MP 7,1 Waiting for Mac Pro 7,1 A1991 (no more)

[wallysb01]

I believe Apple should/would go for Cascade Lake, not SKL. If not for the Spectre and Meltdown issues. But Intel's woes on 10nm might have compromised the design. Are there any ES for Apple to test? If not, the design might even be more or less locked but with no hardware for testing they might be on hold.
Or... as one would say, one more thing... and it's EPYC!!
That would be a good line to wrap things up, maybe on October 30th.
Maybe this would be a good time for Apple to move to AMD (never thought I'd say this) and Rome is already in Apple's hands.
I don't think we'll have any info on mMP in a couple of days, though.

Cascade Lake is 14nm though and I very much doubt, given the lateness of Apple's planning on the mMP, that the delay to 10nm wasn't already a consideration. After all the Xeon class stuff is always a year or two behind the consumer roll outs anyway. I also don't know the pragmatics of the issue, but I'd very much doubt Apple would want to split CPU suppliers between MPs and the rest of the Mac lineup. As it is, Cascade Lake is due in early 2019. That really should parallel MP timelines. Consumer 10nm is probably more like 2019Q2 or later, and being that to the Scalable line must be at least year behind that. Spectre and Meltdown will still be issues, but at the same time, Apache Pass will allow for some interesting options for limited DIMM slot MPs. Something not possible with Epyc.

 

[deconstruct60]

Any kind of move to AMD would be at WWDC since there are programming differences.

The vast majority of developers using XCode ( and other tools) are not macOS developers. Switching to AMD x86 processors would have zero impact on iOS , tvOS , watchOS , and Safari extensions developers. The majority of the WWDC is actually for them and high level frameworks. Even macOS apps that only use Apple Framework libraries there about zero difference. There is a smattering of developers who use the POSIX/Unix base to do java and web services development.... again approximately zero relevant impact to those.

Go look through past WWDC sessions over the last 3-4 years. See anything like "Hand optimizing x86-64 assembly " sessions? Nope.

There is a subset of kernel driver and extensions folks would be impacted by the narrow variances, but the hugely vast majority of AMD opcode space is the exactly the same as Intel's. Apple's kernel folks would need to make adjustments but they don't need WWDC to do that. There is a small set of folks who do have hand-tuned assembler ... that varies by Intel micro-architecture implementation too so no dire WWDC tie-in for that. It is farce to present this as something like the PPC-> x86 shift or the 68K-> PPC shift. It isn't. Mostly hand waving to way Apple needs to boat anchor the Mac Pro to the WWDC conference.... which is largely spacious.

 

[barmann]

That is a big leap having not seen the next iPad Pro.

If it has a USB Type-C port the recent rumors indicate, it won't be that far from being what the MacBook currently is only running iOS. It wouldn't be hard for Apple in 2020 to put a next gen A-series chip in the MacBook frame and call it "iBook". The next thing you'll tell us is that almost nobody using "real" Photoshop on a highly portable laptop. Which is a chuckle.

If Apple keeps the MacBook hobbled on a USB Type-C long term ( just to hit minimal weight specs) then some follow on to the UltraFine 4K docking station monitor will probably work just fine as a external presentation display ( although something less would probably work better is want to balance storage I/O with display) with at least a future iPad Pro ( if not the 2018 ones ... which is a decent chance). Displays aren't probably aren't a huge issue.

Bluetooth is sufficient for keyboard , mouse, trackpads.

Adobe has talked more about this new Photoshop for iOS. It won't match up 100% feature for feature with the macOS on this first iteration. But it is likely to be enough for a sizable group of folks. All Apple and Adobe have to do is iterate on that over time. Short term pointing at corner cases it doesn't does little to outline the value proposition for most of the overall market.


There is a fair amount of focused work done on Pen Displays (e.g., Wacom Cintiqs ). Some of them are already USB Type-C oriented. At first the next iPad Pro could be a replacement. But they aren't that far from using them as an adjunct. (apple's 'Pro' tagged systems aren't absolutely necessary for usage. )

The absolutely exclusive space that the Mac Pro type product from 10-15 years ago is substantively smaller than it is now. That's part of the issue with Apple's priority on a new Mac Pro.
[doublepost=1540413852][/doublepost]

The connectivity of an iPad to an elaborate setup as required for even a fairly simple editing program like Photoshop just isn't there yet .
Given the pace Apple is moving at, and their insistence on third party manufacturers delivering many of the peripherals most times, I'm not expecting speedy progress .

Apple could and should have made the iPads viable input devices for Macs years ago, they still are not even close .
Which is a tad pathetic and makes it unlikely they intend to go beyond their current attempts at creating superficial funcionality for sidewalk artists .

As for the iPad / iOS by itself, tablets are too crude for more than basic editing ; not for just the akward and slow touchscreen operation , the small displays and calibration issues, their limited power, but also because of the lack of a file system and finder, and poor workflow compatibility with any desktop OS .

And no - hardly anyone is editing on a laptop - many, myself included , use laptops in the field with image/video etc. capture and editing software .

However, serious eding is done with keyboard, mouse, maybe a wacom tablet and big screen(s) - which might be attached to a laptop instead of a desktop, turning that laptop into a workstation component .

In that respect I agree, it doesn't need to be a MacPro at the heart of an editing setup, but it won't be a (Apple) tablet anytime soon .

 

[-hh]

Surprised Apple did 300+ logos for the event. Hopefully most of that was commissioned work and weren't soaking up valuable internal resources doing that.

April 2019:

"Well, we have to admit that the Mac Pro design team got distracted last fall on fancy logos..."

