The takeaway I got from it on first listen:
- The nMP was positioned as being a way for Apple to become and stay on the cutting edge. It failed to do that.
It or Apple failed. If go to your last point ( pulled up below to bring into context ).
All the strategic things Apple said the new machine was about, they more or less failed to follow through on, and professionals making investments in a capital intensive production pipeline need that long-term strategic commitment.
That failure to follow through is more a disruption point than the "version 1" hardware design. Being "Version 1" it was probably going to need an iteration or two work out some relatively minor issues. ( MBA changed designs a bit before it took off in the new price category that Apple put it in. )
Apple has a compute GPU and didn't have one of the best OpenCL environments. i know some/most of this is subcontracted out, but failing on this dings Apple more than AMD. [ Yeah, Metal popped up but Apple really needs to figure out how to walk and chew gum at the same time. Being penny wise and pound foolish is going to kill the company over the long term. ]
- GPU advances are happening faster than Apple wants to update the entire machine.
CPU advances are also happening faster than Apple wants to update the entire machine. It isn't just GPUs. Apple is skipping generations inside of Xeon E5 class CPUs to jump from "tick" to "tick" ( even version numbers). That's goofy. Throwing GPUs on top is even more goofy.
I don't think this as anything to do with the new design. Apple let the 2010-2012 go into abandoned active development state to the point product withdraw from EU market because could not keep up with a
2008 design change issue. 2012 and couldn't figure out a new constraint given in 2007-2008 time frame. That has nothing to do with these particular design choices at all. That has to do with R&D commitment and attention to detail. Period.
So in some sense not entirely surprising machine went into Rip Van Winkle mode again because Apple had already done that earlier in the decade. Apple had a chance to break the pattern if they had committed to updates, but they did not.
- Expecting users to buy an entire machine to get new GPUs, when there's been no real progress on anything else is unreasonable.
Again it is unclear what Apple's "Plan" here is. If it is to only pop up ever 3-4 years with something new then "whole new machine" is reasonable. Customers all have to sync up to the fixed upgrade cycle windows but if those are the only customers left ......
None of the big iron vendors drop new systems every year either.
- The "focus" in the new machine has not produced benefits in performance, has not allowed Apple to keep the design current, has resulted in a machine which has less practical flexibility, which costs more, and is less capable for prolonged heavy workloads than the previous design
The design itself hasn't stopped ( "has not allowed") Apple. Far more likely, Apple is stopping Apple. Waiting on limited time slices from Ive's design team would have little to do with design itself and far more due to the people , internal politics, and resources assigned.
Maybe not - but even a "made for mac pro" mac-edition would be better than "buy a whole new computer". The great tragedy, is they called it all about graphics, put moribund GPUs in it, and then made the only expansion bus one which specifically excluded GPUs from the supported devices.
I do think they either planned the upgrade cycle poorly ( some upgrade plan got abandoned/disrupted ) or poorly designed if intended to sit and squat for 3-4 years at a time. For that second case, they need do need some interim GPUs and to resynchronize to the "tick tock" schedule for the Xeon E5.
The (upcoming?) TB3 version of the nMP is what really should have been the version 1 product.
Perfect world. They would have done one more rev on last design 2012-2013 using Xeon E5 v1 and v2. Then could have jumped on board with E5 v3 with TBv2 in early-mid 2014 (with some built up product inventory of old model to smooth out the shock ). Part of the new designs problem was waiting for TBv2. Waiting on TBv3 would have been loopy..... if TBv2 doesn't survive and grow then will never get to TBv3. ( How many desktop FW1300/3200 deployment are there.... practically none. That is partially because of lack of traction by FW800. ).
With an updated Xeon E5 v4 + TBv2 design they could "wait out" the arrive Xeon E5 v5 which is a far better match to TBv3 ( and a couple pairs of TBv3 sockets ). I agree that more of the potential of the current design principles won't be easily evident until get the infrastructure up to v5, TBv3 , and a larger set of the software base catches up (via OpenCL and/or Metal) .
Apple doesn't like to talk about future products. That only works well over a span of 2+ years when can talk about current products. What they far more needed to do to hold many of the pro users leaving at this point is commitment to actually doing something on a regular basis. Not "incrementally innovate a couple of times a decade" level of effort. More "do" less "talk" ( if don't do anything then are not innovating. So yeah in the "can't innovate" status. )
