad-processor)

All using the LGA 2011 socket.

Actually technically not. Confusingly the Intel has taken to mutating the 2011 socket. The number of pins are the same ( hence the 2011 stays constant) but they are not all electrically compatible.

The E5 v1 and v2 version use the original 2011. E5 v3 will use 2011-3. These E7's use yet another variant of 2011 ( not sure what suffix Intel stuck on those.).

You can easily tell in the chart in the column just to the right of the "2011". There are 3 QPI links in this flavor of 2011. In the E5's flavor there is only a maximum of 2. ( on the E5 1600 processors they are just 'dead' as those procesor variants only enable one of the QPI links, but the physical pins for two are there. )

What will find on the Xeon E7 v2 is that the number of PCI-e lanes is gone down from the E5. Since there are typically 2-4 CPU packages, 80+ PCI-e v3 lanes is plenty for all but super extreme corner cases. What is far more limited in a 4 CPU package set up is intra-system bandwidth ( between CPUs ) rather than to PCI-e cards/controllers that typically connect
externally.

" ... The PCI Express has 32 lanes. ... "
http://www.cpu-world.com/news_2013/2013100101_Some_details_of_Ivy_Bridge-EX_processors.html

Xeon E7 v2 is going to have 32. Xeon E5 v2 has 40. Those 'missing' PCI-e pins are being assigned to QPI duty. Still have a humongous number with 4 CPU packages ( 4 x 32 => 128 or even two package set-ups 2 x 32 => 64 ).

However, you most definately have less than a Mac Pro reuquires from the CPU. That is minimally now 40.

Intel is adapting the 2011 pins to fix different subcontexts demands better. That doesn't mean the whole line up inside a generation or inter-generation is necessarily electrically the same or compatible.
 
What's the advantage of the e7 vs the e5

In the context of a Mac Pro. .... nope , zip , zero.


In the context of a substantially more expensive system and software....

workload consolidation. ( multiple instances of same software on fewer systems. If vendor is charging per system this can pay off long run).

more inter CPU package communication bandwidth.

E7' used in single CPU package contexts are highly dubious. Maybe there will be a E7 1xxx v2, but if there wasn't one almost nobody would notice. I suspect those are really only for corner case blade or telecom embedded servers where for some reason they need some esoteric E7 feature for some odd reason.

In most context for single , dual , and a small fraction of 4 CPU packages E5 makes more sense.
 
Not sure where these idiotic, negative and above all unhelpful messages, that clearly ignore what is written, come from.

sorry but you are a bit out there on this one. This threads content has been covered several times before and Intel's product naming conventions for Xeon is quite clear. A simple Sesame Street came can pick out the commonalities and differences between

E5 1620 v2
E5 2267 v2
E7 2880 v2
E5 1680 v2

The ones that work have "E5" at the beginning. The ones that don't have "E7'. It is really, really, really not that hard.

processor-banner-xeon.jpg

http://www.intel.com/content/www/us/en/processors/processor-numbers.html


Xeon E# #### v#

The number after E is a product line. Why would CPUs from a different Intel product line fit in a Mac Pro?????? Seriously why? What even motivates not asking that question?


The first digit is the max number of CPU packages that can used together. "Wayness". The are E5 4xxxx offerings. Don't need E7 to go 4 way.

Which brings us to he second number the socket. Notice how these "E7 will work" examples folks are pointing to have an 8 instead of a 6 there. Even Intel is telling you that the socket is significantly different just actually reading the number. If consistent E5 v3 will have a different digit than '6' . Or that the v# number is going have higher significance on the socket digit than on first glance.
 
Last edited:
sorry but you are a bit out there on this one. This threads content has been covered several times before and Intel's product naming conventions for Xeon is quite clear. A simple Sesame Street came can pick out the commonalities and differences between

E5 1620 v2
E5 2267 v2
E7 2880 v2
E5 1680 v2

The ones that work have "E5" at the beginning. The ones that don't have "E7'. It is really, really, really not that hard.

