You knew someone was going to make this thread eventually.
PCIe 4.0? 5.0?
Faster flash?
T3?
Factory Optane option?
Custom wheels?
PCIe 4.0? 5.0?
Faster flash?
T3?
Factory Optane option?
Custom wheels?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP All Models Waiting for Mac Pro 8,1
- Thread starter Slash-2CPU
- Start date
- Sort by reaction score
I don't think many people will want to play this game for another 8 years. I'm exhausted from waiting forever for the 7,1
well apple can change the cpu / chipset now that way have the an way to route TB video to an video card.
Joking aside, I think the 8,1 will come out as a 2020 model with Cooper Lake Xeon W and will be visually identical to the 7,1. That will be PCIe 4.0.
Think very much like the 4,1 to 5,1 models.
Think very much like the 4,1 to 5,1 models.
Considering the iMac Pro has not received a processor update in two years, don’t expect one next year for the Mac Pro. I believe Apple will update both the iMac Pro and Mac Pro in 2021.
After sufficiently digesting the 7,1 announcement and researching future platforms, I decided to hold off for several reasons:
The only unknown is if the 8,1 will be based on Cooper Lake SP (due early 2020) or skip directly to Ice Lake SP (due late 2020). There's a 6-9 month cadence between each platform, but I'm not sure how quickly Apple will keep up with Intel. (I have a feeling Intel's current schedule is a little ambitious anyway.)
- Similar to autos, I skip the first year model of a redesign.
- No PCIe 4.0 on Cascade Lake SP
- No Wi-Fi 6 on Cascade Lake SP
- Would prefer 3200MT/s memory over 2933MT/s. I know it's arbitrary and we're not in FSB land anymore, but I'm a little OCD about being divisible by a "power of 2"
- The current LGA 3647 on Cascade Lake SP is to be replaced with LGA 4189 in future platforms. Anyone who buys the 7,1 cannot physically upgrade to newer generation CPUs even if the firmware could be upgraded.
The only unknown is if the 8,1 will be based on Cooper Lake SP (due early 2020) or skip directly to Ice Lake SP (due late 2020). There's a 6-9 month cadence between each platform, but I'm not sure how quickly Apple will keep up with Intel. (I have a feeling Intel's current schedule is a little ambitious anyway.)
Last edited:
Intel has yet to release a processor update appropriate for the iMac Pro. The Basin Falls / Cascade Lake-X Ethusiast CPUs (Core i9/i7) have been released and the Xeon versions (W-2000 Series) are expected by the end of this year.
I'm waiting for a "Mac" the missing gab between MacMini and MacPro, with an 5 K display from Apple.
Has not been released . Presently, there is nothing here with a release date in 2019:
https://ark.intel.com/content/www/u...es/123588/intel-core-x-series-processors.html
While the Xeon W 32xx series is in ark.intel.com as released , there are few shipping system with them in them .
Pragmatically, Intel seems to be far more focused in assigning the newer Cascade Lake dies into Xeon SP packages than in the smaller. 2066 socket package. The W-3000 are dribbling out behind those . The most of i9 will I turn trail that .
and the Xeon versions (W-2000 Series) are expected by the end of this year.[/QUOTE]
Some reports are that Intel is abandoning the 2000 series . It isn’t expected by several . I think it probably will but maybe not shipping in volume until early 2020. ( once 2019 EYPC and Threadripper ship, Intel will know if they have ‘extra’ dies they can put in 2066 package or not ) .
The notion is that is that Ryzen 3 and later Treadripper will make the 2000 series non competitive this year so Intel will just ‘walk away’ for this specific package . They would try using W-3000 and maybe a new E-23xx update relatively late in the year . If Intel follows that path , the iMac Pro is stuck until 2020 if they stick with same baseline system design .
Apple,s inability to walk and chew gum at the same time probably means all of 2019 will get consumed with getting the Mac Pro ( and server variant ) out the door and won’t do anything with iMac Pro until next year . IMac Pro could jump to Intels Ice Lake which is probably a better fit to something sized around the 2066 package ( but s new socket so need new logic board ) .
There is decent chance it could . In 2020 Intel is roadmapped to realest two different tracts of Xeon SP class processors . Copper Lake ( which is yet another 14nm +++ revision with some instruction add ons and maybe small clock bump ) . Drooped in same socket package all Apple would need to do is update the firmware and declare a new Mac model number. Wouldn’t be a major change, but Apple could present facade that they are working super hard .
