Bear in mind that a big justification for the price gap is not so much number of cores, but Xeon, ECC RAM and "workstation class" GPUs - which provide more I/O bandwidth (=more TB3 ports), improved stability/reliability for sustained heavy computing and (arguably) optimized OpenCL performance (although that might just be down to drivers).
Optimized OpenCL? Maybe optimized Metal. but far more lkley the bigger horsepower "grunt" of the iMac Pro's GPU (AMD Vega) relative to the iMac ( AMD Polaris ... which was always meant to be less grunt than Vega) is really the gap. I doubt Apple has put time and effort into optimizing Vega better than Polaris since the latter is far widely deployed ( iMacs and MBP 15" ) .
ECC is vastly more expensive than regular RAM is a myth. Not sure why Apple would want to blow more hot air into that fire. The top end iMac Pro BTO configured to top CPU , 32GB RAM , and 1TB SSD is $3,699. That's a $1,300 gap. The Xeon W 2145 8 core is less than that ( $1,200) (and the i7 4770 is $350 so that gap is $850).
The 6-core i7 is likely to give the 8 core Xeon a run for its money on well-chosen benchmarks (I'm wondering if new iMacs might max out at i5 hex core - apparently faster than the current quad i7 - which would also reduce the heat dissipation requirements.
Yeah but at 6 core Xeon W would have far more overlap with that 6 core desktop in those same "well chosen" benchmarks. A 6 core in the iMac Pro would bring them closer together where the differentiation would be a more involved , more complication explanation. Apple doesn't want that. If as the question if have a high fraction of multi-core workloads and the answer of 8 cores is better than 6 is a vastly simpler explanation.
For a Mac Pro it is easier because essential combine it with "do you hate embedded displays" or "do you have a special display you are very happen with already". In that case then even if the iMac benched out sightly higher the "distaste" for the integrated screen would be the differentiation. Folks would give up the marginal difference for the more broad match to other features the Mac Pro had.
I'd be surprised to see a Mac Pro price point that gave you a Mac Pro + 5k display for less than the cost of a base iMac Pro.
Apple is kind of crazy if they think they can drive the Mac Pro price up another $1K or more and not loose a substantial number of customers. There are ballpark $600 and $800 options in the 6 core range with Xeon W. Dropping the processor cost in half (and avoiding apple's 30% mark up of that) would help). Four cores ( that base out at 4GHz so run singled thread very well ) drops that to the $450 range. If there is a way for the customer to put in their own storage than the "minimal" SSD doesn't have to be 1TB big. It could be half that which also reduces the price. Similar with easy access RAM the entry configuration could drop below 32GB ( 16GB is still more than the entry level capacity of the iMac ).
The BTO of the Mac Pro back up to the iMac Pro entry level configs could be closer, but the minimal baseline really shouldn't be the same as the iMac Pro. iMac Pro is trying hard not to be the "regular" iMac. The Mac pro shouldn't have to try very hard to present a different. It should
be different in very obvious ways in terms of range of configurations the user can do.
Today... I suspect that the mass market has largely shifted to laptops/all-in-ones and the appeal of a "modular" Mac is purely for pros/enthusiasts/power users - who are in danger of defecting to PC anyway. Even outfits like Dell, HP and Lenovo seem to be offering (and presumably shifting) 'premium' Windows laptops and all-in-one's at increasingly Apple-like prices... despite also offering a full range of cheaper options. So, I'm not sure cannibalisation is such a danger.
Dell , HP , and Lenovo tolerate a much higher degree of fratricide between their adjacent products than Apple will. Their margins are much lower and one of their primary goals is the make that up with volume. Apple doesn't want to be as small as possible, but they also don't want being "biggest' to be a critical path to being profitable.
People moving to "smaller' computers over time because their workload a plateaued more than computers' abilities have increased is cannibalization. Two systems that have same base CPU and GPU is far closer to fratricide than cannibalize. Two products going after the same set of users with essentially the same abilities is largely infighting. What can be added to the Mac Pro should be the major difference. You'd get folks weren't in the same target pool.
[doublepost=1523388793][/doublepost]
Price could be lower if Apple used dual CPU chips instead of a single chip with high core count. Price of 12 core Xeons is insane.
Not for the same base/turbo clock ability. On apps that don't linearly ( or almost linearly) scale with core count you'd be throwing performance out the window.
To get to the same base clock ability of the Xeon W line the equivalents in the Xeon SP Gold line are more expensive; not cheaper. You can get to cheaper if will to through the base clock out the window along with your max Turbo ability.
So you can either have a two cheaper Xeon SP Silvers and be roughly 10-15% better on loads that run close to 80-90+% fully parallel or throw about 20-25% of your turbo out the window. For a system that mostly involves a single user interacting with an individual program throwing the turbo out the window it is generally not a good trade off. The single user means there is going to be a mix of workload ( between more single and parallel.) .
For systems that are carrying the workload of several users ( and almost always somewhat independent parallel workloads ) then having more and somewhat slower cores works. The Xeon SP processors are optimized for data center server work where the typical job is to aggregate multiple users' workload onto a single system.
The Mac is highly targeted at being a GUI operating system. It isn't suppose to be a computational node in a grid somewhere. Yes,
Even with Xeon SP Silvers it isn't hugely cheaper (like down to mainstream CPU pricing). the Xeon SP bronze stuff is dirty cheap but also dirt slow too. It is just clocked even slower to be even cheaper. You are under the Intel W level of performance at that point.
Fewer cores on a chip also allows for higher clock speeds.
at cheaper prices. Intel will sell high count and relatively high clock speed but the price premium doesn't scale linearly. It only makes sense for very corner case multiple user/thread loads.
And for crap's sake, sell all systems with dual sockets even if the second socket is empty, so the second CPU can be added at a later time.
Second socket empty? Never going to happen in Mac systems. Yes, HP, Dell, etc do it, but Apple isn't a "Monkey See, Monkey do " operation. Pointing to HP/Dell and saying "you have to ape exactly what they do" isn't going to fly.
Apple didn't before and they are even more unlikely to start now. Apple isn't going to sell "loose' CPUs. Nor put that into the support matrix.