While a 28-core CPU configuration is a crazy-powerful option that we've never had the luxury of before (and will cost more than a car), the "theoretical ultimate-ist" in me is still slightly disappointed that the new hotrod Mac Pro tower doesn't have a dual-socket option or seem to even have room for a 2nd socket in the future.
There is a pretty good likelihood that Intel will attempt to throw two CPU dies into a single LGA 3647 like package if they can get back on the fab shrink path again. The short term stuff that Intel is doing in "two dies" Apple probably won't follow, but over the next 2-3 years what was "two dies' is likely coming to the W line up. ( just as multiple dies as chiplets have come to AMD's line up).
In contrast, those Z8 and other workstations that Apple compares to have dual-socket options, so they can go up to 56 cores. Hate to cede the ultimate high-end config options to the competition.
Not with a broad set of offerings at the general base/turbo clock speed that Apple is offering and not at the same price levels ($/perf).
The only place where it might make some sense is the rack mount version. However, if that turns out to be bascially the same logic board as the desktop version (with some minor port placement tweaks) then it is highly indicate that Apple only wants to do one board for both. ( Which in turn is indicative don't want to do the dual socket one.).
Given how expensive the Mac Pro is to start with, and how pricey the brand-new processors are, I'm not only thinking about a 56-core arrangement (or whatever dual-socket config) as an original purchase, but also for long down the line, kinda like what many of us have done w/ dual-CPU MP4,1/5,1s by maxing out at 12 core X5680 or X5690 setups.
I highly doubt Apple is putting tons of effort into optimizing the path to folks buying used , Intel end-of-life CPUs into this system as an operational effort. These new systems are probably not going to make for being good fillers into the used xMac hand-me-down crowd demand fillers.
Two major paths Apple could be on. One they go back into Rip van Winkle mode. ( they sell relatively low numbers of these and then do nothing). if there never is another Mac Pro the used prices for these won't drop to 'rock bottom'. And if Intel iterates on the socket design this is a 'dead ender' socket. Again limited 'used' supply will pop up later.
Second path is that they iterate. If they get back to a 2 year cycle then this Mac will get term'ed out for support relatively close to when intel end-of-life the CPU. 10 years in these will probably be on the "obsolete" list on that path; nothing like the 'life preserver' Apple through the 4,1/5,1 over the last couple of years. What happened with the 4,1 was a fluke that probably won't get repeated. .
That's something that would've been nice on the MP6,1 as well— yank out one D3/5/700 and put another 12-core CPU in for 24 total, now that you can get them for way less than originally priced by Apple as CTO.
Dual CPUs never would have worked in the MP6,1 design space.
[doublepost=1560996503][/doublepost]
Yeah, seems like a reasonable assessment. Even then, though, your server comments remind me of the keynote showing a rack-mount version. As you say below, no server products worth a crap coming from Apple, but if rack-mounting, that could perhaps be considered the "scale out" you're talking about. Seems like an even higher core count might be useful in that sense.
Higher core count doesn't necessarily mean more CPU sockets.
https://www.nextplatform.com/2019/04/24/why-single-socket-servers-could-rule-the-future/
For the folks selling dual socket servers tossing out a dual socket "workstation" makes some sense in that can mostly reuse much of the grunt work to get the CPU , firmware and basic board going. Throw on top being a company that 'sells everything to everybody" and it is easy tick to box and throw yet another system out there.
But the movement is to fewer and fewer dual ( or >= 2 ) servers. Even Dell sees that.
Why Apple would swim upstream to a trend moving to drop multiple sockets doesn't make much sense in light of them
not being a "sell everything to everybody" company. The rack system is probably going to be aimed more at virtualize mac instances rather than general usage server workloads ( MacStadium , MS Azure , etc. ) . Renting out a mac vm instance to someone who doesn't need a full time Mac Pro. Another sizable chunk will be folks who just want a workstation in a form that they can put into a box with wheels that has workstation , storage, and some industry specific do-dads. Again not particularly a server.
Few , if any, folks are going to buy the rack version to build a docker cloud infrastructure for generic VMs images. Office/workgroup file server ... not very cost effective there either. ( someone desperate to keep AFP as a deployed option maybe will throw gobs of money at this, but not cost effective.)