Image
http://www.intel.com/content/www/us/en/processors/processor-numbers.html


Xeon E# #### v#

The number after E is a product line. Why would CPUs from a different Intel product line fit in a Mac Pro?????? Seriously why? What even motivates not asking that question?
Yet another one:
What motivates this question: 15 cores

Why would a different line fit?

Are you seriously asking? They can name it what the hell they want, it uses the same socket and chipset, that's a stronger indicator than 'but it's e7, not e5, end of story' silly line of reasoning you use. In fact, if you used your brain instead of spewing info taken from the web that you obviously barely understand, you would know that compatibility between lines is good for intel, for motherboard developers, for basically anyone. Especially now that Intel has frequent updates.
But it's also incredibly easy to prevent it by software.

Idiotic refers to the claim that 2011 was discarded or that e7v2 used an old socket. Please don't join that club, and don't assume more than is written or fire before reading.

Finally, xeons e5v2 are supported in hacks now that the nmp is out.

----------

sorry but you are a bit out there on this one. This threads content has been covered several times before and Intel's product naming conventions for Xeon is quite clear. A simple Sesame Street came can pick out the commonalities and differences between

E5 1620 v2
E5 2267 v2
E7 2880 v2
E5 1680 v2

The ones that work have "E5" at the beginning. The ones that don't have "E7'. It is really, really, really not that hard.

Image
http://www.intel.com/content/www/us/en/processors/processor-numbers.html


Xeon E# #### v#

The number after E is a product line. Why would CPUs from a different Intel product line fit in a Mac Pro?????? Seriously why? What even motivates not asking that question?


The first digit is the max number of CPU packages that can used together. "Wayness". The are E5 4xxxx offerings. Don't need E7 to go 4 way.

Which brings us to he second number the socket. Notice how these "E7 will work" examples folks are pointing to have an 8 instead of a 6 there. Even Intel is telling you that the socket is significantly different just actually reading the number. If consistent E5 v3 will have a different digit than '6' . Or that the v# number is going have higher significance on the socket digit than on first glance.

Could you point to where this has been covered? I actually looked.
 
Finally, xeons e5v2 are supported in hacks now that the nmp is out.

Including all of the Speed Stepping and whatnot? Mavericks brought with it some support for the Hacks running E5 V2 chips, but no one could get them working at 100%. The 10.9.1 update didn't help.
 
Not sure where these idiotic, negative and above all unhelpful messages, that clearly ignore what is written, come from.
But thanks lixuelai, I indeed referred to the E7 v2, that's why I wrote that in the title. And indeed mentioned that lga 1567 is discontinued, which it is. And finally, all indications (thx 808?), short of the cpu in our hands, point to e7v2 using lga 2011.

I am also aware that their tdp makes then fine for the nmp, the question is whether Apple has a history of tightly controlling which cpu is allowed ( it would be quite easy to refuse to boot if an e7 is detected). On the other hand, I believe apple would want to have room to grow beyond 12cores and maybe 15 is already a planned update for next year - it brings a good perf. Boost at little effort. So I had the same thought as dalupus.

It's possible that I deserved that response. Anyway I've never known intel to make such things electrically compatible. It seems even less likely from Apple, especially. In the past they've ignored the most expensive cpu options. The 12 core isn't that much of a stretch, as it aligns somewhat with the price of 2 x 6 cores from the last generation. Haswell EP will supposedly bring the base chip from 4 cores to 6. They've been sitting at 4 cores at the bottom level since 2009, so that's significant. I'm not sure how many the top options will have at that time.
 
It's possible that I deserved that response. Anyway I've never known intel to make such things electrically compatible. It seems even less likely from Apple, especially. In the past they've ignored the most expensive cpu options. The 12 core isn't that much of a stretch, as it aligns somewhat with the price of 2 x 6 cores from the last generation. Haswell EP will supposedly bring the base chip from 4 cores to 6. They've been sitting at 4 cores at the bottom level since 2009, so that's significant. I'm not sure how many the top options will have at that time.

