It's also a bit confusing because there's no indication that we'll get Cove Xeons next as opposed to continued Lake iterations.
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 7,1 Waiting for Mac Pro 7,1 A1991 (no more)
- Thread starter Mago
- Start date
-
- Tags
- macpro 2016
- Sort by reaction score
It's also a bit confusing because there's no indication that we'll get Cove Xeons next as opposed to continued Lake iterations.
To be fair to Intel I quite like that architecture. Its wider both: on front end and back end of the execution, so the IPC increase should be quite impressive, like in the range of 20-25%, even.
The other side of this coin is that it appears it will not clock as high as Skylake/ Kaby Lake/Coffee Lake/Coffee Lake/Comet Lake - Refresh will clock. Partially because of the architecture, partially because of the process, itself, which will have lower performance than 14 nm +++ that CFL-R is based on.
well apple should not be locked into the heat / power needs of one cpu gen.
But chipset / ram channels / pci-e lanes do change a bit with Intel each gen. But they should not be locked into any gpu that needs to be on it's own card / Module
I don't think that I understand this comment.
Having the GPU on its own PCIe card is a huge advantage - and not having that is a big part of the failure of the MP6,1.
Or by "its own card" do you mean a GPU on a proprietary card like the MP6,1 - rather than interchangeable PCIe cards like almost all other workstations?
The other side of the coin which neither article goes into is exactly what the hardware changes are to mitigate against Spectre and Meltdown. Current chips are still doing a blend of sw and hw mitigations.
Intel already said, that all of 2018 and later CPUs will have hardware level mitigations.
We'll have Lake Xeons for a while still is the answer.
From this at Anandtech. Its sounding like Cascade Lake is sooner in 2019 then later (reference here) and Cooper should be a 2020 product. So 2021 at the earliest for 10nm to hit the Scalable Family.
Yeah and swap the GPUs from AMD to Nvidia while their at it
Nope, all AMD; CPUs, GPUs, & GPGPUs...!
With Zen 2 - sure. Right now, with Zen/Zen+, you would lose Single Threaded performance. Zen 2 appears to be around Skylake in IPC, and will end with 16 cores on AM4, most likely.
Yes, I should have clarified; I am looking at the 7nm stuff forthcoming from AMD as a solution for an all AMD Apple desktop lineup...
If cascade is set for 2019, and Sunny Cove also, well - what's it gonna be like?
Will mMP come with Cascade or Sunny?
Confusing?!
Sunny seems like a new arch, sort of, at least with some enhancements.
Where does that leave Cascade?
Anyone figured out yet?
Will mMP come with Cascade or Sunny?
Confusing?!
Sunny seems like a new arch, sort of, at least with some enhancements.
Where does that leave Cascade?
Anyone figured out yet?
Sunny Cove will come at first in low-power, low-core chips. There will not be anything serious from Intel until at least 2020. Comet Lake for Mainstream Socket 1151 and Cascade Lake for Server chips are slated for 2019.
Sunny Cove -- a microarchitecture.
*-Lake -- some implementation of some microarchtecture with perhaps different incremental features (intro of an new opcode(s) and/or fixes or small adjustements to cache. ) .
In the Xeon SP class there are two coming in 2019.
Cascade Lake - - sampling now to "big" clould players and more mainstream in 1st half of 2019.
This is a variant of Skylake baseline microarchitecture but with fixes for security (Specter ) which impacts VM host business far more than mainstream and some Optane memory tweaks.
Cooper Lake --- yet another 14nm Skylake tweak. Probably more security tweaks and adding in some "machine learning" math oriented opcodes to the mix. This is "due in 2019" which probably means around Q4 2019. It probably has the last of the clock speed bumps in it also to goose the last performance out of 14nm+++
"... Intel officially put up the signs declaring Cascade Lake for 2018, Cooper Lake for 2019, and Ice Lake for 2020. ..."
https://www.anandtech.com/show/1320...tels-dcg-discussing-cooper-lake-and-smeltdown
Several indications that this will be a socket change and really a placeholder for Ice Lake. It won't just be the CPU that changes but will get an updated PCH also that would have come if the 10nm transition hadn't been such a huge log jam.
