Who said the servers need to be off site? What do you think the 10 Gb ethernet port will be used for in iMac Pro? See what Aiden has setup. Local servers. Aiden, are the workstations as powerful as the servers? Naive? Hardly - open minded to other solutions than a big box of full of graphics cards under your desk. If iOS development needs extensive compute power, Apple will resease a Mac pro or server doing just that.
I still do not get why Apple should compete with HP for general purpose workstations. They have not succeeded particularly well the last two decades. You have not provided any arguments about this.
Do you say the current hardware is suffient for iOS app development unless it is centered around ML, AR, VR?
I personally use 10GBe with an slave compute server (a Xeon-D with an single nvidia GP100) in my operation I do a lot of double precision compute on this host, I'd like to not need a black box running linux along my iMac (non-pro 1st gen 5K iMac with TB2 10G adapter), I do this thanks to nVidia and intel specific cross-platform remote toolchain, of course I'm not programming anything to run on iOS impossible today nothing in Xcode (ML,AR) allows such setup (remote cross-platform debugging.compile, etc), not to remember you Adobe Premeire dont do Video Edition on external compute servers, and a long etc.
Aiden was specific to their production server, not they R&D workstations, all we know what he uses.
The Problem is, Apple Needs to release a product for iOS development well before it is needed otherwise iOS developers will be for ever on Android developer's Mirrors, AR/ML,VR trends where so obvious even when Apple launched Siri (one of the first widely available AI developments).
Luckly Siri was developed 100% on linux then integrated into iOS (was intended to be multi-platform before Apple purchased Siri creators), this is not an easy toolchain for AI, neither Core ML (aka ML-kit) foresee such toolchain, it requires to run ML training on the Mac.
[doublepost=1505819601][/doublepost]
People keep suggesting this but it isn't any quicker. The internals would have to be redesigned for current chipsets, and all it does is distract from the next Mac Pro. You'd be talking at least a year to restart the classic Mac Pro design, especially once you include time to get the factories churning them out again.
It really doesn't make a lick of sense, especially cause you'd have to toss the Thunderbolt ports that Apple has already started the transition to.
And if you solve how to put Thunderbolt ports in a Mac Pro tower... well you've already solved the next Mac Pro then.
The production lines for the classic Mac Pro were dismantled long ago, along with the classic Mac Pro engineering team. It's not like just flicking back on a light switch.
I mean, a classic Tower-like design as the Cheese-grater, NOT to to force new Motherboards on the legacy cheese-grater chassis, its much quickier to release a cheese-grater like solution since its more flexible and quicker to develop than a Trash-can like, even dont scare about Motherboard design, it its almost automated an all new motherboard based on Intel or AMD reference designs cost less than 50K for a dozen or more prototype cards, Apple could commision the chasis to one of thousands chinese factories already doing that, a cheese-grater like wont include 5.1/5 bays maybe even not spinners bays, always will be quicker and cheaper than a trashcan-like design, or the hybrid tc-chesee-grater designs which requires more tricky R&D about thermals, form and function etc, Apple now has almost 6 months on that, most pc Manufacturers comes to the market quicker than that starting from zero.