/S

Apple is also looking looking for a new major US Campus ( and some rumors has the pegged for PA , NY (not the city) , or "NorthEast" ). NYC makes sense a place to scout some cities from. ( "Oh we came to NYC early to practice for dog and pony show ...." cover for doing some in person scouting. ) . Apple is running their search 100% opposite of Amazon's spectacle contest search process. On the not so sneaky side, this Bloomberg article puts a chunk of Apple's advertising biz operations in NYC (https://www.bloomberg.com/graphics/2018-apple-new-campus/ ) . [ similar suspicion scouting Chicago area at March Education / iPad event. ]

Interesting ... and I'd forgotten about the Campus search. YMMV, but if I were one of these firms, I'd pick Philly. Its conveniently located in the middle of the action of the Northeast (halfway between NYC & DC), has some solid .EDU's, etc ... but the kicker is on cost: it is surprisingly depressed/affordable.

 

[deconstruct60]

I believe Apple should/would go for Cascade Lake, not SKL. If not for the Spectre and Meltdown issues. But Intel's woes on 10nm might have compromised the design.

There is nothing particularly compromised. Intel has pulled a small subset of features back to 14nm for Cascade Lake and Cooper Lake. There are 2018 (which probably won't ship with volume until 2019. ) and 2019 (which will pick up some backported features and socket changes. )

https://www.anandtech.com/show/1320...tels-dcg-discussing-cooper-lake-and-smeltdown

Intel isn't going to have any substantive core count and major throughput increases, but 'compromised' is the wrong notion. Intel will probably be running them hotter ( to get some small clock bumps. ). Adequate cooling that shouldn't be a major problem. ( probably would have trouble seeing a difference in an iMac Pro since can't really leverage the clock bump much. )

Ice Lake is probably later than this "mid 2020' prediction in the artucle above because Intel has been consistantly wrong by a quarter or two for years at this point.

Are there any ES for Apple to test? If not, the design might even be more or less locked but with no hardware for testing they might be on hold.

Engineering Samples for Cascade lake. Yes there are. in fact Intel is probably already shipping "early production" to all the major clouds ( AWS , Azure , Google , Facebook ) already (or in a few weeks). Those are the Scalable Processor variation. The refactored "workstation" line up seems to be a bit muddled. Intel has thrown another socket into the Xeon W mix trying to keep up on the count count wars. I doubt Apple would use that at all. The 3175X seems more a marketing gimmick that useful in a Mac Pro ( or iMac Pro).

https://www.anandtech.com/show/1345...-xeon-w-processors-fixes-for-spectre-meltdown

It is basically a specially binned Skylake era die. The relative scaricity o the 9900-9700's points to why Cascade lake is sliding for general volume availability. Intel's Fabs are backlogged ( in contrast to claims that all their orders were rapidly drying up. )



[quote
Or... as one would say, one more thing... and it's EPYC!! [/quote]

Apple might use Threadripper but not EPYC. There is not really a good reason to at all in terms of hitting reasonable price points and performance points for single user workstations. EPYC isn't primarily targeted at single user workstations.

That would be a good line to wrap things up, maybe on October 30th.
Maybe this would be a good time for Apple to move to AMD (never thought I'd say this) and Rome is already in Apple's hands.

EYPC and Rome aren't what is critical for AMD to get some wins. If Apple was going to move the whole desktop line up ( or desktop + most of laptop line up) to AMD then I could se a shift on the Mac Pro side. But one, probably less than 2% of mac sales, system being out of step with rest of the Mac line ( forked kernel , drivers , etc. ) doesn't make any sense at all.

Apple should have at least some internal hackintosh project looking at switching to AMD. Intel has made some bonehead moves over last couple of years. They might clean up their fabrication act in next year or so but Apple should be turning up the "we are getting ready to leave" heat on them to keep them focused.

I don't think we'll have any info on mMP in a couple of days, though.

If Apple based the core foundation of the next Mac Pro off of what they had already learned and did with the iMac Pro ( Intel W and Vega GPU ) then there is not any decent reason why they shouldn't be about finished at this point. Even if they are waiting for the 'Cascade Lake' single socket derivation the socket is exactly the same. All Apple would be waiting on is firmware and some edge differences to finish getting worked out. The core firmware , the T2 , the central core of the workstation would just be largely done. Adding some empty slots and more DIMM sockets would be just incremental additions to a working infrastructure. And they would had about 18 months to do those increments and put the GPU on a custom card ( to open up the TPD power/thermal window) and facilitate upgrades later .
[doublepost=1540497126][/doublepost]

Cascade Lake is 14nm though and I very much doubt, given the lateness of Apple's planning on the mMP, that the delay to 10nm wasn't already a consideration.

Not only is Cascade lake 14nm , the next in line , Cooper lake is 14nm too. 10nm for intel means waiting probably until Q3-Q4 2020 and that's practically suicide for the Mac Pro as a product. ( and really wouldn't pragmatically buy much anyway. 10nm is likely to run about as hot and heavy at the 14nm+++++ stuff that is options in next several months into 2019 ).

After all the Xeon class stuff is always a year or two behind the consumer roll outs anyway.

Not really. In terms of Spectre/Meltdown fixes the upper level Xeon class is going to have more hardware fixes than the desktop stuff pretty soon now. (Cascade Lake has more ). Intel's roadmap for SP ( and likely Xeon W derivatives ) has yearly ticks for next two years to make up for fact missing out on 10nm for so long.



I also don't know the pragmatics of the issue, but I'd very much doubt Apple would want to split CPU suppliers between MPs and the rest of the Mac lineup. As it is, Cascade Lake is due in early 2019. That really should parallel MP timelines.