The iMac Pro would shift to the ice Lake baseline on a different socket . Then Apple could shift to new Mac Pro logic boards in 2021 . The Mac Pro is extremely unlikely to shift to a different socket in less than 18 months . Apple isn’t going to toss this motherboard quickly .
Three moves in three years would dig Apple out of the trust hole they have dug over last approximately 8 years .
The iMac Pro got caught in the 10nm shift debacle couple to the even bigger debacle of updating the Mac Pro . If Apple sticks with Intel then in 2020 the iMac Pro should jump out in front of Mac Pro on higher boost and modest core count throughput again . It should also be able to do something substantial on GPU too.
Last edited:
Uhm... yeah. That'd be nice, but taking into account the Apple of the past decade, the next update to Mac Pro will be around 2025 or never. "Okay, here's your Cheesegrater, now it looks more like one. There ya go! Now please shut up and go away for a few years. Once the hysteria reaches another crescendo we'll release another Mac Pro."
If you're hesitant about buying nCMP then wait a few months for firmware updates to everything that's critically wrong (T2 chip is primary concern for me, and being able to boot around it into PCIe RAID0). After that, I strongly suspect there will be a bunch of cards tested/released which will all work with nCMP, and ... absolute silence/nothing, Apple goes back to sleep and forgets about Mac Pro just like they always do. I somehow do not see a renascence of Mac Pro ... I see a small handful of people and a bunch of studios buying them, and if there's enough of those people who buy it, they may update it after 3 years since it's back to being a big tower where they can re-use the same design.
In my humble opinion, thinking there will be an updated Mac Pro in 2020 or 2021 is beyond wishful thinking.
Cooper Lake does not have PCIe 4.0. It is the same basic micro-architecture as the current Cascade Lake and Skylake one. Intel is already announce that all Cooper Lake is getting is primarily some tweaked instructions and some more persistant memory support.
https://en.wikichip.org/wiki/intel/microarchitectures/cooper_lake
Intel baked their PCI-e v4 design into their 10nm logic implementation. ( similar to why LPDDR4 suppose is still missing from their laptop line up. ). It is relatively highly coupled and they aren't likely at all to backport it at this point.
From the roadmaps it looks like Intel won't be able to do the "Extreme Core Count" ( XCC) dies that currently have more than 18 cores. 10nm+ yields won't scale that big so they are going to be stuck on 14nm for "extreme" sized dies. That's why Cooper Lake is there in the lie up. It is just gong to be mainly be there for the big count count option with single die.
Ice Lake may only get up to 10-12 cores and they may have to use two just to cover the span the HCC ( high core count) 12-18 core solutions cover now. Ice Lake is 10nm so it will finally uncork Intel's PCI-e v4 implementation. Intel will likely leave PCI-e v4 only on the Ice Lake product line to pull folks into that new socket (and new set of motherboards ).
Cooper Lake is far more targeted at more so being a socket update for the current deploy base of SP motherboards.
which was about zero change in PCI-e and extremely narrow change in instruction set. Yeah like that ( and no PCI-e version change).
Intel will have another product that will go to PCI-e v4 but it is in a fundalmentally different socket which is exactly not like the 4,1 to 5,1 transition.
Is Cooper Lake really going to be a significant "bang for buck" change in terms of performance? Not really. The two new instructions are mainly to run server AI inference loads (not a core competency of Mac Pro) and perhaps mild teak of the clock speed ( which will be nothing much to hold out for). Will Apple resist the urge to bump the model numbers to show that they are "working hard" on the Mac Pro ? The imagery will probably prompt them to move with little to show ( they can make a performance difference by bumping 1-2 MPX modules. That's where significant performance would come from. But new MPX modules shouldn't generate a new model number. )
Cooper Lake may bring some more saner pricing from Intel if they are both getting their butts kicked by AMD and Intel has a "new shiny" Ice Lake model to up-sell. That would only be a change to $/Performance ratio though. ( which Apple may need also if AMD is royally kicking Intel butt. )
Last edited:
I seem to remember reading somewhere that Cooper Lake will move to the LGA 4189 socket and 8-channel memory lanes, which would rule it out for the 2019 Mac Pro LGA 3647 socket if true, but I agree that the loss would be minimal.
I also have a bad feeling about Ice Lake and a Workstation version. Not sure the clock speeds will be high enough for single threaded performance. So we could be looking beyond/ice Lake for the next Mac Pro refresh.
I wanted to "thumbs up" this comment...