Some folks won't necessarily have to throw out Cascade systems for Cooper lake. ( some will upgrade their skylake-SP to Cascade and probably sit out another couple of years. ). Cooper Lake will probably be for folks doing "new" (green field) deployments or major overhaul upgrades ( swapping out racks ).
Ice Lake -- due in 2020 and apparently will use the Sunny Cove microarchitecture in the Workstation-Server part of the market . Probably when 10nm+ or 10nm++ is ready. Some 6-8 cores in the mainstream Core line up might squeak in at the end of 2019, but that's doubtful (given Intel's track record).
I suspect that Intel Xeon W won't follow all three of those. I think that Intel with match the Xeon W will pick up Cascade Lake ( ignoring the current Skylake-X 'refresh' that does exceeding little for the Xeon W product line. ) so W v2. Since that will be 1Half of 2019 then a more than reasonable chance that they'll skip Cooper Lake and just drop new W v3 around middle of 2020 with the Ice Lake / Sunny Cove foundation.
Perhaps Intel will throw a Cooper Lake variant at the Core i9 crowd in Fall 2019 since they got a funky refresh in Fall of 2018. Improvements for random VM hosting security and Optane probably have limited impact for the Desktop hotrodders so Cascade doesn't "do" much for them so skip it. The clock bumps of Cooper may entice them a bit and will 'future proof" the board with the socket change. For the bigger workstation vendors I'm sure their will be single Xeon SP offerings in a larger workstation for those who particularly anxious to get access the new "machine learning" opcode sooner rather than later, but I don't think Apple (and probably Intel) is going to be pressed by that for the Intel W line up. ( those couple of MI opcodes and diminishing return clock bumps aren't really going to help again Zen2. Price adjustements will, but that stuff isn't going to make big difference over that relatively short amount of time in the Xeon W workstation market. Tossing out some Xeon E (derived off the mainstream Core i might help stop the bleed more in early 2012. ) )
Someone at Apple could have foolishly "beat the farm" on a Ice Lake class procesor PCH and socket design and Apple would be more pressed about a new socket change for Intel W before Ice Lake arrives. Hence the next Mac Pro would have slid to end of 2019 and perhaps into 2020. Spectacularly goofy given how late the product already was in 2017, but can't put bozo moves past Apple's scope at this point.
[ If Apple wasn't grossly late, then it would be better to "jump in" to a new Mac Pro major upgrade at the beginning of a socket change cycle. At this point though, it would be better long term for Apple to just "eat" the pain of doing a major board design and have to turn around and do another within 2 years. The giant screw up here was not being active in development and at some point they need to "eat" that. It was a screw up .... pay for it and move on. It isn't like Apple is going to go broke over it. ]
[doublepost=1544890264][/doublepost]
Actually, technically no. Technically, there are some hardware fixes, but they haven't fixed all of the problems. Even with there hardware fixes there are still some OS fixes required in some contexts ( a bit less "painful" performance fixes, but still not cleaned up completely in hardware ). There is a chart in this article:
https://www.anandtech.com/show/1330...e-intel-clarifies-whiskey-lake-and-amber-lake
It is still a mix. Even Cascade Lake for the SP ( and likely Intel Xeon W in 2019) will still be a mix. There are variations of Meltdown that are still OS mitigated and several are mitigated still in firmware. Intel will probably squeeze in a bit more into Cooper Lake for the SP class late in 2019.