Consumer 10nm is probably more like 2019Q2 or later, and being that to the Scalable line must be at least year behind that.

I would be more on the "or later" for most of the consumer stuff. Intel is dropping $1B to expand 14nm production capacity.

https://www.anandtech.com/show/1341...4nm-demand-prioritizing-highend-core-and-xeon
"... so any investment now is unlikely to increase fab throughput until end of Q1 next year. ..."

if Intel was doing a huge ramp of 10nm to shift off of 14nm in Q2 2019 they wouldn't need a big boost in 14nm capacity at the end of Q1. As the scalable line shifts to multiple dies and EMIB packaging to get to very high core counts, the gap won't be as big because the dies will get just a bit closer in size.

Spectre and Meltdown will still be issues, but at the same time, Apache Pass will allow for some interesting options for limited DIMM slot MPs. Something not possible with Epyc.

But the Optane DIMMs would need some kernel support too. It would nice if Apple did that. But I also don't see Apple putting the resources into making that happen. Apple's mark up on Optane DIMMs would be 'ugly' too.

EPYC has substantive NUMA and scheduling issues too. ( so does Intel W but much less so. )

 

[Lammers]

If Apple is ready to ship in Feb/March about something they talked about on October 30th ( defacto November ) there is exceedingly little risk. The risk isn't October per se. It is whether they are ready or not. At some point they have to ship and probably won't ship an update for at least 12-18 months. For the majority of the life span of the product, the competitors will have a window to respond.

If Apple has no substantive, coherent system done at this point then yes there is a higher substantive risk. They probably shouldn't show anything. The competitors may have more competent R&D teams that can finish off a new design in 9-10 rather Apple's sloth-like 14-16 months.

Yes, I originally included the delivery time dimension in my post but then removed it for brevity. But yes, the latency between any sort of "sneak peak" and the product shipping is what drives the risk. My point is not to evaluate that risk but simply that it is a non-zero risk, and therefore will factor into Apple's decision-making over whether or not to talk about Mac Pro at the event.

And as I said in my post, for a "sneak peak" to have enough value to actually bother with it needs to include sufficient detail to serve its purpose - so I assume if Apple doesn't have a "substantive, coherent system" at this point then they wouldn't do a "sneak peak" anyway.

Major events is not where you go to talk about your major screw ups.

On a related note to this, I do also wonder whether an iPad-focused event, presumably intended for a mass-market audience, is the sort of event where you talk about Mac Pro - perhaps the most niche product in their portfolio - at all.

The laptops that is far more driven by "thinness" and having 2D space of required keyboard to play with.
The iMac is in trying to hide the ram door in case shell behind the pedestal arm along with the exit fan vent. The mini is trying to hide it in a bottom placed fan ( which is its own problem.) The iMac Pro compromised on narrow slits at the bottom back for airflow but door got dropped as soon as folks would be able to see it.

The Mac Pro doesn't necessarily have those problems. As a deskside unit, it does not all. The older design did a relatively good job of lining up the lid seal with a sight line the device already had. From a reasonable distance folks couldn't see the line and frankly stored below/beside/under a desk the whole system is largely out of sight in most places too. However, even relegated to being a literal desktop the Mac Pro 2013 case also completely came off without any huge "door' disruption of the case design.

That Mac Pro is probably going to be big enough that a large door/lid that comes off in one whole large piece won't be disruptive. The MP 2013 was designed after/during 2012-12 era for the laptop Mini. The 2013 Mi A couple of standard slot "doors" really should be either unless Apple got hugely anal about what the bck of the machine looked like. ( if it were to stay tagged as being a literal desktop then that's possible. ).

The "why" is irrelevant, in my view. And actually I disagree with your suggestion (if I'm reading it correctly) that the loss of upgradeability is a consequence of other things. I believe that the loss of upgradeability is a deliberate product strategy and design principle, and one that permeates across Apple's portfolio. And I make this point because I think it can help inform our thoughts on the new Mac Pro. I believe that the sort of upgradeability that people here want and expect of the new Mac Pro is an anathema to Apple's product principles. And I think that's clear from looking at Apple's behaviour.

The 21.5" iMac in most cases just "hard to get at" upgradability. Not whether it is upgradable or not. There are so-DIMMs slots there; it is just hard to get to them. Same with iMac Pro. If the screen came off easy it would be a hassle. If there was a visible door cut in the case it wouldn't be hard at all. The Mini 2014 was a shift following the Mac Laptops.

It's not about whether it's POSSIBLE, it's about whether it's INTENDED. These products are INTENDED, deliberately, to not be upgradeable by end users. And that's explicitly stated in Apple's support documentation too. The trend, again, is very clear that Apple has been progressively removing all upgradeability from its products.

The real question is why would Apple lock up the case on a new Mac Pro? IMHO, there are only extremely hand waving justifications for that.

I'd say things like:
- To improve system reliability by only selling Apple-certified "modules" that are engineered and deeply tested with the Mac Pro and its ecosystem for performance and reliability.
- To avoid the traditional driver issues that people have on Windows from the ability to plug in any third-party hardware.
- To be able to sell Apple-certified modules at a margin.
- To be able to sell complete system "solutions" with a set of modules/components/whatever designed for specific use cases (photography, video editing, whatever).

In my view, such a practice would be more in-keeping with Apple's established behaviour.

Laptops are extremely different where weight and battery capacity are far more intertwined. Phones and tablets even more so. But decoupled from the keyboard , display, most laptop oriented parts , and from literally placed on the desktop there is no reason why the even Apple's design trends for those other products the Mac Pro should fall under those constraints.