Unless someone (not Tim, not the dearly departed Jony) decides that it is mission critical to re-establish credibility with the creatives by updating the MP7,1 -> MP8,1 -> MP9,1 *whenever* there's something significantly better - even if the bean counters say "we haven't broken even on the R&D costs for the MP?,1".
If there are visionaries at Apple (yes, that's a 1980's sentiment) who want to get back into mind-share with the creatives, they need to hit the bean counters with arguments like "the cost of a new motherboard to update the MP7,1 is less than the profits from one day of iPhone sales".
The armchair CFOs here who wail that "OMG - it will need a new mobo" simply don't understand how cheap it is to knock off a new mobo.
[doublepost=1563572545][/doublepost]
So - a new mobo. Shouldn't be a big deal.
That is the SP version. As long as the die fits inside of the 3647 socket package they can just snip off the external connection to the "extra' 2 memory channels exactly the same way the package 2066 only has 4 instead of 6.
I'm not sure if the bigger socket is also a way of also getting two dies in there also. the Platinum 9282 are BGA-5903. I'm not sure how much that is for the additional pins for that the die pins outs force that large (and there is empty more space than they need).
But the core issue is just how big the die gets. Cranking up the pin outs 542 doesn't necessarily means the die grow a lot if that is just one more memory controller subsystem. ( as oppose more L3 for addition cores). I think the max core count is staying the same (or extremely small increase). As long as the die doesn't grow much can still stuff it into a 3647 package.
Long term the Xeon W pin outs don't have to be the same as the SP. In fact, Intel probably don't want them to be the same. W series didn't start off that way. Going to 3646 in 2019 is at least as much about Intel 10nm FUBAR as anything necessitated.
At one point in the series of "revised , revised , revised" timelines the Cooper Lake was moving to the Ice Lake socket as a "bridge" to Ice lake sliding into 2020 ( back when Cooper Lake was suppose to arrive late 2019 .... which it isn't at this point. ). W-series was far more vague about how they were moving the chipset and socket forward but wouldn't necessary needed to move all the way up to 3647 ( something around 2600-3000 would have worked fine).
The clock rate should be at least the same. The IPC is better so it should have better throughput on single threaded.
The MacBook Air has a Y-series that max Turbo's at 3.6GHz. Intel and OEMS have 'leaked' a bit about the 'not quite shipping yet' Y-series 10th generation and they Turbo out at about 3.4-3.9 GHz. Chart here in this article: https://www.anandtech.com/show/14436/intel-10th-gen-10nm-ice-lake-cpus
Those are probably 10nm+ like the Ice Lake SP class die will be ( or the Ice Lake will be 10nm++ ... which would even more so indicate that clock won't drop).
It is the die sizes, not the clock rate that is the main 10nm problem. Intel isn't going to be a very substantive clock boost but that is OK because the microarchitecutre will have some upticks of its own so get speed bump that way. (and internal and external bandwidths go up ).
Intel is just going to have more problems chasing high core counts. That is not necessarily what the Mac Pro needs at this point.
The iMac Pro could just a medium sized die with some clock increase with no TDP increase. Whether Intel will put that in a W-series in 2020 early enough in the year will have impact on iMac Pro.
Last edited:
Great input, thanks.
I really hope that there is a chance that we can upgrade the 2019 Mac Pro with future CPUs, especially if Intel makes a Ice Lake-W or later CPU for the 3647 socket with the Sunny Cove speed improvements.
This is something that would be great to get disclosed from Apple before the 2019 Mac Pro ships, although I’m not sure Intel want to tell about their W roadmap.
I’ll be surprised if even Cooper Lake chips work in MP7,1, based on MP4,1 5,1 and 6,1 history. You can totally forget Ice Lake or Sunny Cove.
It may be that MP8,1 is like 4,1->5,1, where some EFI mod/update is needed to boot with Cooper Lake. Hopefully(doubtful) this will be an Apple-supplied update. More likely, it will be an end user with the technical know-how who does it.
Current rumor is that Cooper Lake will be MCP design. Multiple Chip Package. Two or three chips under one heat spreader. That may make Cooper Lake completely incompatible with 7,1 if it requires different voltages or pinouts.
