AMD has not been a trustworthy CPU supplier - and needs to rebuild customer confidence that AMD can consistently innovate and deliver. Intel had been a trustworthy supplier - although there have been some hiccups on timetables.
Over last 2-3 years Intel has hiccuped about as badly as AMD 4-6 years ago. But AMD has hiccuped on GPUs ( Polaris was slow in coming and Vega was in the same boat on timeliness). AMD is getting those produced at quantity once shipped, but they haven't been the "we are delivering early" vendor for Mac graphics.
So it is more than just hiccups. Frankly, Apple's delays on refactoring the Mac Pro make both of their hiccups look tame.
AMD's current and upcoming core count advantages are mostly meaningless - since in the workstation space most applications struggle to utilize more than a handful of cores. (Obviously, some apps do use many-core effectively - but only people depending on those particular apps will care about high core count. For the rest, Intel's better per thread performance is an advantage.)
The focus on the mega die combo package and ultra sized socket solutions is probably missing the point for the Mac Pro ( and rest of Mac line up). If AMD can stay a process node update ahead of Intel then the single threaded advantage is going to slide backwards a bit. if it is just simply math metrics with mundane branching, then just cranking the clocks a bit higher with incrementally improved branch predicators will have a substantive effect. AMD will simply throw more transistors and clock at it.
So if they make their 6-18 core option to be within 1-3% of Intel and it is $100-200 cheaper and do custom work for Apple and it doesn't look like AMD is going to shoot themselves in the foot again, then Apple may bite. That's a number of "ands". It is 'safer' ( known quantity) for Apple to ride with Intel for a while longer.
Switching to ARM desktops and high end laptops would be a disaster. What if pro app builders (like Adobe) simply say "we're not going to do another architecture port" and drop Apple as a platform?
the bulk of the Intel CPU+GPU packages that Apple buys though are in the lower-mid laptop space though. ( even the entry level Mac Mini. and 'edu' iMac 21.5 ).
If Apple dumped Intel (and AMD) for all of those then probably would change the dynamic for Apple using AMD for the "rest" of the Mac line up. I think it would be a very bad idea for Apple to split the Mac line up on architecture. half-in , half-out would likely cause more problems than helps. Apple might be able to goose some extra Scrooge McDuck money pit money out of it is just trim off the lowest fringe of Macs, but that's just largely stock option/grant gyrations.
However, going Rip Van Winkle for 5 years is kind of goofy too.. If the Intel bulk discount isn't there then AMD's better pricing would play a bigger factor. Plus it is lower volume of chip needed for the remaining x86 line up so AMD's smaller capacity volume is also less of a factor.
[doublepost=1533774806][/doublepost]
Evidence that Nvidia is working against Apple?
1. Possible burnt bridge.
https://techcrunch.com/2014/09/04/n...-samsung-galaxy-devices-blocked-from-the-u-s/
Apple has been planned to jump into GPU market with both feet. At this point they have and are shipping large volume. If Nvidia has sent lawyers sniffing around Apple about infringement and Apple's need to give them a stream of the iPhone revenue, then they probably burned a bridge with dynamite.
This happened in 2014 and
2013 is last time saw Nvidia GPUs in a Mac. None. Coincidence... could be. Maybe not. How man Qualcomm raidos are going into this years iPhones now that Apple and Qualcomm are trading legal blows ? Is Intel's radio insanely great better? No.
Could Apple and Nvidia have worked out some patent detente by 2019? Sure but it would take two to tango in that process. The Nvidia fanboys who blame the whole absence on Apple would be missing the bigger picture.
2. Nvidia kneecapped OpenCL.
AMD, Intel , Imagination Tech all had OpenCL 1.2 drivers long before Nvidia. Nvidia
was "prepping" OpenCL 2.0 support in 2017 ( OpenCL 2.0 got
released in 2013 ) Fours years later .... that's like Apple Mac Pro like development cycle time. If ranting about Apple about being slow is harmful to customers , that kind of delay is in practically the same boat.
if Samsung was 4 years late supplying new OLED they'd be dropped as an Apple supplier. Not materially different in being grossly late as a software supplier to Apple. Being timely on CUDA isn't a value add as an Apple subcomponent supplier.
Nvidia could have done CUDA and OpenCL (one didn't have to 'loose' for the other to 'win'). About the same time Nvidia got 'sue folks' happy they also made moves to create bigger moat around CUDA. Slow rolling OpenCL. Chopping "open compute" parts out of Portland Group Stack after acquisition. Yes that is they right to get better profits for their own company. However, that isn't making them a trustworthy Apple subcontractor.
Apple released Metal in 2014 ( another coincidence on timing?) . Probably was in development before then. There is some unclear chicken-and-egg thing here. Did Apple start telling the GPU vendors they were bolting away from OpenCL ( so Nvidia felt justified in back burnering it as a Apple contractor) or was Apple nudged into Metal in part because the "committee' shepherding OpenCL was going off track from their perspective. I don't 100% blame just on one party here, but I highly doubt Nvidia is pristine clean here.
[ Apple has similar sore point. They shouldn't drop all open compute framework. ( or at least make hooks for them to plug in cleanly). Similarly, not having a empty, secondary x16 standard slot so folks could optionally plug in something for CUDA is shortsighted. Metal should have to compete harder on a mostly level playing field.
P.S. external PCI-e enclosures ( ePGUS) pragmatically means just about all Macs in the future (those with TBv3 that form the large base at that point) can have end empty secondary x16 (physical) slot added to them. It isn't like another GPU is going to extra extremely rare. Not mainstream, but they will be around. Keeping a open secondary slot out of a Mac Pro isn't going to 'help' much if have reasonable requirements for its size. ]
