Then it would be nice if you provided a link to that statement.
Hint: You won't find it.
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
Hint: You won't find it.
The next Mac Pro is late; grossly late. Actually grossly late is pretty close to an understatement. It is 2017 and with that context Apple goes to Intel has asks for power point slides on what he future looks at. Apple spots a processor that doesn't exist yet and then bet the entire farm at on that. Also in the context of Intel sliding on roadmaps consistently over last 3 years. The 'brillant' product management made gambling on the nonexistent a the cornerstone of the design. ( Not trying to design ourselves into corners .... just flakey products don't aid to that. ).
If that is the bozo product mangeent that Mac Pro has then folks should just quit waiting. I understand Apple wanting to engineer something relative new. But instead of asking Intel for power point slides, it should have been a request along the lines of "which processors in your line up in the workstation class can you guaranty you can hand us at least engineering sample of by the middle of 2018?". Essentially, what does Intel really have because we have a urgent demand that requires a timely solution. ( go to AMD and ask the same question. And tell both Intel/AMD that asking the same question to both. )
WWDC is too late. Pragmatically, Apple will have to show in April around that anniversary date they have now laid the groundwork for. Two reasons. Between April 2019 and June 2019 a giant hailstorm of hate is going to ran down on them. "The dog ate my homework wait until June for only a tease" is equally as screwed up as "We'll we screwed up and it is now 2020" would be. Both are a Lucy pull the football move. Many folks are going to be pissed by the end of 2018. The anniversary date is only going to unleash the 'hate' in a giant wave. Apple the vapourwave king is going to tech porn press headline for months by WWDC.
Two, they are going to need an extended session longer than WWDC can reasonably allow as to why this colossal screw up was in any way necessary. They could stuff it into a NDA "don't open" box, but at least it should be we have something running in a prototype stage.
The above and the below are not rationally consist plans. Those are two excuses for exceedingly poor planning and product management.
Actually it isn't. Intel knew about Meltdown/specture bugs in 2017. Where those fixes were going in had to relative well know by the end of 2017. That Intel was pushing out Ice Lake with at least Cascade Lake would have been known. And absolutely certainly before April 2018. By April 2018 Intel would have also had the notion that Cooper Lake was at least possibly going to be a second stop gap injected.
By April 2018 that plan would have been in the crapper if any real NDA communication was being down between Intel and Apple.
[doublepost=1533683897][/doublepost]
"they" == Intel. Pragmatically, yes through relatively recent leaked roadmap slides.
"they" == Apple. No. Just the whole year 2019 as far as granularity goes.
If Apple made some sane choices looking at Intel's roadmap then Intel's said may be closely coupled to Apple. If Apple was a bit looney then the schedule could very well be on Mr. Toad's wild ride.
Last edited:
I think it will be 2020.
The likelihood of the execs using broken Apple keyboards is pretty fair. Statistically, it would be more likely for the broken keyboards to cause significant communication interference with critical inter-office memos using the 2019 date ( 4 different keyboard buttons ), than the 2020 date ( only 2 keyboard buttons ). Plus it has a level of symmetry that makes Jony Ive's silk thong uncomfortably moist.
The likelihood of the execs using broken Apple keyboards is pretty fair. Statistically, it would be more likely for the broken keyboards to cause significant communication interference with critical inter-office memos using the 2019 date ( 4 different keyboard buttons ), than the 2020 date ( only 2 keyboard buttons ). Plus it has a level of symmetry that makes Jony Ive's silk thong uncomfortably moist.
Charlie was calling about Intel problems with 10 nm process for past 2 years, and people were calling him hater, idiot, fanboy, shill, etc.
He was dead right from the ground up. What makes you think this time he is wrong, when he in previous article: https://semiaccurate.com/2018/07/17/amds-rome-is-going-to-be-a-monster-cpu/ even pointed out to specific configuration Rome will have?
Don't read clueless Bloggers.
It's their foundry manufacturing group that is sinking Intel down. It is directly because TMG is not able to push out 10 nm CPUs, Intel is going to struggle technologically for the next 2-4 years. If they will ditch all of their technology, what they will do, when they are going to be stuck on 14 nm process next couple of years, when everybody will be on 7 nm? Intel made good decisions to search for another markets to survive, till they can push out new architectures, and new process nodes.