I'm not suggesting the loss of upgradeability is a symptom of a physical constraint. I'm suggesting the loss of upgradeability is a deliberate design choice.

 

[ActionableMango]

I do also wonder whether an iPad-focused event, presumably intended for a mass-market audience, is the sort of event where you talk about Mac Pro - perhaps the most niche product in their portfolio - at all.

Everyone is billing the Oct 30 event as an iPad and Mac event, so I wouldn't call it an iPad-focused event.

Apple just registered in their Eurasian database three new desktops and one new laptop model. Apple is expected to unveil new Mac Mini, iMac, and MacBook... so what's the third desktop? I guess it could be a updated iMac Pro or Mac Pro.

 

[JesperA]

I've been excited for the launch of a revised Mac Pro ever since 2013, nowadays i have become so indifferent due to the extreme long wait that i am looking forward more to their new monitors instead. Hope they show them soon, maybe at the OCT30 event, not likely but yeah...

No rumors about them though which is kinda weird

 

[goMac]

If Apple was going to move the whole desktop line up ( or desktop + most of laptop line up) to AMD then I could se a shift on the Mac Pro side. But one, probably less than 2% of mac sales, system being out of step with rest of the Mac line ( forked kernel , drivers , etc. ) doesn't make any sense at all.

AMD is going to need a different set of drivers but I think the kernel already runs on AMD? I don't think they'd have to fork it. I'm not even sure the kernel needs any AMD specific code to run.

 

[jameslmoser]

AMD is going to need a different set of drivers but I think the kernel already runs on AMD? I don't think they'd have to fork it. I'm not even sure the kernel needs any AMD specific code to run.

There are people running hackintoshes with AMD.

 

[Joe The Dragon]

I believe Apple should/would go for Cascade Lake, not SKL. If not for the Spectre and Meltdown issues. But Intel's woes on 10nm might have compromised the design. Are there any ES for Apple to test? If not, the design might even be more or less locked but with no hardware for testing they might be on hold.
Or... as one would say, one more thing... and it's EPYC!!
That would be a good line to wrap things up, maybe on October 30th.
Maybe this would be a good time for Apple to move to AMD (never thought I'd say this) and Rome is already in Apple's hands.
I don't think we'll have any info on mMP in a couple of days, though.

Yes, I originally included the delivery time dimension in my post but then removed it for brevity. But yes, the latency between any sort of "sneak peak" and the product shipping is what drives the risk. My point is not to evaluate that risk but simply that it is a non-zero risk, and therefore will factor into Apple's decision-making over whether or not to talk about Mac Pro at the event.

And as I said in my post, for a "sneak peak" to have enough value to actually bother with it needs to include sufficient detail to serve its purpose - so I assume if Apple doesn't have a "substantive, coherent system" at this point then they wouldn't do a "sneak peak" anyway.

[/QUOTE]

What about just saying on sale DEC 2018 -- shipping early 2019?? But saying late 2019 may just kill sales.

On a related note to this, I do also wonder whether an iPad-focused event, presumably intended for a mass-market audience, is the sort of event where you talk about Mac Pro - perhaps the most niche product in their portfolio - at all.


The "why" is irrelevant, in my view. And actually I disagree with your suggestion (if I'm reading it correctly) that the loss of upgradeability is a consequence of other things. I believe that the loss of upgradeability is a deliberate product strategy and design principle, and one that permeates across Apple's portfolio. And I make this point because I think it can help inform our thoughts on the new Mac Pro. I believe that the sort of upgradeability that people here want and expect of the new Mac Pro is an anathema to Apple's product principles. And I think that's clear from looking at Apple's behaviour.


It's not about whether it's POSSIBLE, it's about whether it's INTENDED. These products are INTENDED, deliberately, to not be upgradeable by end users. And that's explicitly stated in Apple's support documentation too. The trend, again, is very clear that Apple has been progressively removing all upgradeability from its products.


I'd say things like:
- To improve system reliability by only selling Apple-certified "modules" that are engineered and deeply tested with the Mac Pro and its ecosystem for performance and reliability.
- To avoid the traditional driver issues that people have on Windows from the ability to plug in any third-party hardware.
- To be able to sell Apple-certified modules at a margin.
- To be able to sell complete system "solutions" with a set of modules/components/whatever designed for specific use cases (photography, video editing, whatever).

In my view, such a practice would be more in-keeping with Apple's established behaviour.


I'm not suggesting the loss of upgradeability is a symptom of a physical constraint. I'm suggesting the loss of upgradeability is a deliberate design choice.

but say BS like to stop an sata port is just a ripoff
useing driver issues that people have on Windows from the ability to plug in any third-party hardware??? why also lockout non apple usb and TB hardware as well?

if they are ok with e-gups why not have full size pci-e slots? (other maybe can't fit in thin case)
[doublepost=1540515313][/doublepost]

AMD is going to need a different set of drivers but I think the kernel already runs on AMD? I don't think they'd have to fork it. I'm not even sure the kernel needs any AMD specific code to run.

stock Linux ones work on AMD and intel.
apple just puts stuff in to lock out non apple stuff.

 

[JesperA]

stock Linux ones work on AMD and intel.
apple just puts stuff in to lock out non apple stuff.

