nMP CPU Upgrade

[DaveP]

I've seen a lot of speculation (and seemingly some solid information) regarding CPU upgrades on the nMP and that is user replaceable. However, something that is beyond my level of knowledge is what compatible chips/sockets are in the pipeline from Intel.

I would love to get a nMP, but am more CPU, RAM, and disk speed bound so I'm leaning towards the top end iMac with an SSD. However, if I had some confidence that in a couple years I could pop in a faster CPU into a nMP that would likely sway my decision. Simply putting in a CPU with more cores is not of much interest to me and would be looking for more single thread speed. I have a hard time following the Intel roadmap.

Thanks.

 

[d-m-a-x]

Personally i believe it can be done, but the fact remains to be seen

 

[deconstruct60]

However, if I had some confidence that in a couple years I could pop in a faster CPU into a nMP

A simple "pop in" upgrade? There wasn't much to support that and newer info is suggestive of plain not. It is will likely be a "project" to get to and replace the CPU.

. Simply putting in a CPU with more cores is not of much interest to me and would be looking for more single thread speed.

There is not much else to buy other than what Apple is selling now. Doubtful it is gong to get substantively different in a couple of years for a couple of reasons.

a. this current variation of socket 2011 is at a dead end. E5 v1 and E5 v2 and stop. Intel isn't going to put much, if any effort into more versions for this socket. The next variation on the socket coming with E5 v3 won't work in this motherboard. Physically similar/same but electrically different.


b. If cores don't make any difference than not much particularly faster. E5 1660 v2 turbos to 4GHz but the entry level Mac Pro's E5 1620 v2 turbos to 3.9GHz. Highly doubtful that 100MHz is really going to be worth the drama of disassembling the Mac Pro to get to the CPU socket. Folks primarily interest in single core drag racing..... that Mac Pro is already shipping.

I have a hard time following the Intel roadmap.

If primarily just interested in single threaded drag racing then the Mac Mini (and the Intel roadmap associated with it ) are where should be looking.

Intel's roadmap for the Xeon E5 line up is far more focused on "more cores" than on max clock drag racing.

 

[Quash]

I think it will be doable.

Were gonna have to wait for the ifixit teardown to see how hard it would be
And then still apple could probably disable it, but they haven't done so in the past.

It's not very interesting though at the moment. Maybe in 4 years orso it's interesting to replace your quad for a heavily discounted octo core...

 

[DaveP]

a. this current variation of socket 2011 is at a dead end. E5 v1 and E5 v2 and stop. Intel isn't going to put much, if any effort into more versions for this socket. The next variation on the socket coming with E5 v3 won't work in this motherboard. Physically similar/same but electrically different.

That is the info I was looking for. Thanks so much for all the responses. And yes perhaps "pop in" should be replaced with "technically feasible to replace" though we should know for sure shortly.

 

[d-m-a-x]

I'm pretty happy with the 4 core stock model. I mostly do photo editing, so graphics cards are not a huge concern. I am haphazardly planning on finding an 8 or 12 core to swap with the 4 down the line.

 

[AidenShaw]

there's also the BIOS to consider

It's common for a BIOS to have a list of supported CPU ID codes. (This is good coding - don't initialize a CPU if you don't recognize it.)

Even if a CPU fits the socket, it might not work unless it's the same ID as one of Apple's supported CPUs.

 

[flat five]

a. this current variation of socket 2011 is at a dead end. E5 v1 and E5 v2 and stop. Intel isn't going to put much, if any effort into more versions for this socket. The next variation on the socket coming with E5 v3 won't work in this motherboard. Physically similar/same but electrically different.

swap the board too

 

[DaveP]

It's common for a BIOS to have a list of supported CPU ID codes. (This is good coding - don't initialize a CPU if you don't recognize it.)

Even if a CPU fits the socket, it might not work unless it's the same ID as one of Apple's supported CPUs.

That is another good point.

I don't think I can justify the nMP over an iMac.

 

[jasonvp]

It's common for a BIOS to have a list of supported CPU ID codes. (This is good coding - don't initialize a CPU if you don't recognize it.)