Everything at this moment points to a situation, where Icelake architecture will actually compete with Zen 3 architecture.
And remember, we will see next year Zen 2 architecture. Zen 2 will be slightly faster per core than Skylake, but will offer double the cores, and much better efficiency, because of 7 nm process.
Matisse/Rome is not what People think it is. When you will see guys what AMD did on 7 nm CPU you will be amazed at brilliant technical and business genius, Lisa Su have decided to do.
Because this decision translated into efficiency of the design, and its simplicity, that allowed AMD to perfectly scale it.
Last edited:
I doubt a Q1 release. They were very forthcoming that the Mac Pro “would not be coming this year” with Apples track record of actually delivering (2013 Mac Pro, AirPods, HomePods, AirPower) I wouldn’t expect anything before WWDC, NAB if they really had the **** to say they were serious about Pro users at a Pro venue.
Some "Pro" dog and pony show is absolutely not necessary. When they have something reasonably done (and have some relatively solid production ship timelines ) they should do a demo. Waiting for a fixed in time dog-and-pony show for some collection of other products is looney toons for a product this late. This product is soooooo late it is a spectacle all by itself. They don't need any more spectacle than they already have.
If Apple's shovel all Macs into the June - October window philosophy have worked extremely well over the last 4 years, that would one thing. In many respects it hasn't. Log jamming more systems in smaller parts of the calendar year has turned into a highly constipated (inhibited ) product flow over last 4 years.
Shifting Mac Pros away from the iPhone launch , macOS beta window, and masOS initial (probably buggy) release would be a good thing. If Intel's and AMD new speed-bumped product line ups match up to that time shift window then all the more so. ( 2013 Mac Pro release was about same Quarter new Xeons shipped and Thunderbolt 2 was ramping up in volume. And tease was about time Intel was throughly leaking info on Xeon bump; just not officially released. Ramped volume for MP 2013 was pragmatically Q1 2014. )
June for Intel mobile based products made sense because Intel has typically done April-June releases in that space. The Xeon workstation stuff has been in Q4 ish time frames over last several years. So picking a following Q1 would tend to back accurate even in the context of an Intel slide for a quarter. The safe pick is a quarter or so after what Intel is shooting for. What Apple sorely needed was a safe pick; not some yippie-ki-yay flyer.
The AirPods and HomePods and AirPower were first generational products. In some sense, that can't be "late" because there were never offered before ( they had a promise but Apple had a "gee this is buck rogers stuff" excuse. Nobody else had done them either technically). The Mac Pro is about 180 degrees opposite of that. It is horribly late. Some substantive fraction of the angry mob with pitchforks wants them simply to ship the old, old machine with 'as minimal as possible" changes. So the whole "this is new and super fantastic Area 51 like technology" reality distortion field isn't going to fly.
To be blunt, Q4 2018 would have been painfully late. That they are talking 2019 is only indicative that they were screwing around engaged in deeply unproductive( or just plain no ) activities in 2015 and 2016. They already were on the clock. They've been on the clock since the MP 2013 was 2-3 years old. They absolutely are explicitly on the clock since April 2017; over a year ago. The Airpods , HomePod, and AirPower didn't come two-three years later.
Cherry on top is that they have working iMac Pro. A working Xeon W and current top end AMD GPU can't possibly be some deep, dark mystery to getting a macOS up and running on.
There’s no requirement for a dog and pony showing... it is interesting that both the initial mea culpa and follow up we’re both in April.
Let’s be honest Apple has f***** up in the pro space. The MacBook Pro had a press release “release”, and there is little precedence for Apple to do much. They did preview the iMac Pro, but this Apple is something else.
[doublepost=1533699774][/doublepost]I think we’ll be lucky to see something by NAB or WWDC, more than likely it’ll be closer to 2020. But Tim/Craig/Phil can try to prove me wrong.
There probably is some coupling to chatter (grumbling ) going on at/before/after NAB in Las Vegas which typically happens in early April. However, explicitly highlighting the show itself only tends to tends to enhance the echo chamber that the Mac Pro is only good for video ( Final Cut) and Audio ( Logic X). That is the scope of its imprint. Which is a problem. Those two apps needs far more than the Mac Pro in scope. And vice versa, the Mac Pro needs more than just two apps to reasonably exist.