Where does Apple put stuff in to lock out non Apple stuff? In Linux? No, Apple does not put anything of the sort into Linux. They also don't put anything into XNU, that would make even less sense than your Linux comment. You think Apple develop their kernel with support for AMD and THEN lock AMD out? Obviously not, they just don't put any AMD support into their current non-internal kernel. Third parties have added support for AMD in the kernel so hackintosh users can use AMD hardware. Apple most likely have an internal version of XNU that have full support for AMD but that is not released to the public (only if and when Apple starts to offer/sell AMD CPU:s in their computers)

 

[Lammers]

Everyone is billing the Oct 30 event as an iPad and Mac event, so I wouldn't call it an iPad-focused event.

Apple sells several times more iPads than they do Macs, has an install base of iPads several times that of Macs, and is growing sales of iPad whereas Mac sales have stagnated recently. FAR more people care about iPad than about Mac. Plus we're entering the holiday season, and iPads make good Christmas presents. So accordingly I expect iPad to get more attention at the event than Mac, and likewise I expect press coverage to lead with the iPad news first and foremost.

 

[Lammers]

What about just saying on sale DEC 2018 -- shipping early 2019?? But saying late 2019 may just kill sales.

I'm not sure it makes sense to announce a release date for a product but not tell people about the product.

if they are ok with e-gups why not have full size pci-e slots?

This is actually a good example of what I'm talking about. It's not that they're "ok" with eGPU - they actually specifically designed and engineered a solution for eGPU, including both the software support in macOS, and the hardware solution in collaboration with a partner (that was initially only available to buy from the Apple Store). You don't have the ability to just plug any GPU card directly into the motherboard. You only get to use a specific set of GPU cards, with a specific set of eGPU expansion chassis, and use it with Apple's native eGPU implementation (which has its own set of constraints). It's quite different from what happens in typical workstation-land in Windows.

 

[Mago]

Apple sells several times more iPads than they do Macs

 

[askunk]

AMD on their 2019 roadmap:

[doublepost=1540556449][/doublepost]It seems AMD is down on multi-GPU. Will Apple insist on the matter with the nMP?

 

[filmak]

Apple sells several times more iPads than they do Macs, has an install base of iPads several times that of Macs, and is growing sales of iPad whereas Mac sales have stagnated recently. FAR more people care about iPad than about Mac. Plus we're entering the holiday season, and iPads make good Christmas presents. So accordingly I expect iPad to get more attention at the event than Mac, and likewise I expect press coverage to lead with the iPad news first and foremost.

Many people care for different things. You cannot be sure that they will focus on iPads because they sell a lot more. It's an ecosystem, each product has a direct relationship with the others.
One can claim that the Mac App store or the iTunes store, or Apple services in general, are more profitable so they will have to focus on them, but what are Apple's services without the hardware? What are iPads without the Macs to build apps for them and support them? How can a large organization manage iPads without Macs?

So, I think that,
iPad will get more attention from iPad users.
Macs will get more attention from Mac users. (And there are still a lot of them)
Macs and iPads will get a lot of attention from people who care for both of them.

If Mac sales have stagnated recently, this is another great reason to focus on them.

And everything from Apple's products is a great Christmas present..

 

[Mago]

Dont get crazy, Oct'30 is an PRO Keynote, the diversity of logos, only means apple sees (or they want us to believe that) a diversity of PROs users, from that, you can deduce all the launches will be PRO related and conclude with the mMP introduction (but avail 2019).

Analize that.

What I spect:

1. Opening: a 30 minute video showing any kind of profeesional using Apple Stuff, from vaccums salesman (as T.Cook) to Neil deGrasse Tyson seaching for black holes (or a good joke, which seems he is chasing all the time, w/o good luck)
2. Then Updated iPad Pros, and updated iPad mini (becuse only vertical pros markets still use iPad minis).
3. MacBooks
4. Then updated iMacs, and maybe an updated iMac Pro.
5. New peripherals, as TouchBar enabled Magic Keyboard (also maybe an premium Magic Keyboard with two color e-Ink keycaps for those on video editing, etc requiring multiple Kbd)
6. Closing with an updated "PRO" Mac Mini (dont expect more than 6 cores unless Apple switched to AMD).
7. And One Last Thing: a Sneak Peek or introduction to the all new "modular" Mac Pro...

And closing with the blatant smiles from Cook, Federighi, and "Cant Innovate My Ass" Schiller...

 

[Joe The Dragon]

I'm not sure it makes sense to announce a release date for a product but not tell people about the product.

This is actually a good example of what I'm talking about. It's not that they're "ok" with eGPU - they actually specifically designed and engineered a solution for eGPU, including both the software support in macOS, and the hardware solution in collaboration with a partner (that was initially only available to buy from the Apple Store). You don't have the ability to just plug any GPU card directly into the motherboard. You only get to use a specific set of GPU cards, with a specific set of eGPU expansion chassis, and use it with Apple's native eGPU implementation (which has its own set of constraints). It's quite different from what happens in typical workstation-land in Windows.

so why just have a small range of pci-e gpu chips that work?? or let AMD / nvidia just do the drivers?

 

[mattspace]

It seems AMD is down on multi-GPU. Will Apple insist on the matter with the nMP?

AMD is big on whatever AMD is capable of making.

Closing with an updated "PRO" Mac Mini (dont expect more than 6 cores unless Apple switched to AMD).

So who wants to bet the next mini doesn't include thunderbolt / eGPU capability. It's USB-C only.

 

[deconstruct60]

AMD on their 2019 roadmap:

View attachment 798450 View attachment 798451
[doublepost=1540556449][/doublepost]It seems AMD is down on multi-GPU. Will Apple insist on the matter with the nMP?

Errrrr, when has gaming been the primary macOS (or even top 10) focus ? Even less so for the Mac Pro ... primarily built for gaming? Go back and watch that Mac Pro 2006 introduction video clip someone linkedi in a page or to back. Gaming mentioned even once? No. That AMD is 'down' on multiple GPU dies for gaming doesn't particularly have any Mac or Mac Pro relevance.