Not that Apple would be using Amber-Lake in a Mac Pro, but it got no hardware fixed and arrived later in 2018. It is only in the higher classes typically used in hosting VMs from "random" folks is where they really meant "all" (and frankly haven't shipped to everyone in 2018). [ Intel might tap dance and say Amber Lake was more of name rebadge than something new for Fall 2018 ... but they are trying to have it both new and 'old" depending on context. ]
I doubt going to get substantially deeper penetration into hardware Meltdown fixes until Sunny Cove (Ice Lake) microarcchtecture arrives. Even then not sure they are going to get them all. Development of Sunny Cove probably started before Intel had full scope of Meltdown in 2017 got illuminated. The stuff that is showing up as firmware fixes they could work in but the OS only fix stuff may be too much out of scope for Sunny Cove to completely lock down. Intel has working silicon at this point which means there was finite amount of time to weave in fixes.
Part of what Sunny Cove objective is to push "wider" ( broader scope of instructions down in parallel). That is one of the driving functions of these with trying to do more "into the future" opcode instructions before fully realized the impact/scope. Intel will have to go through the whole pipeline and clean that up.
Last edited:
" ... Sunny Cove microarchitecture and 3D die stacking for processors potentially making their way into future Mac devices, with arrival more likely in Xeon-equipped iMac Pro and Mac Pro lines first. ... "
Holy cow.... that is about as backward as you could get out of the Intel briefing. Intel demo in the briefing was working silicon on the smaller scale; not the largest one.
3D packaging of the CPU , PCH , and some RAM into a chip SoC. ... does that sound like a Mac Pro or iMac Pro ?
Integrated GPU Gen 11 graphics... Do any Mac Pro or iMac Pro have an Intel iGPU in them? No. So this demo was a fast path to the Mac Pro ? Please. you have got to be kidding me. Clickbait? Yes. Insightful analysis? No
( mention Mac Pro and will get hits from the folks circling the airport on fumes. Lament that it is painfully neglected and get even more clicks. But delivering insight ... only to make more money off the articles. )
The other working demo was of a one "big core" , 4 smaller cores and 2mW standby power draw.... does that sound like "big iron" Mac Pro or iMac Pro which stay plugged in 100% of the time to work or have the need for "small" x86 cores for lightweight, mainstream workloads? Again no. Not even slightly applicable.
Discrete GPU coming in 2020-2021 time frame .... Mac Pro really is on a "grossly late" constraint so 2021 parts not particularly applicable.
Will there eventually be a Mac Pro / iMac Pro with Sunny Cove microarchiecture .... 2020-2021 probably yes, but there is no good reason why Apple couldn't drop new version of both Mac Pro and iMac Pro in 2019 with what was a year (or so ) ago roadmapped to be late 2018 tech. There should be a Cascade Lake derivative of the Intel W in first half of 2019. It is socket compatible with iMac Pro so that's a simple "bump" upgrade. The Mac Pro has been in Rip van Winkle mode for a ridiculous amount of time. "2018" tech will be an upgrade over "2013" tech easily. Apple doesn't need some Area 51 , Buck Rogers tech ... they just need current stuff at this point with the Mac Po. In 2017 shooting for "late 2018" tech would have been an entirely sensible move to make. That happened to slide into 2019
Apple may do something strange like update iMac Pro in 1Half 2019 and for some goofy reasons stretch out the Mac Pro pain as far as possible into 2019 looking for quirky tech that never was going to be late 2018 tech. If they were shooting for late 2019 tech all along, then I suspect that it will bite them in the butt and will slide into 2020 to rather dubious market outcomes. Even fewer are going to wait that long. After Apple does another pulls another "Lucy" to Mac Pro users "Charlie Brown" in April 2019 and pulls the football and says "sike it isn't 2019 ... really 2020" that will be it for more than a few.
Over the much longer term though the fact that Intel (and AMD ) are getting deeper into packaging multiple dies onto the CPU package it won't be surprising if perhaps a future Mac Pro gets a dGPU die stuffed into the "CPU" processor package. That doesn't mean there would be a slot for an additional GPU, but that could roll up the CPU+GPU+Thunderbolt integration all into a neat (almost complete except some support chips) package for the Mac Pro also.
Since Microsoft opened itself for OpenSource initiatives, and Linux development on Azure, there is ongoing speculation, that is becoming more and more loud, that Microsoft could release their own, full blown, Ubuntu-Like Distro of Linux, for all of computing.
