Ports generally want to be close to the motherboard; these are placed on the same side of the machine as the motherboard. There's nothing especially contrived about that.
Sure, but they still have a space claim, as well as many other subassemblies.
There's a lot that's not going to be really answerable until the nMP's start to be delivered and an independent does a tear-down.
More generally... are you suggesting they should have used a four-sided heat sink despite only having three items that needed to be attached to it, in order to maintain a more conventional shape? That's the opposite of form follows function.
Not really. My point is that most of the stuff that's being stuffed into the box are rectalinear objects, so "squares in squares" makes for a generally easier space claim effort than with 'squares in circles' or 'squares with triangles'. Its a question of how much of a packaging complexity headache do you really want, when there's not any particularly profound reason for the external case to be 6.6" in diameter versus 6.75" or 7.0", etc.
Even if we want to point to the deep draw metal case, there's no major engineering design constraint that says that 6.6" is golden and that 10" (or more) is impossible...or even hard.
Placing the GPU boards parallel to or perpendicular to the motherboard would have prevented the use of a unified heat sink,
So what? Just what is the specific engineering justification of a 'unified heat sink'? Don't be particularly surprised if the real reason is merely to reduce the product's parts count for less expensive manufacturing. And sure, that's a good thing, but the touch labor in comparison to Intel CPUs that cost $1000 each is utterly insignificant.
and the latter approach would have created considerable dead space within the machine.
Again, so what? Sure, smaller is generically
nicer, but given the shrink from the cheese grater, they could have easily been 25% larger and it still would have been a 'Wow its small' customer reaction.
Plus, given that we already have some indicators that the nMP's performance is constrained (IIRC, something with the parallelism in the GPUs? The thread was elsewhere on MacRumors), it sounds to me that there's either a Power Supply constraint and/or a Thermal Constraint present within this design. Once again, the question will come back to what their trade-off decisions were...personally, I don't like the engineering implications of a hard constraint on future capability growth --> it invariably becomes a cost driver.
None of these other devices have the requirements that lead Apple to the design of the new Mac Pro, namely exactly three large boards sporting components requiring serious cooling...
Sure and that would be all to write about ...
if there were utterly no other components that also had space claims which needed to be also packaged.
Insisting that Apple should have enforced ...
Sorry, not at all; I apoologize for creating that impression.
My point is that other than the history of the Cube fiasco, there's no precidence for Apple to have embarked on a 'cylinder' language.
If you look at their other products, they are frequently reminicient of the old Cube, but the wrinkle is that they've all avoided being _regular_ cubes (eg, all three dimensions the same).
FWIW, I find the "Tower Cubes" of the Airport to be very interesting in particular, as they do hint that there probably was another nMP design under consideration in this general form ... my guess would be that it included a few 2.5" and/or 3.5" bays in it, adding to the height. I'd not be surprised to hear this come out eventually in a biography in 5-10 years.
One might want to consider taking an 'outside view' approach here instead of trying to analyze all of the specifics of this product, look at what has occurred in previous similar cases...
Of course. The problem is that the past precident of the last time that Apple sold a 'small form factor' performance workstation ... was the Cube. Maybe the market has changed
enough to be different this time; time will tell.
In what fraction of cases do critics end up being right, and in what fraction is Apple ultimately vindicated?
Think outside of the box: given their current corporate divisions' performance, to what degree does Apple now have the luxury of not really having to care if they just constrained MP sales by functionaly abandoning segments such as the SMB 'independent creative' who needs MacPro performance but who lacks the business to rationally pay for a Fibre Channel Server down the hall?
Actually not if have to cool both the inside and outside of the triangle. the triangle will fit entirely in the circumference of the circular fan. ( i.e. the fan's circle is not bounded by the triangle. It is bound by the case that encloses the triangle. )
The same is true for any other geometry pairs too.
Not to mention that the trapezoid power supply, I/O board, and RAM DIMMs (all not members of the triangle) also need cooling.
And as I said above, we're ultimately going to have to wait for an independent third party teardown to really gain insight on just how profound or necessary this "unified triangle" feature really was - or wasn't.
total smoke that not even present at all in this design.
Correct: there's not been much "CUBE!" angst manifested, probably because it isn't cube shaped, plus there's been much more emphasis to pack in performance (particularly GPU) so that despite its inherent trade-off limitations, it is less likely that computational "horsepower" will be a shortcoming...this was a factor with the old Cube in that its hardware specfiics failed to be a progressional step up from the its PowerMac counterpart: what killed the Cube was in no small part that Apple was asking its poweruser-centric customers to pay MORE to get LESS. That's a backwards 'Value' paradigm, so it is no real mystery as to why the Cube failed in the marketplace.
Apple has doesn't have to dissipate anywhere near this amount of thermals in the rest of their portfolio either.
Is that statement with, or without, any consideration of the legacy 2012 Mac Pro?
There is a function here. It is to get rid of the heat in the geometric shape. In this case, the cylinder shape is driven by the fan.
A system is going to have to get rid of its waste heat regardless of what geometric shape it happens to be.
And there's really many design choices and options which could have been employed, even if one has a specific objective of a large diameter fan: take a look at the ducting work in a lot of big flat laptops for an illustration.
The only thing that would drive it back toward a rectangle would be injecting more rectangular components and need to break down into more thermal zones.
Not really: that's really more of a question of to what degree are they motivated to expend resources to minimize the system's overall volume.
Insofar as the nMP's thermal package, I fully expect that the eventual teardowns are going to reveal multiple thermal sensors and as such, "logical" thermal zones. From there, the simplest engineering approach for a single speed fan control logic would be to base its speed simply on what the hottest 'zone' needs...just a straightforward 'MAX(Zones A,B,C,D) --> table' line of code.
No one has ever complained about having their electronics being exposed to excessive cooling, and since its a desktop, the only reason to try to limit fan velocity really isn't power management, but noise signature .. and that's really already been addressed by maximizing the fan diameter - - any additional potential complexity won't likely result in a meaningfully significant benefit to justify itself.
-hh