I don't think that's the case with Macs. Families of chips, perhaps. But not individual CPU IDs. See the bunch of us that upgraded our 4,1 and 5,1 Mac Pros with the 3.46GHz x5690s as one example.

The concern will be: how easy is it to pull the PCB that the CPU is on away from the thermal core, thereby revealing the top of the CPU as well as the standard 2011 slot lock-down levers, etc? I don't know. We need to wait for a tear-down before we know. My guess is that if you're adventurous, it'll be little more difficult than the 5,1s are.

 

[mrsavage1]

just would like you to know the next gen of e5 cpus are using different sockets so the options here on the nmp will likely be the highest it will go

 

[d-m-a-x]

Wait a couple years and find a 8 or 12 core to come down in price and become available in used condition

 

[deconstruct60]

Wait a couple years and find a 8 or 12 core to come down in price and become available in used condition

That isn't going to buy anything in single core drag racing. In a couple of years, there will be E5 v3 8 and 12 core models that are much more sanely priced and have more features. Even then, it is somewhat questionable. A used Mac Pro 2015 ( in 2016).

 

[avemestr]

As far as I recall:

There's a going to be a physical change in the socket of Haswell Xeons: The new socket (LGA 2011-3) looks the same, but are wired differently, so you will be denied the option of upgrading to a better CPU introduced i.e. next year or two years from now.

You will probably be able to upgrade to other LGA 2011 socket CPUs (i.e. the other ones in the current nMP line-up) though, but you'll be stuck in Ivy Bridge Country, when tech moves to Haswell and later on Broadwell.


It is quite strange, that the nMP is introduced with an "old architecture", though Apple had no choice at this point in time. But they could have updated the oMP to Sandy Bridge/Ivy Bridge architecture and let the old chassis survive another generation - then made an 2-month exclusive deal with Intel on the Haswell Xeons for the nMP. That would have sent shockwaves through the industry.

Introducing the nMP with Ivy Bridge is, just as everything with the nMP, a strange and totally unnecessary compromise.

 

[flat five]

As far as I recall:

There's a going to be a physical change in the socket of Haswell Xeons: The new socket (LGA 2011-3) looks the same, but are wired differently, so you will be denied the option of upgrading to a better CPU introduced i.e. next year or two years from now.

You will probably be able to upgrade to other LGA 2011 socket CPUs (i.e. the other ones in the current nMP line-up) though, but you'll be stuck in Ivy Bridge Country, when tech moves to Haswell and later on Broadwell.


It is quite strange, that the nMP is introduced with an "old architecture", though Apple had no choice at this point in time. But they could have updated the oMP to Sandy Bridge/Ivy Bridge architecture and let the old chassis survive another generation - then made an 2-month exclusive deal with Intel on the Haswell Xeons for the nMP. That would have sent shockwaves through the industry.

Introducing the nMP with Ivy Bridge is, just as everything with the nMP, a strange and totally unnecessary compromise.

what do you mean? nothing in the above paragraphs describe the 'uncompromising' solution..

seems like it was pretty necessary for them to introduce the nmp with ivy bridge.

 

[AidenShaw]

Motherboards are cheap to design....

As far as I recall:

There's a going to be a physical change in the socket of Haswell Xeons: The new socket (LGA 2011-3) looks the same, but are wired differently, so you will be denied the option of upgrading to a better CPU introduced i.e. next year or two years from now.

You will probably be able to upgrade to other LGA 2011 socket CPUs (i.e. the other ones in the current nMP line-up) though, but you'll be stuck in Ivy Bridge Country, when tech moves to Haswell and later on Broadwell.


It is quite strange, that the nMP is introduced with an "old architecture", though Apple had no choice at this point in time. But they could have updated the oMP to Sandy Bridge/Ivy Bridge architecture and let the old chassis survive another generation - then made an 2-month exclusive deal with Intel on the Haswell Xeons for the nMP. That would have sent shockwaves through the industry.

Introducing the nMP with Ivy Bridge is, just as everything with the nMP, a strange and totally unnecessary compromise.