That doesn't mean that the standard configuration for the Mac Pro would be dual/mulitple GPUs. However, it would be a extremely bozo dubious move to exclude more than one GPU. If Apple is serious, there should be a slot for 'compute' GPUs. ( if someone wants to add an Instinct , 'Tensor' , Nvidia, or other ) learning/inferencing card that should be an option. Apple doesn't have to sell, just enable the easy usage.

Multiple dies ( and likely multiple RAM stacks) on the same card likely means it will run on the 'hotter side'. Apple needs a thermal/power envelope that is higher than the 400-500W range that the iMac Pro ( MP 2013 ) have. That means more space (volume) , power , and airflow . Multiple dies on the same heavy computation card is likely to increase the bandwidth demands. ( so breaking it out into a seperate box with a slower interconnect won't help. That is why dies from potentiall two cards are being brought closer... better bandwidth/latency. )
[doublepost=1540566714][/doublepost]

...
What about just saying on sale DEC 2018 -- shipping early 2019??

In Apple's Apple 2018 meeting they explicitly said there would be no Mac Pro sales in 2018. That was given as fiscal guidance so some people could long term plan how to spend their money in 2018. Starting pre-orders (but calling them 'on sale' ) in December and technically just pushing the fulfillment into January as some kind of a loophole move is silly. if they really won't be ready to ship until mid-late Q1 then can just take the pre-orders in 2019 also. There is no upside to starting pre-orders months earlier than can ship.

For the MP 2013 Apple started 'sales' before their factory had time to ramp up on inventory for the launch. That was a 3-4 month angst of shortages. Folks :standing in line" isn't really a 'good thing' in the pro space. Especially when grossly late. A month ramp so that know they do not have any production problems would be far more prudent.


If Apple has somehow has nothing physical to show, but would in December then it would be better just to say that there is another, smaller dog and pony show coming in December for the Mac Pro and that it is a 2019 product. The 'new' news is that they are putting a date on giving out details, not the product shipping. And yes that would be helpful to some. Others with a time issues it won't be, but if not buying anything until after March-April in 2019 anyway it doesn't really make a huge difference. Apple will have information they can fold into their 2019 purchase planning process before the end of the year. That's a good thing.

But saying late 2019 may just kill sales.
....

Like a 6 ( and/or 9 ) year old Mac Pro isn't going to kill sales also? Apple is in the which bad outcome does the least damage zone. All they have is bad choices. So yeah not talking about 'late 2019' ( in late 2018) is part of the least worst choices, but all of their choices at this point are sales killers.... all of them.

Some people are going to leave because they have stuff to do with newer equipment and Apple has none for large chunk of those squatting on a older Mac Pro ( or looking to move off an approximately equally as old Windows systems . Some folks who 'left' in 2013-2014 are up on a renewal cycle. ).

 

[wallysb01]

Apple sells several times more iPads than they do Macs, has an install base of iPads several times that of Macs, and is growing sales of iPad whereas Mac sales have stagnated recently. FAR more people care about iPad than about Mac. Plus we're entering the holiday season, and iPads make good Christmas presents. So accordingly I expect iPad to get more attention at the event than Mac, and likewise I expect press coverage to lead with the iPad news first and foremost.

They might sell more iPads, but iPad sales are not growing any longer.

That's been true for about 4 years now. There might be a very small bounce here after hitting the bottom in Q2 2017, but its effectively flat.

I also wouldn't equate more sales to more people caring about iPads. They are cheaper, so its easier to do things like give one to your kid for educational apps and youTube or the Christmas present thing you mention. In an 'Apple family', the kids often get the hand-me-downs as far as the computers go. Or they might just get Chromebooks or cheap a Windows POS. But none of that means people *care* more about iPads.

Forward looking, I expect tablets to continue to get squeezed by Phablets and newer MacBooks/MacBook Airs. Both phones and computers have more use cases than tablets in the same price ranges. Its going to continue to become a very niche thing to need a ~8-12 inch screen for something, when you already carry around a 5-6+ inch screen and a 12-15 inch screen. Also, iPad margins have been getting squeezed. Something that's not really happening with iPhones or Macs. That indicates that tablets are more price sensitive. And that makes a lot of sense if its a give it to your kids or as christmas gift item, instead of something I buy for myself because I really need it and I want it to be the very best it can be.

I would also be skeptical that the iPad line gets more play than the Mac line at this event. The iPad will probably open (after the usual pat yourself on the back moments), but the Mac line has potentially every single computer outside the MacBook Pro to talk about: Air, Mini, MacBook, iMac, iMac Pro and Mac Pro. And probably in that order of likelihood of appearing and amount of time given if introduced at all. Two of the most likely things to appear are completely new models (Air and Mini). The iPad isn't going to see 'new model' time, just new capabilities time. The MacBook, iMac and iMac Pros would be spec bumps, so probably not a ton of time given, but still a slide each, maybe 2 (fastest iMac and Mac ever comments could soak some time). And the Mac Pro, if we see it all, would a probably just be a quick tease. Even crossing a few of those things off the Mac list, that's a lot of stuff to talk about compared to beating the 'full version' app drum on the iPad side repeatedly.
[doublepost=1540568899][/doublepost]

Not only is Cascade lake 14nm , the next in line , Cooper lake is 14nm too. 10nm for intel means waiting probably until Q3-Q4 2020 and that's practically suicide for the Mac Pro as a product. ( and really wouldn't pragmatically buy much anyway. 10nm is likely to run about as hot and heavy at the 14nm+++++ stuff that is options in next several months into 2019 ).