Apple product releases get covered by the press whether there is a dog-and-pony spectacle put on or not. Just press release makes it sound like there is simply a quarter page insert in the coupons insert in the Sunday paper. That is actually not the scope at all of a the releases.
Directed ads and eventually Apple store (and 3rd party retail ) placement will put the product out there for hands on. The days where had to be in some conference hall and a booth babe to show you the way to the demo table is rather antiquated at this point.
Apple is likely to "preview" / tesase this one also, because have already done a bunch of pre 'nonannouncing' (when it won't be).. So there will probably be a lead of about a quarter or so of "ok it is coming" since it is hardly completely a secret.
Apple could possibly wait until the end of 2019 and perhaps slide into 2020. But would the Mac Pro be long term viable at that point? They would have spent 2+ years creating an initial demand bubble. Perhaps initial shortages again as with the MP 2013 and then what 8-12 months down the road? A big bubble somewhat follow by a significant bust would probably put Apple right back into "Rip van Winkle' mode.
By the end of 2018 the MP 2013 would have hit the 1800 day old mark. creeping deeper into 2019 gets that every closer to the 2,000 mark which many will have a hayday out of . By the end of 2019 Apple will hit around the 3,200 day mark from the 2010 release. Is that really going to attract customers that have 5-6 year cycles (or shorter ) on system buys?
The real surprise really should be that that executive crew actually thought dragging their feet until the end of 2019 and perhaps sliding into 2020 was a good idea.
@decontruct60
I agree with quite possibly all you said. I think the “final cut appliance” is a joke and the Mac Pro should be something that transcends various fields that require a workhorse. I am in architecture and want a machine that can render models with little downtime. Right now a hi res cover page in Revit (parallels or boot camp) can take easily 20-30 minutes I want something with some serious GPU and CPU. There are other fields who would like the same thing or at least competitive to what’s out there today from the competition.
In my honest opinion there should be no one else’s timetable (but this is Apple we are talking about-the precedents have been set by them; I think the mastered the tech keynote and it has significantly run its course, the last townhall/events have been lackluster, not only in what they were actually releasing but the presenters themselves). The longer Apple goes silent about this the more blowback they will get when they release something crappy. Let’s be honest if it’s another cube or tube, everyone will see the writing on the wall. I would love to see something that inspires future pro hardware development but my expectations are low.
I agree with quite possibly all you said. I think the “final cut appliance” is a joke and the Mac Pro should be something that transcends various fields that require a workhorse. I am in architecture and want a machine that can render models with little downtime. Right now a hi res cover page in Revit (parallels or boot camp) can take easily 20-30 minutes I want something with some serious GPU and CPU. There are other fields who would like the same thing or at least competitive to what’s out there today from the competition.
In my honest opinion there should be no one else’s timetable (but this is Apple we are talking about-the precedents have been set by them; I think the mastered the tech keynote and it has significantly run its course, the last townhall/events have been lackluster, not only in what they were actually releasing but the presenters themselves). The longer Apple goes silent about this the more blowback they will get when they release something crappy. Let’s be honest if it’s another cube or tube, everyone will see the writing on the wall. I would love to see something that inspires future pro hardware development but my expectations are low.
Charlie always loved to spin up doom and gloom scenarios for Nvidia and Intel before. Intel is definitely having some trouble with their 10nm and their future process but lets not act like sky is falling down on them lol. Intel is most likely counter Rome with MCM solution and will not have problem competing in terms of core count. Intel will no longer be able to retain their sweet margins on their Xeon line of processors anymore but they are more than capable of competing in terms of performance. Intel's 10nm is comparable to other foundaries 7nm high performance node. (First EPYC 2 with TSMC is going to fabbed on 7nm lp) Assuming if they hit their target of 2019-2020, they shouldn't have much trouble competing against AMD.
Real question would be if Intel will have same kind of problem with 7nm process. if their 7nm EUV (comparable to other foundaries 5nm) comes as scheduled, then they will not have any problem. If not, then they will have some real trouble in the future (assuming other foundaries next node comes without any delays)
Yeah.
350W Cooper Lake Triple die CPU, with low core clocks, that will cost eye watering 40-50k$ will be perfectly capable of competing with 64 Core, 180W-250W TDP CPU that will cost around 10000$.