So, Apple will just tweak the motherboard with the new socket and support chipset when Haswell is ready. Slap it in the tube, test and sell.

Look at all the different minor variations on motherboards that Asus/MSI etc. sell -- clearly it can't cost a lot of money to design one.

 

[avemestr]

what do you mean? nothing in the above paragraphs describe the 'uncompromising' solution..

seems like it was pretty necessary for them to introduce the nmp with ivy bridge.

They could have updated the oMP with Sandy/Ivy Bridge and made a big splash with nMP (exclusive Haswell Xeon chips!).

nMP as it is today, is "big news, except for 'old' arch", "great with 2 GPUs, except no SW utilize it", "filled with TB2 ports, except there's hardly any peripherals", etc.

So, Apple will just tweak the motherboard with the new socket and support chipset when Haswell is ready. Slap it in the tube, test and sell.

Look at all the different minor variations on motherboards that Asus/MSI etc. sell -- clearly it can't cost a lot of money to design one.

It's no big problem for Apple to move to Haswell. I simply responded to the OP that asked if he could upgrade the CPU in the future, and for him it'll probably be impossible to upgrade to newer CPUs. He might only be able to upgrade to the CPUs already on the market.

 

[flat five]

They could have updated the oMP with Sandy/Ivy Bridge and made a big splash with nMP (exclusive Haswell Xeon chips!).

i admit, i don't follow this stuff too closely but hasn't one of the more probable speculations towards the nmp holdups been with the cpus? and aren't the current ivy bridge processors fairly exclusive right now?

dunno, if apple waited til haswell, that seems like it would of been a much bigger compromise.

[EDIT] this is not to say that it may very well be a great decision for an individual buyer to wait for haswell and get a 7,1(or whatever).. it's just for the lineup as a whole, i don't think it would of been wise for apple to release this thing much later than they already have.

 

[AidenShaw]

i admit, i don't follow this stuff too closely but hasn't one of the more probable speculations towards the nmp holdups been with the cpus? and aren't the current ivy bridge processors fairly exclusive right now?

"Limited supply" perhaps, but not exclusive. The x64 server vendors have been shipping systems with the E5-v2 for a month or more.

However, those same vendors have just a few E5-v2 workstation SKUs available, and few or no BTO systems.

On CPU availability alone, Apple could have started shipping much earlier in December. It could be a simple issue with wanting to do more testing with final production CPUs that added a couple of weeks.

 

[deconstruct60]

i admit, i don't follow this stuff too closely but hasn't one of the more probable speculations towards the nmp holdups been with the cpus? and aren't the current ivy bridge processors fairly exclusive right now?

dunno, if apple waited til haswell, that seems like it would of been a much bigger compromise.

Bigger than withdrawing the product from the EU market for almost 11 months?

It would not be all that much of a compromise if had a viable product on the market at the same time as the infrastructure Xeon E5 and 602 chipset ( and maybe a single TBv1 controller as conceptual transition to this new concept. Take away one-two slots for TB and an embedded GPU).

If Apple had finished off a E5 v1 design in 2011 they could have still spent a heavy fraction of 2012 and most of 2013 on the Haswell version.

Speculations were that E5 v2 (Ivy Bridge) was going to buy them something "big" on TDP. It doesn't. About the only thing v2 buys is single CPU package 12 core options. So that the new single package 12 core can beat the old dual package 12 cores on several benchmarks. Therefore, Apple isn't backsliding by dropping dual packages.

A single E5 v1 would have topped out at 8 cores if had restricted the systems to just one package. The old 12 cores would have been faster. Similarly if had deployed 16 cores systems ( dual 8's ) then the Haswell system likely would be a backslide ( probably max out at 14-15 cores ). Given there never were any 16 core Mac Pros waiting till 2016-2017 to get there means the progression to the new design constraints won't have a hiccup.

There was more likely a conjunction that Apple was waiting for. Both Thunderbolt and graphics are at major transition points also in 2013-2014. Thunderbolt v2 is one. That dragged the Mac Pro to the end of 2013 since TB v2 was/is primarily targeted at 2014. If Apple had done a TB v1 board with C602 for E5 v1 they would have been tracked into using the same board for E5 v2. The Mac Pro would have rolled through most (if not all waiting on a delayed v3 (Haswell ) ) 2014 on TB v1 when the rest of the line up had gone v2.