Last edited:
I wouldn't count on Ice Lake W heading for the same socket. Even stuffing Cooper lake into a 3647 is in the maybe status ( if the Intel is feeling yield heat on moderate-big dies really bad then may have to, but I don't think they really want to. ). if Look at the "Platform" cadance on their roadmap
https://fuse.wikichip.org/news/2336...r5-pcie-5-0-for-2021-granite-rapids-for-2022/
Intel is on track to churn through more sockets over the next two years, not fewer. The 10nm logjam will cause the socket that was aimed at 10nm to disappear in a year. At Purley now , Cedar Island was suppose to be a bridge to Whitley and then they'll both be dump at Eagle Stream ( Intel back on track on 7nm). Once get back to a non overly complicated process they'll go back to the two year cycle on server sockets. But 2020 is going going to b e abnormal. That abnormal may bring an illusion of a bump for the Mac Pro 2019 but primarily Apple picked a dead-ender to start the "new" Mac Pro with. ( If they had gotten off their butts earlier then they'd be in a different situation but they didn't. )
[ if snipping off the extra memory controller and some other non used pins could allow the Cooper Lake die to be matched back up with Purley; so maybe viable even though targets at a different support chipset/platform/etc. If turn off enough stuff so that look like Cascade Lake then should be able to match up to older chipset. ]
I think the prices of these processors are so high that a significant fraction of the "what is left " Mac Pro market isn't going to care much. They'll be looking for cheaper processors in 2-4 years not new ones ( e.g., what has been going on for the Mac Pro 2010-2012 for last 3-4 years. )
If Apple actually wants to get back onto a path were selling a Mac Pro where there are future CPU upgrades then the next update should be in 2021 ( if sticking with Intel. AMD is suppose to shift sockets in 2020 , so then if jumping off the Intel train. )
The iMac Pro can be put on Ice Lake ( sunny cove arch) because 1. it needs the IPC bump worse (more coupled to the mainstream iMac updates ) and 2. few are going to look to that for procssor upgrades so no big harm if that socket disappears in a year.
Apple would get back on track if went 2019 Mac Pro , 2020 iMac Pro , 2021 Mac Pro , 2021 iMac Pro. The Mac Pro jumps in at the beginning of "tick/tock" and iMac Pro jump onto the second half (but this initial move is screwed because the previous years were screwed. )
If Apple is still treating the Mac Pro as a 'hobby' product then probably three years to next Mac Pro and not going to be trying to line up with the Intel (or AMD ) cycles at all.
Once Intel moves to PCI-e v4 pragmatically need new motherboards anyway. ( I know AMD talked a game about how it won't matter but right now it does.... It is extra work on a tangent that Intel probably won't see the need to put effort into. AMD may close to loop on that eventually but it was a priority for them either. ) . Even if kept 3647 it would be 3647-A or 3647-B or something like that. ( socket 2011 I think went through a sequence like that).
The socket upgrade path for Cooper Lake is going to Ice Lake. Yeah, it looks like they are being sold at the same time during 2020 but Cooper Lake numbers actually shipped, bought and deployed in 2020 are highly likely to much bigger than the Ice Lake ones ( at least in the Xeon SP space ). The "upgrade to Ice Lake" buys will come later for a sizable subset of folks.
I think the Xeon W is probably on a different track. They put the W-2000 on Ice Lake and the W-3000 series will go comatose until get back to 2021 to clean it up. More saner pricing on Single only SP units basically is all the W-3000 series is anyway. (Intel introduced a few of those with Cascade lake SP and could just do more on Cooper Lake if need something for that 'hole' in the line up.). It is is more facade have a workstation solution than what they initially started out trying to do.
They my stuff a Cooper Lake into a 3647 W-3000 to make it look a bit less comatose, but that is just another band-aid layered on top of another band-aid.
Apple isn't in the discrete CPU spare parts selling business so I highly doubt there are going to trumpet this open even if it existed. Nor do I expect they even spent much time aligning the Mac Pro around this. The drought on the Mac Pro had been so long that they just needed to ship something. That a Mac Pro exists at all is the major selling point here.
Apple would have to be blind to not know this is a 'dead ender" CPU socket. They jumped on a 'dead ender' for the Mac Pro 2013 also. This wouldn't be the first time.
if they sell "enough" over first 12 months then perhaps they'll do the "right thing' in 2021. If they don't I'm sure they'll go right back into Rip van Winkle mode.
Thanks. I don’t think Cooper Lake would be worth upgrading to anyway, but the new Sunny Cove architecture could be a nice step up. I guess we will just have to cross that bridge when we get to it, but it makes the 2019 Mac Pro a slightly tricky proposition.