Cascade Lake is just a refresh of Skylake-SP. 28 cores, with higher TDPs(200-250W), the same core configs. And that will cost 20000$.
Lets say that 64 core/128 Thread, 180W TDP CPU will cost 10000$.
How do you see Intel offering competitive?
AMD will destroy Intel on price/performance/power factor.
In all of the above by a factor of 2X!
First, I didn't say anything about him being wrong about 10nm. I said he sometimes goes too far. He takes nuggets of truth and then starts to wrap histrionics around them. That Intel had/has a problem 10nm. After year one, that wasn't some super unique insight. That they have a problem isn't the real issue with his over-the-top commentary. It is that this specific problem is going implode or decimate the company somehow. That's what is lacking. It was surprising that Intel let this fester for so long.
Second, as for wrong now... go back and re-read my response. Everything he screwed up I went through point by point.
As for his article about AMD.... actually that is illustrative of why isn't to be solely trusted about the doom of up coming processor products. To quote:
"
... If AMD stuck with it’s 2016 roadmaps, it would be in for a big fight with Intel next year. ...
....
This is the long way of saying two things, one we got the information right two years ago but that information changed and it is wrong now. ..."
'we were right but it s wrong'. Carefully claiming correctness and throwing wrong into misdirection. That's the problem with his narratives. If he is right about something he will mention it over and over and over again into every article he can stuff it into. And when not really on it is mainly about misdirection. I get it ... these open articles are sales pitches for his "paid for" content. But sales pitches aren't objective analysis.
Roadmaps change over time. Intel's has changed as had AMD's. His doom and gloom is wrapped up in Intel being rigidly stuck with their 2015-2016 roadmap. They aren't. Nor are they 100% stuck with their 2017-2018 pricing.
Intel can probably make stuff at 10nm. Part of Intel's problem is that they are hooked on super fat margins like a crackhead. The problem is more likely that they can't make relatively large dies at super fat margins at "normal" prices with 10nm. They aren't going to completely junk their TMG and start over. They do have to wait though to get the 7nm gear ( which for some of their equipment is upgrades..... but can't go down to parts bin at Fry's and buy these. They have multiple year lead times to buy and deploy. )
Intel bet somewhat wrong on this
"..The crux of the matter is that EUV would shorten time to market and arguably make the process easier (if only more expensive), and several fab companies are waiting for Intel to jump onto it first. With EUV not ready, Intel has had to invest into deeper multi-patterning techniques, which raise costs, decrease yields, and increase wafer process times considerably. ..."
https://www.anandtech.com/show/11722/intel-reveals-ice-lake-core-architecture-10nm-plus
The 'clever' multiple pattern stuff has been tougher than they thought. The "spend' is real.
Intel's diversification moves have been slow on payback and extremely expensive. The basic ideas were OK but the execution has not been stellar. Some stuff like McAffe was pretty bad. ( bought for $7B in 2010 and spun out into a joint venture in 2016 for revaluted $4B . That was $3B down the drain... enough to build a modest foundry for 3rd party customers. ). But not devastating losses (every investment can't be a winner.).
Infineon (cell radios ), Altera , and Mobileeeye have all been blockbuster acquisitions that haven't been huge home runs. They've often been relatively slow in getting those profitably merged into their own TMG fabs. ( Nervanna isn't quite blockbuster in cash outlay, but that too isn't deep integrated and creeping toward 3 years in. Not late yet, but certainly not early. )
Yes, but is Ice Lake the same Ice Lake it was back in 2015 ?
Feb 2017:
" .. On speaking with Diane Bryant, the 'data center gets new nodes first' is going to be achieved by using multiple small dies on a single package. But rather than use a multi-chip package as in previous multi-core products, Intel will be using EMIB as demonstrated at ISSCC: an MCP/2.5D interposer-like design with an Embedded Multi-Die Interconnect Bridge (EMIB). ... "
https://www.anandtech.com/show/1111...n-core-on-14nm-data-center-first-to-new-nodes
Intel was shifting away from monolithic dies also. The data center stuff was suppose to be first to 10nm++ ( but it looks like 10nm++ is probably going to be the real first, high volume iteration. )
Zen 2 has yet to ship in high volume. It probably will scale, but at what yield rates? Intel has a problem, but the notion that AMD is 100% problem free needs alot more completed phases to it.