Linux is and always will be Unix-Based OS, just like Mac. If MS's Linux would allow the same performance as Windows, 1:1 compatibility with software, what would be the point of buying a Mac? For any type of usage?
I know it is dreamland, but... https://www.zdnet.com/article/ms-linux-lindows-could-microsoft-release-a-desktop-linux/
I so much love Unix based OSes. But the fact that Windows is so powerful, and that there is plenty of software on it - makes me stuck to it.
Linux is and always will be Unix-Based OS, just like Mac. If MS's Linux would allow the same performance as Windows, 1:1 compatibility with software, what would be the point of buying a Mac? For any type of usage?
I know it is dreamland, but... https://www.zdnet.com/article/ms-linux-lindows-could-microsoft-release-a-desktop-linux/
I so much love Unix based OSes. But the fact that Windows is so powerful, and that there is plenty of software on it - makes me stuck to it.
There was some time ago a discussion that If Apple would drop Intel from their computers, they would either pick Semi-Custom AMD solutions, or their own ARM. Ashraf Aeassa'a argument was always that Intel could do custom hardware for Apple if Apple would want to because of their cutom foundry initiative.
Well:
https://twitter.com/david_schor/status/1074681154765250560
Its being dropped. No more custom foundry business for Intel, Im affraid...
Well:
https://twitter.com/david_schor/status/1074681154765250560
Its being dropped. No more custom foundry business for Intel, Im affraid...
Thamx dec
I'm aware that Sunny is a new arch, said it in my post, but maybe I got it wrong but I was under the impression that Intel mentioned availability for 2019. With a roadmap too. Followed by Willow Cove by 2020 and golden Cove in 2021. Here:
W 3175X doesn't seem an option.
I'm aware that Sunny is a new arch, said it in my post, but maybe I got it wrong but I was under the impression that Intel mentioned availability for 2019. With a roadmap too. Followed by Willow Cove by 2020 and golden Cove in 2021. Here:
W 3175X doesn't seem an option.
I'd like to see mMP get PCIe4 but I guess that's a no go.
From the Intel camp, at least.
Just to make it last a bit longer... by the time it gets updated again PCIe5 should be out
From the Intel camp, at least.
Just to make it last a bit longer... by the time it gets updated again PCIe5 should be out
I'd like to see mMP get PCIe4 but I guess that's a no go.
From the Intel camp, at least.
Just to make it last a bit longer... by the time it gets updated again PCIe5 should be out
From the Intel camp, indeed, as the AMD X570 chipset is supposed to support PCIe4...
X570 ITX motherboard, Ryzen 9 3850X (7nm, 16c/32t, 4.3GHz base/5.1GHz boost, 135w), & Radeon RX 3080 (7nm, Navi 10, 8GB GDDR6, 150w); seems like it would make a solid xMac...
Switch to a X599 mATX motherboard, 7nm Threadripper CPU & 7nm Vega 2 GPU(s), would make a solid Mac Pro...
I would love to see that lineup. A lot of power and prices could start well below the current Mac Pro.
But that already tells us, that Apple won't do it.
This might be the real issue : the lack of competing operating systems on the mainstream PC level .
There are two, and one of them is dragged down by proprietary hardware and an ancient marketing philosophy .
As much as I like Mac hardware ( of 5 years ago ... ) for built quality and looks, the only reason I'm still using Apple is OSX .
It's not like Apple has ever provided more advanced computers than anyone else, and I highly doubt they will start now .
The ideal MP would be whatever is running a cross platform OSX .
Which basically means: Microsoft's Linux Distro that is fully 1:1 compatible with Windows applications, with 0% performance difference.
I'm afraid you are right .
Ideally it would be an independent 3rd party Unix based OS, but the OS competition train has left a long time ago it seems .
Too complex I assume, and the market fenced off by MS and United Fruit .
Apple is too lazy to support more than a ball pen at this point, so no hope there .