Once getting into the custom, much larger GPU card business they probably added time to get a "known" GPU into the system and running right. With those two coupled to probably having done alot of not very much in 2010-2011, they probably missed having anything viable for early 2013 so at that point were aligned with E5 v2 timing. If late might as well wait for the late E5 v2 to surface also.

[EDIT] this is not to say that it may very well be a great decision for an individual buyer to wait for haswell and get a 7,1(or whatever).. it's just for the lineup as a whole, i don't think it would of been wise for apple to release this thing much later than they already have.

The flaw is here is mixing transitioning to this new set of design constraints and going to Xeon E5 (v1 or v2). They didn't have to be coupled. In very late 2014-early 2015 going to TBv2 and some better graphics would have been just as viable. I think the notion that folks are getting at is that if had not done a major switch of design constraints Apple could have gotten something "new" out the door in 2012. Going even a day into 2013 would have been avoided and would have released much sooner.

 

[flat five]

Bigger than withdrawing the product from the EU market for almost 11 months?
[...]
If Apple had finished off a E5 v1 design in 2011 they could have still spent a heavy fraction of 2012 and most of 2013 on the Haswell version.

yeah, you're probably right.. they probably could have done that..
but we see how it played out and we can see they were simply over it.. 5,1 was the farewell ride i think.. and that's what they did-- just rode it out..


bad for the company too if they didn't just drop it.. the designers who would of caught the assignment of "you guys go work on the 5,1 but meanwhile, we'll be messing around with this thing"...well, the 5-1 people would of quit.

of course there's a lot of wouldshouldacoulda talk but i'm pretty sure there's never a perfect solution.. (if they did X then Y which was handled via scenario B now becomes Z problem.. etc)

The flaw is here is mixing transitioning to this new set of design constraints and going to Xeon E5 (v1 or v2). They didn't have to be coupled. In very late 2014-early 2015 going to TBv2 and some better graphics would have been just as viable. I think the notion that folks are getting at is that if had not done a major switch of design constraints Apple could have gotten something "new" out the door in 2012. Going even a day into 2013 would have been avoided and would have released much sooner.

yeah, i see what you're getting at.. and it makes sense as an immediate thought.
using the mp1 history, they quick fired intel into the lineup and people had to wait awhile to get native apps running.. in a similar light as the dual gpus..
but they had to do it in order to make the 3,1+ as good as they were.

by the time version3+ of this comes around, it's going to be well tuned.. will some of the decisions they're making now make more sense then? the true design might not be realized for another 4 years or so (dual 4GB ssd, 18core cpus, tbolt4, 256ram etc).. the rest of us buyers are just paying lots of cash to be beta testers


-----------------------

[edit] but once this day occurred:
https://forums.macrumors.com/threads/1677182/

the 5,1 was a done deal..

(and even if you don't believe the story of 2010-germany, that day did occur somehow someway)

 

[AidenShaw]

I don't think that's the case with Macs. Families of chips, perhaps. But not individual CPU IDs. See the bunch of us that upgraded our 4,1 and 5,1 Mac Pros with the 3.46GHz x5690s as one example.

True, but it's still a risk.

Past behavior by Apple BIOS engineers is a hint, but not a guarantee of what future systems will do....

 

[jasonvp]

True, but it's still a risk

I'll bet good money against it being a risk at all. There's no reason to believe that the 2600 or 1600 V2 series will be incompatible depending on the specific chip ID.

 

[AidenShaw]

I'll bet good money against it being a risk at all. There's no reason to believe that the 2600 or 1600 V2 series will be incompatible depending on the specific chip ID.

There's also no reason to be certain that the BIOS will accept unknown CPU IDs....

It goes both ways.

 

[jasonvp]

There's also no reason to be certain that the BIOS will accept unknown CPU IDs....

It goes both ways.

You willing to put some money on it? Friendly little wager?