Last edited:
As I said, Intel won't be able to retain their margins on ther Xeon processors but they will be able to compete on performance, and price if necessary.
Lets not throw spec numbers and pricing on unreleased products as if those are facts.
We will find out in 2019 if Zen 2 based EPYC comes as planned, as well as Intel offerings by then.
You know perfectly why Charlie is "wrong"
.Its because he writes about stuff 12-24 months before they happen, and a lot changes on the road. Why he changed his point of view on AMD?
Because in 2016 there was one design, that would happen from AMD: 16C/32T CPU, and EPYC 2 would be made from 4 dies plus something additional, for IO, on the package. Current design for Matisse, and Rome are MUCH different.
If they could make 10 nm products, we would already see them, in massive quantities
.Here is good article on 10 nm Launch Intel did earlier:
https://semiaccurate.com/2018/05/29/is-intels-upcoming-10nm-launch-real-or-a-pr-stunt/
The problem with 10 nm process is not margins.
10 nm process essentially is not yielding at all to the point where Intel would LOSE MONEY on each wafer, if they would sell products made on it, and they would not be performance competitive with their own 14 nm++ and 14nm+++ process, because of lower clock speeds.
Of course not. Intel changed a lot in Ice Lake architecture, while it was, and they wanted to do something different with the architecture(15-20% IPC increase with lower clock speeds).
BUT, 10 nm process, or rather 12 nm process as it should be called right now, per:
https://semiaccurate.com/2018/08/02/intel-guts-10nm-to-get-it-out-the-door/ will bring performance and density decrease from initial performance/density targets, for 10 nm process.
At what % will AMD get 120 mm2 die on 7 nm process, that is pretty mature right now?
You know perfectly why Charlie is "wrong"
Its because he writes about stuff 12-24 months before they happen, and a lot changes on the road. Why he changed his point of view on AMD?
Because in 2016 there was one design, that would happen from AMD: 16C/32T CPU, and EPYC 2 would be made from 4 dies plus something additional, for IO, on the package. Current design for Matisse, and Rome are MUCH different.
If they could make 10 nm products, we would already see them, in massive quantities
Here is good article on 10 nm Launch Intel did earlier:
https://semiaccurate.com/2018/05/29/is-intels-upcoming-10nm-launch-real-or-a-pr-stunt/
The problem with 10 nm process is not margins.
10 nm process essentially is not yielding at all to the point where Intel would LOSE MONEY on each wafer, if they would sell products made on it, and they would not be performance competitive with their own 14 nm++ and 14nm+++ process, because of lower clock speeds.
Of course not. Intel changed a lot in Ice Lake architecture, while it was, and they wanted to do something different with the architecture(15-20% IPC increase with lower clock speeds).
BUT, 10 nm process, or rather 12 nm process as it should be called right now, per:
https://semiaccurate.com/2018/08/02/intel-guts-10nm-to-get-it-out-the-door/ will bring performance and density decrease from initial performance/density targets, for 10 nm process.
At what % will AMD get 120 mm2 die on 7 nm process, that is pretty mature right now?
Christ, you are basing all your arguments on Charlie's articles? There is a reason his website is called Semiaccurate
But I guess it doesn't matter as long as it says whatever you want to hear right?
No. Its because I payed him the money for the articles, and I can see that what he was writing in them was 90% correct, and just the cosmetics were not correct.Christ, you are basing all your arguments on Charlie's articles? There is a reason his website is called Semiaccurate
But I guess it doesn't matter as long as it says whatever you want to hear right?
These are both written in 2018. They both smack of "Monday Morning" quarterbacking. There are huge issues they are missing.
First, Intel has done a rather peculiar track on 'get more fab customers" for their manufacturing..... they bought them.
Infineon ( $1.4B) acquired in 2011 started produce celluar radios on internal fab in 2018 . ( Gasse's article has a lament about Intel missing out of the iPhone CPU jobs originally. Well Intel has finally looped to having the next iPhones celluar radio). This step misfired a bit. Infineon thought LTE and especially CDMA weren't worth high priorities, so right as Intel bought them, Apple disappeared as their highest revenue customer and the European headset makers (their other deep base of customers ) somewhat got streamrolled by Apple. So a double pothole.