Apologies if it sounded harsh, and your graceful answer is appreciated. I really did not want the tone of this thread to go negative; I want to get opinions/insights. As many others, I suspect, I find that the new mp is great (low noise!) but is lacking in terms of cpu power. The price of the 12 cores upgrade in addition make thinking of about a post-buy upgrade economically viable it seems.

I hear you also in terms of Intel making chips incompatible, but I hope that they won't. Technically speaking, ivy was a tock and yet uses the same socket as sandy bridge, I can't imagine a technical reason that would force an incompatibility. Without any competitors using the same socket, I think they would make it incompatible only if big buyers such as dell wanted that, but I again can't imagine why. Eg if you built a supercomputer with 1000 CPUs and want to upgrade, you'd be quite annoyed if there are no options because of marketing gimmicks. So that's why I think limitations, if any, will come from apple. But they could have blocked all CPUs except the ones they use, but there are reports of working 1o cores.... So that was what led me thinking of putting 15 cores in here.... I am very aware that there are many ifs however.
 
Last edited:
Apologies if it sounded harsh, and your graceful answer is appreciated. I really did not want the tone of this thread to go negative; I want to get opinions/insights. As many others, I suspect, I find that the new mp is great (low noise!) but is lacking in terms of cpu power. The price of the 12 cores upgrade in addition make thinking of about a post-buy upgrade economically viable it seems.

I was in a grumpy mood without realizing it at the time, and it was a topic that had come up before. I'm sorry about that.


I hear you also in terms of Intel making chips incompatible, but I hope that they won't. Technically speaking, ivy was a tock and yet uses the same socket as sandy bridge, I can't imagine a technical reason that would force an incompatibility. Without any competitors using the same socket, I think they would make it incompatible only if big buyers such as dell wanted that, but I again can't imagine why.

There are various things such as is compatible with different chipsets. I don't recall E7 and E5 lines sharing the same pin configuration in the past. Workstation and server boards typically carry chipsets for 2 generations, or at least that has been the case with intel in recent history. For examples nehalem--->westmere brought the popular quad--->hex upgrade after the price of hex cpus dropped by 40% to $600 retail. There are other restrictions by intel. E5-1600 cpus can't be used in dual configurations. I'm not sure whether that has to do with the way chips are binned or if it's just about margins.


Eg if you built a supercomputer with 1000 CPUs and want to upgrade, you'd be quite annoyed if there are no options because of marketing gimmicks. So that's why I think limitations, if any, will come from apple. But they could have blocked all CPUs except the ones they use, but there are reports of working 1o cores.... So that was what led me thinking of putting 15 cores in here.... I am very aware that there are many ifs however.

The 15 core cpus are still likely to be clocked lower than the 12 core EP. It's often the case, and I don't know whether or not that would be beneficial to your average workload. Extreme core counts at low clock might be the way to go when dealing with massive numbers of virtual machines on something like ESXi, but I'm not entirely sure, as I've never needed to research options for that kind of thing. For computationally expensive things like render farms, E5-2600 cpus seem to be quite common. You're right about cto markups. Many of those who are concerned about the ability to upgrade are looking for ways to extend the life of their system. If they are purchasing a 6 core today, they may later look for retired 12 cores as the hardware ages.
 
It's very unlikely Apple would ever consider adding processor support for the E7s in OS X

I'm not so sure... this thread is here because people are looking at a list of publicly announced future Intel CPUs and picking out ones they would like to see in their Mac Pro.

Over in Cupertino, there will undoubtedly be people looking at a similar (but more official, longer and much more secret) list of future Intel CPUs and picking out the ones they would like to see in their Mac Pros.

If we've come to the conclusion that those 15 core E7s look pretty cool, who's to say that Cupertino won't come to the same conclusion, solve the QPI issues and put them in 6,2 or 7,1 Mac Pros?
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.