I wouldn't count on Ice Lake W heading for the same socket. Even stuffing Cooper lake into a 3647 is in the maybe status ( if the Intel is feeling yield heat on moderate-big dies really bad then may have to, but I don't think they really want to. ). if Look at the "Platform" cadance on their roadmap
https://fuse.wikichip.org/news/2336...r5-pcie-5-0-for-2021-granite-rapids-for-2022/
Intel is on track to churn through more sockets over the next two years, not fewer. The 10nm logjam will cause the socket that was aimed at 10nm to disappear in a year. At Purley now , Cedar Island was suppose to be a bridge to Whitley and then they'll both be dump at Eagle Stream ( Intel back on track on 7nm). Once get back to a non overly complicated process they'll go back to the two year cycle on server sockets. But 2020 is going going to b e abnormal. That abnormal may bring an illusion of a bump for the Mac Pro 2019 but primarily Apple picked a dead-ender to start the "new" Mac Pro with. ( If they had gotten off their butts earlier then they'd be in a different situation but they didn't. )
[ if snipping off the extra memory controller and some other non used pins could allow the Cooper Lake die to be matched back up with Purley; so maybe viable even though targets at a different support chipset/platform/etc. If turn off enough stuff so that look like Cascade Lake then should be able to match up to older chipset. ]
I think the prices of these processors are so high that a significant fraction of the "what is left " Mac Pro market isn't going to care much. They'll be looking for cheaper processors in 2-4 years not new ones ( e.g., what has been going on for the Mac Pro 2010-2012 for last 3-4 years. )
If Apple actually wants to get back onto a path were selling a Mac Pro where there are future CPU upgrades then the next update should be in 2021 ( if sticking with Intel. AMD is suppose to shift sockets in 2020 , so then if jumping off the Intel train. )
The iMac Pro can be put on Ice Lake ( sunny cove arch) because 1. it needs the IPC bump worse (more coupled to the mainstream iMac updates ) and 2. few are going to look to that for procssor upgrades so no big harm if that socket disappears in a year.
Apple would get back on track if went 2019 Mac Pro , 2020 iMac Pro , 2021 Mac Pro , 2021 iMac Pro. The Mac Pro jumps in at the beginning of "tick/tock" and iMac Pro jump onto the second half (but this initial move is screwed because the previous years were screwed. )
If Apple is still treating the Mac Pro as a 'hobby' product then probably three years to next Mac Pro and not going to be trying to line up with the Intel (or AMD ) cycles at all.
Once Intel moves to PCI-e v4 pragmatically need new motherboards anyway. ( I know AMD talked a game about how it won't matter but right now it does.... It is extra work on a tangent that Intel probably won't see the need to put effort into. AMD may close to loop on that eventually but it was a priority for them either. ) . Even if kept 3647 it would be 3647-A or 3647-B or something like that. ( socket 2011 I think went through a sequence like that).
The socket upgrade path for Cooper Lake is going to Ice Lake. Yeah, it looks like they are being sold at the same time during 2020 but Cooper Lake numbers actually shipped, bought and deployed in 2020 are highly likely to much bigger than the Ice Lake ones ( at least in the Xeon SP space ). The "upgrade to Ice Lake" buys will come later for a sizable subset of folks.
I think the Xeon W is probably on a different track. They put the W-2000 on Ice Lake and the W-3000 series will go comatose until get back to 2021 to clean it up. More saner pricing on Single only SP units basically is all the W-3000 series is anyway. (Intel introduced a few of those with Cascade lake SP and could just do more on Cooper Lake if need something for that 'hole' in the line up.). It is is more facade have a workstation solution than what they initially started out trying to do.
They my stuff a Cooper Lake into a 3647 W-3000 to make it look a bit less comatose, but that is just another band-aid layered on top of another band-aid.
Apple isn't in the discrete CPU spare parts selling business so I highly doubt there are going to trumpet this open even if it existed. Nor do I expect they even spent much time aligning the Mac Pro around this. The drought on the Mac Pro had been so long that they just needed to ship something. That a Mac Pro exists at all is the major selling point here.
Apple would have to be blind to not know this is a 'dead ender" CPU socket. They jumped on a 'dead ender' for the Mac Pro 2013 also. This wouldn't be the first time.
if they sell "enough" over first 12 months then perhaps they'll do the "right thing' in 2021. If they don't I'm sure they'll go right back into Rip van Winkle mode.
God knows when we will see a Ice Lake based W chip, that could slip into 2021. Anyway, I would be very surprised if the Mac Pro will see an upgrade in 2020. I think that the Eagle Stream Platform will be the next step.