Altera ( $16.75B ) acquired in 2015 and I think they were already doing some FPGA at time of acquisition. 2013:
" Altera also will work with Intel as a fab partner at the 14-nm process technology node, while continuing to use TSMC at 20 nm. "
https://www.electronicdesign.com/fpgas/altera-makes-deals-shrink-real-estate-and-reduce-power
MobileEye ( $15B ) acquired in 2018 nothing so far but 4-5 years down the road ....
If Intel can shift the bulk of those ( and the other acquisitions ) largely onto Intel's Fab then there would be lots of work to keep the Fabs busy. Similarly getting back into non-volatile memory could be a more work on some slightly different fab methods. ( dogma of never, ever do any memory again is silly. )
The question is more of whether Intel has used the money wisely than of whether "Integration" is someone inherently evil ( ... it is not. That too is rigid dogma, more than truth. ).
Second, speaking of using money wisely .... the parts about focus on margin rather than product is on track ( but going fabless or becoming a pure fabless would not necessarily change that myopia. )
One of Intel's core problem was not so much that the Data center group was peaking out so much as several of the other products were stumbling. They tried to buy their way into mobile as opposed to defending closer to the end of their "territory". Tablets and Chromebooks it is mostly working.
Third , the that first article is whacked about why folks are going fabless. This is primarily driven because modern Fabs cost billions and the price keep going up as the nm resolutions go down. That means needs a collection of products to generate the revenue to cover the costs. That can be done by collecting a large number of low revenue customers together to share the same factory. Or it can be done by running alot of higher revenue that have sustained volume products products through. It is dropping down to where everybody has to put their stuff through 2-4 companies. If that shrinks to 1-2 that probably won't be a good thing long term. Intel has the right product mix to be one of the last 4 standing. ( chucking the processor group relatively would be a dual edged sword. It wouldn't be necessarily 100% better. )
Krzanich was promoted out the manufacturing group. He should have been someone who could spot folks in manufacturing blowing smoke and putting on the "Emperor's New Clothes" tales. He was the one to started the "fab for others" business
"... In less than an hour, he took a decision that would pave the way toward a new business for Intel: Achronix became its first foundry customer. This past February, Altera Corp became its first major client for the business, and industry insiders believe it could eventually reach a similar deal with Apple. ... "
https://www.reuters.com/article/us-...brian-krzanich-at-intel-idUSBRE9420WG20130503
The problem has been that has stalled since 2010. Largely because folks have problems jumping on board with the process. 14nm somewhat stumbled out of the gate before 10nm completely stalled. The core problem wasn't whether fabbing for others would be a problem or not. The problem was whether they had a fab process that others highly wanted. ( it probably would have helped to do it more so in a siloed, subsidiary where folks were not as fearful of showing Intel the crown jewels of competitors to other Intel products. )
I suspect Kraznich got show the door as much for this as much as the "affair" stuff. The blame for the pain over the next 2 years is going to get heaped on him and he was a convenient candidate.
Intel has the base renevue stream to stay solidly in the fab business. They need some supplementary stuff ( and probably cheaper not to acquire all of your supplementary customers. ) going forward . Pushing Intel's processor core business to a smaller number of available fabs vendors could backfire for alot more folks than just Intel. ( e.g, once get to a single point of failure a single vendors process hiccup could hit the whole market; not just a couple of vendors. )
Cosmetics? Chuckle. His articles doom and gloom about how SP Skylake is grossly underperformance what Broadwell did. How that stalled, weakened line up and that 10nm was just going to put it down from there.......
And to look out for Intel's meeting this week. Well OK. ( from anandtech's live blog)
https://www.anandtech.com/show/13192/intel-datacenter-day-2018-live-blog
A. Fastest ramp since Ivy Bridge. Sure if you only looked at the 4th from the right bar on that graph Skylake is low, But that's because they have all been low initially. Data center folks tend to buy conservatively and at a pace.
B. A lead in virtualization is also a factor for big multiple client cloud shops.
C. They have had higher growth in the high core count options ( So $8-10K was painful for folks but a substantive number saw the value. ).
D. A large chunk of the talk is about Intel's networking ( which AMD doesn't have ) , non-volatile memeory ( which AMD doesn't have) and AI solutions (which outside of throwing at GPU at it, AMD doesn't have. )
A couple of other highlights from the blog where he grossly missed the boat.
" ...
