It's possible that Apple Engineers could write some NVIDIA drivers for the Mac
I think there's 2 reasons why Apple are going with AMD this time for GPU.
1. Better compute performance.
2. Excellent discounts from AMD.
It's not in Apple's interests to see AMD go down the plug hole as Nvidia would have an effective monopoly on GPU. A reasonably prosperous AMD also allows Apple to leverage Intel for discounts on what I presume to be exclusive use of Intel CPUs.
There's already rumours going around in other threads about Intel losing interest in further development of Iris Pro units with doubts over Kaby Lake HQ series processors which would go into the Macbook Pro 15".
As it is, we would in theory have a very long lifetime for the Skylake HQ Iris Pro chips destined for the Macbook Pro 15" and Apple appear to be fighting a losing battle to get Intel to reserve ever more space on the ever shrinking die for GPU by the time Cannon Lake comes around.
Instead, Intel would appear to be concentrating on squeezing more CPU performance (and cores) into Coffee Lake and Cannon Lake preparation for the 2017 release of the AMD Zen CPU which is reputed to have very strong on board GPU thanks to Vega technology.
So now Apple (and Intel) have a problem:
Vendors appear to be increasingly abandoning the quad core Iris Pro chips in return for a lower end CPU plus discrete GPU for added performance. When you're spending loads of money, people are increasingly expecting graphics performance. An
i7-6700HQ with HD Graphics 530 is almost $70 cheaper than an
i7-6770HQ (as seen in the now infamous - but niche market -
Intel Skull Canyon NUC). That $70 can be put towards an AMD Polaris GPU which would perform better and the 15" Macbook Pro case cooling solution is already designed for a GPU.
So Apple are only gunning for improved battery life by going without GPU on base model, so they offer better integrated budget graphics performance, and price tiering when they offer the base model 15" Macbook Pro.
Bear in mind that the oft quoted
i5-6350HQ could form the basis of an even lower end model for Apple if they so wanted to go that way.
So where does the Mac Mini come into this?
Well, I think Apple may want to consolidate the CPU lines they purchase from Intel to maintain a decent discount while simplifying their product lines.
We know that the Macbook Retina will use the Y series 5w Kaby Lake CPUs which could bring with it the option to use a genuine USB-C/Thunderbolt 3 interface, or even two!
At some point I expect the 4K iMac to go fully Skylake for the Thunderbolt 3 ports, which should hopefully mean the base model (based on an ancient Macbook Air) gets to be quietly killed off.
The same fate could have fallen on the lowest model Mac Mini which also uses the same generation MBA CPU.
But we also know that the
Macbook Air has been specifically name checked by Mark Gurman and the 15w U series Kaby Lake is due imminently too.
The U series Kaby Lake 28w Iris Graphics model comes next year. Apple could just bump the storage or discount prices on the existing Broadwell model to delay things a few months, or they'd have to go with the existing Skylake CPU with Iris Graphics -
i5-6267U for example - and add the Alpine Ridge Thunderbolt 3 chip if they pull the trigger on a full Thunderbolt 3 refresh across the range.
The HQ series 45w Iris Pro mobile chips by Intel may be slowing in development due to lack of interest from 3rd party vendors, or refocusing by Intel on bread and butter CPU performance after recent job cuts, or perhaps because Apple informed Intel that they had decided to return to discrete GPU where applicable because they were developing a 5K monitor or valuing compute power from discrete GPU more.
Either way, if Apple strike a deal with AMD for Polaris GPU and use Intel's cheaper quad core i5 mobile CPUs as entry level they may have a decent upgrade for all users of their 15" Macbook Pros at the possible expense of battery life.
So we get left with 4 increasingly 'way-out-there' options for the Mac Mini, which traditionally borrows from the Macbook Pro 13" (option 1 in this case):
1. Use Skylake i5/i7 dual core 28w CPU because the Retina Macbook Pro 13" is getting a refresh at the same time and the Kaby Lake version isn't due until next year with no indication of what kind of benchmarks to expect. No need for case redesign with or without Thunderbolt 3.
2. Use Kaby Lake i5/i7 15w CPU all round and call it the most power efficient desktop yet - possibly with a case redesign and with or without Thunderbolt 3. Loads of moaning from our power users pining for a quad mini.
3. Add a discrete AMD Polaris GPU to the Kaby Lake i5/i7 CPU for extra desktop powered compute on a Mac Mini because there's a new Apple 5k monitor in the offing and they want users to be able to connect directly to it. This will need a case redesign which might not be in Apple's budget but the dual core segregation keeps the Mini away from cannibalising any other Mac. Phil Schiller can hand wave away the reduced Geekbench of option 2 by offering up superior compute performance for serious users (or gamers!).
4. The Macbook Retina 15" gets a refresh with Skylake Iris Pro, a chip which may be the end of the line for Iris Pro and therefore could sit there for 2 more years - perfect for the Mini especially if there's an i5-6350HQ quad core with no hyper threading option. Not sure the Mac Mini sales volume will help secure a big enough additional discount and such a Mini would still not be able to drive a 5k display without an external GPU. It's only going to please some folks on this forum though.
Sorry, I wandered into the realms of fantasy there