Thanks for the info, I hadn't payed attention to Cooper lake on 14nm.

I would be more on the "or later" for most of the consumer stuff. Intel is dropping $1B to expand 14nm production capacity.

https://www.anandtech.com/show/1341...4nm-demand-prioritizing-highend-core-and-xeon
"... so any investment now is unlikely to increase fab throughput until end of Q1 next year. ..."

if Intel was doing a huge ramp of 10nm to shift off of 14nm in Q2 2019 they wouldn't need a big boost in 14nm capacity at the end of Q1. As the scalable line shifts to multiple dies and EMIB packaging to get to very high core counts, the gap won't be as big because the dies will get just a bit closer in size.

True, but as you can tell. Even if consumer things move to 10nm sometime soon, there will be plenty of 14nm chips being produced for some time, including the chipsets. Also, there are market competition issues at hand. If Intel doesn't boost capacity soon, they supply issues might lead to loosing contracts to AMDs fairly competitive CPUs and once people start transitioning it may be hard to win them back. Intel isn't going to have too hard of time keeping up with AMD on performance in a pure sense, but if they can't supply them, it doesn't matter.

But the Optane DIMMs would need some kernel support too. It would nice if Apple did that. But I also don't see Apple putting the resources into making that happen. Apple's mark up on Optane DIMMs would be 'ugly' too.

EPYC has substantive NUMA and scheduling issues too. ( so does Intel W but much less so. )

Optane DIMMs would be expensive, true. But getting up to 512 GB out of a single DIMM and being able to populate the other DIMMs with standard DDR4... whoa. That opens a TON of doors for relatively small footprint computers with low overall costs otherwise. I don't think we've seen prices yet, but many users won't be particularly price sensitive if it does something that ordinarily would require a server with 16 DIMM slots to be populated with lots of not-cheap-either DDR4. I'd certainly be sad if Apple couldn't bother to support Apache Ridge. Enough so, they might lose 10s of thousands of dollars of sales from me.

 

[ActionableMango]

Apple sells several times more iPads than they do Macs, has an install base of iPads several times that of Macs, and is growing sales of iPad whereas Mac sales have stagnated recently. FAR more people care about iPad than about Mac. Plus we're entering the holiday season, and iPads make good Christmas presents. So accordingly I expect iPad to get more attention at the event than Mac, and likewise I expect press coverage to lead with the iPad news first and foremost.

Nobody is questioning what the sales numbers are. New iPads and new Macs will be announced, and so this is an iPad and Mac event.

 

[goMac]

Where does Apple put stuff in to lock out non Apple stuff? In Linux? No, Apple does not put anything of the sort into Linux. They also don't put anything into XNU, that would make even less sense than your Linux comment. You think Apple develop their kernel with support for AMD and THEN lock AMD out? Obviously not, they just don't put any AMD support into their current non-internal kernel. Third parties have added support for AMD in the kernel so hackintosh users can use AMD hardware. Apple most likely have an internal version of XNU that have full support for AMD but that is not released to the public (only if and when Apple starts to offer/sell AMD CPU:s in their computers)

Yeah, and XNU is open source. You can download XNU today from Apple and I'm pretty sure run it on AMD. Apple used to distribute all of Darwin together with no lockouts to Apple gear. You could run it on any generic PC.

Hence my confusion on custom kernel whatever.

 

[deconstruct60]

They might sell more iPads, but iPad sales are not growing any longer.

That's been true for about 4 years now. There might be a very small bounce here after hitting the bottom in Q2 2017, but its effectively flat.

They are growing. If look at the 4 year beginning and ending window they aren't but last 12-18 months they have. Apple's delusion that they have high price elasticity is what they corrected. There is a "low cost" iPad now to deal with the damage that Chromebooks and Windows S type alternatives are offering. Amazon is also healthily taking chunks at the low end consumption end.

That Kuo predicts a revised iPad mini with a lower cost screen ( and likely lower than current "low cost" iPad ) should put the iPad product line as a whole back on track. Trying to retreat into ever more expensive "Pro" iPads (dramatically driving up the average selling price) is a bad solution when competitors are beating you on better value proposition. Apple needs both. Ignoring the minis ( iPad and Mac) and the entry Mac laptops ( MacBook Air) has problems. Dropping the iPhone SE off a cliff is probably going to bite them in the butt in a couple of years too. Apple probably doesn't like looking at those but selling every more expensive stuff can easily turn into a death spiral.

Phablets are phones so that means they have a monthly service cost. A wi-fi iPad mini doesn't. The lifecycle ownership costs are substantially different. What Apple should have seen was that most folks were not going to dump their iPads on a 24 month cycles ( many folks aren't even doing that for phones now either).

To loop back tot he Mac Pro, Apple needs to put a line in the sand on its cost. A Mac Pro baseline price higher than the iMac Pro would be self destructive move. Even moving the base price deep into the $3K range is highly dubious. If they want to tank the overall revenues over the long term there are few thing they could do have have a worse negative effect on sales and revenues.

I would also be skeptical that the iPad line gets more play than the Mac line at this event. The iPad will probably open (after the usual pat yourself on the back moments), but the Mac line has potentially every single computer outside the MacBook Pro to talk about: Air, Mini, MacBook, iMac, iMac Pro and Mac Pro. And probably in that order of likelihood of appearing and amount of time given if introduced at all.

Depends upon whether they also do a iPad mini now rather than later. Kuo's prediction seem to lean on later but if it is sooner rather than later the iPads will get more time budget. It isn't just iPad. If Apple has a new pencil that will likely take up demo time. If they have to explain that the new pencil is "Pro" only... a little bit more.