01:38PM EDT - Xeon-D was built out of joint innovation with facebook
01:38PM EDT - Nearly 30 SKUs are custom off-roadmap uniquely for the service provider
01:39PM EDT - They want to keep it off roadmap
01:39PM EDT - Amazon is a big customer
...."
Intel is building stuff that folks having given them spec guidelines for. When vendors go out of their way to help you, most folks won't quickly, causally throw them under the bus.
Core counts aren't everything....
"..
12:26PM EDT - Performance is 8x in certain use cases, such as SparkSQL compared to DRAM only systems
12:26PM EDT - Enable 9x read transactions, 11x users per systems
12:26PM EDT - Delivering the promises of the hardware..."
....
"...
02:39PM EDT - A huge part of SAP's business in having a database up and running for analytics - with growing of dataset sizes means that loading datasets is a key bottleneck
02:39PM EDT - Faster startup times by 12.5x
02:40PM EDT - >3TB memory per CPU
..."
AMD is going to match 3TB of direct attached to CPU storage at the same price? Nothing in his histrionics indicates that. More cores isn't alot of traction if more extremely low latency storage is what you need. There are spots where Intel will backslide, but they probably won't backslide in market share everywhere. So gross, wide scale, imminent collapse is "arm flapping".
" ...01:40PM EDT - Can also make custom ASICs for customers .."
...
01:43PM EDT - Intel talks a lot about software for a hardware company
01:43PM EDT - It's the key to creating new services
01:44PM EDT - 200 engineers are engaged with cloud service providers for hands-on side-by-side engineering
01:44PM EDT - Already working on 150 cloud service projects this year already ..."
( I think the "talks a lot" part above is the blogger, not Intel. But is a salient point completely lost in Charlie's hype machine. )
Core counts aren't going to matter if have fixed function circuits to do custom algorithms faster.
( that isn't particularly relevant for single user workstation that the Mac Pro primarily address. Apple doesn't necessarily need a Xeon with a FPGA/ASIC in it. )
We're talking about Apple, so seems like things check out.
[doublepost=1533765248][/doublepost]
I give 75/25 odds that they'll wait until WWDC.
Cause it's not going to ship until the end of the year anyway. They'll put a few units out at the end of 2019 to claim 2019 ship, and ship in bulk in 2020.
I know people won't like that, but if you don't think that's what's going to happen, then you haven't been paying attention.
There's also AMD floating around somewhere out there. Maybe they still end up using AMD. I'd guess Intel in since the date is so fuzzy. If they were using AMD they'd probably be shipping this year. Or last year.
I also still think it's likely that Apple is working on something around upgradable graphics, dragged Intel into it, and that's locking them to some future Intel chipset or architecture.
This is Charlie' Twitter:
https://twitter.com/CDemerjian
Read his feed, because he was at Intel's conference, and look what he has to say about Intel, then read his latest articles.
Intel is lying, and trying to spin things out so that analysts cannot see how murky waters are(THE LAKES!), that things are really bad.
And yes - they are. But that you will find out, only in his articles, because he has solid data about the technology behind 10 nm process.
P.S.
What If Apple cannot give you Mac Pro, because of Intel's 10nm fiasco? What if Apple wanted to do something interesting, but lack of tech from Intel does not allow them to do so?
What Lack of tech? Workstation design should be flexible enough to update to whatever new tech that comes out every year. Be it PCIE 4.0, New thunderbolt, etc, etc.
If Apple can't release Mac Pro because of Intel's 10nm fiasco, then might as well not come out because that means it might be another MacTrashCan 2.0, putting themselves in thermal corner again.
Look up "TPM".
That would simply mean that Apple is stupid. (Although not much about the MP6,1 or MP7,1 would argue against the point that "Apple is stupid".)
The primary design goal for the MP7,1 should have been to be "as quick to market as possible". Apple has certainly ignored that one.
HP/HPE/Dell/Lenovo/SuperMicro rush to get the latest tech from Intel out in systems.
Apple seems to only move when the components that they buy hit "end-of-life" and can no longer be purchased.
There is nothing radically different about IceLake other than it is on different node potentially offering more core count, extra memory channels and extra PCI lanes. If they are going to offer Mac Pro with enough flexibility for expansion and enough power to drive it, it shouldn't matter if it is based on Cooper Lake or IceLake, or AMD Rome if they decide to go that route.