I don't think Apple is going to break out the iMac Pro here. If the Mac Pro isn't ready and it is extremely likely that the iMac Pro isn't ready for a bump either. If can see them both being lumped into a 'come back in December and will have more to say' line, but details seems doubtful.

I doubt the MacBook is going to take long at all. ( There may not even be an update. Amber Lake brings No Meltdown/Spectre fixes. No GPU update. only some better binned dies for small clock bumps which probably means no gain in battery life. ). The one system in the Mac line that might be on balance better off with an Apple A-series is this 'one port wonder' design. I can see Apple letting that coast for another 6-10 months until perhaps Intel gets their 10nm production cleaned up for limited high volume.

True, but as you can tell. Even if consumer things move to 10nm sometime soon, there will be plenty of 14nm chips being produced for some time, including the chipsets.

Chipsets is part of what the capacity build out is for. From the $1B investment article I had liked in before another quote.

"...In particular, the company had to develop 22-nm new version of its H310C chipset in a bid to free up its 14 nm capacity, according to Tom’s Hardware. .."

The log jam on 14nm is screwing up their chipset roadmaps too. Even if intel moves some CPUs off of 14nm (e.g., Y-series mobile since they are 'smallest' ) then they still have chipsets were scheduled to drop onto 14nm to soak up a large portion of that vacated space. Intel's FPGA business is has an even bigger problem than the CPU line up in that it needs to get onto a good 10nm process in a bigger way ( more competition. ). The cell phone modem is also in worse shape. ( they'll get a bump out of this and maybe next years iPhone but being behind on process is going to hurt later. And it is alot easier for Apple to dump them there. )

Also, there are market competition issues at hand. If Intel doesn't boost capacity soon, they supply issues might lead to loosing contracts to AMDs fairly competitive CPUs and once people start transitioning it may be hard to win them back.

Intel is going to loose some to AMD anyway. Intel's pricing as much as their 10nm woes is going to cause that to happen. In the server space it Intel has some advantages in some subsets and AMD has some advantages on a substantively different subset. There is definite overlap, but there are also differences.

That uptick in AMD is actually going to be a bit of a relief value for Intel's log jam. Tactically if they maneuver their stronger products through behind their weaker products they should do OK and AMD's encroachment will be capped.

Intel isn't going to have too hard of time keeping up with AMD on performance in a pure sense, but if they can't supply them, it doesn't matter.

In GPU computation space yes they do. They will make up bigger chunks in about 18-24 months but for now they are running a big deficit. As for as x86 core counts. It will be a challenge for those who don't care about NUMA effects. It probably should tip a bit negative in late 2019 - late 2020.

For the Mac product context, they are probably not going to be too far apart.

Optane DIMMs would be expensive, true. But getting up to 512 GB out of a single DIMM and being able to populate the other DIMMs with standard DDR4... whoa. That opens a TON of doors for relatively small footprint computers with low overall costs otherwise. I don't think we've seen prices yet, but many users won't be particularly price sensitive if it does something that ordinarily would require a server with 16 DIMM slots to be populated with lots of not-cheap-either DDR4. I'd certainly be sad if Apple couldn't bother to support Apache Ridge. Enough so, they might lose 10s of thousands of dollars of sales from me.

This will likely run into similar problems the "dual GPUs" ran into with the Mac Pro 2013. The high level software programs will have to adapt a bit and many won't. A few will but is that enough to motivate Apple to do kernel changes that only get to the Mac Pro (and perhaps the iMac Pro). Intel may also kneecap the W series on this feature also.

Far more off the shelf apps could adapt to an Optane M.2 SSD scratch drive slot than to the Optane DIMMs with zero changes to kernel or app.

I think Intel will bring Apache Ridge functionality down to the Xeon W level but not before the next socket change ( or perhaps socket and 10+++ nm change) . It would be dubious for Apple to wait for that for the Mac Pro. They should start to scoping the work to put it in later at this point though so it can be ready 18-24 months from now. If they were now getting massively sidetracked with porting macOS over to ARM for a 2020 target for some lower corner case Mac products , then that probably wouldn't happen.

There are just WAY too many usecases where just a reasonably up to date Mac Pro would have leverage than to chase this relatively narrow niche. As Intel makes it easier to get into and expand, Apple should do that. But this bleeding edge and small , that isn't really what they do best. if Apple could master 8 DIMM slots on one CPU package that would be very good progress in and of itself. They could move one from that solid foundation.
[doublepost=1540580639][/doublepost]

AMD is going to need a different set of drivers but I think the kernel already runs on AMD? I don't think they'd have to fork it. I'm not even sure the kernel needs any AMD specific code to run.

fork probably was too narrow a meaning in that concept. It wasn't necessarily a complete different line of code development. It was more so that were going to need to branch (fork) the resources and development process also. A scheduler for a high NUMA system isn't the same scheduler for a 2-4 core laptop, or 2 core low energy phone.

Microsoft has substantive modified kernels to run windows on this 886 core beast then they have on running their Surface products.

https://www.anandtech.com/show/1352...x86-datacenter-class-machines-running-windows

covering all of those spaces requires more resources ( people , equipment , and time ). That is more so the 'fork' I was trying to get at.

AMD NUMA is different than Intel NUMA even at the same product levels. Apple could throw an unoptimized kernel at those CPU packages that 'worked' (as in booted and ran), but they won't be strictly competitive running at the top end if other folks (Linux/Windows) are running tuned kernels and Apple isn't.