Last year, Apple held a small event with a mere handful of big names in the Mac blogosphere. It was at this event we got the first whiff of the iMac Pro, which was merely mentioned as a "great" new iMac model in the pipeline. The event was to address the aging Mac Pro 6,1, which Apple discovered was designed in such a way the newest graphics processors can't be cooled by the innovative (but oddly specific) cooling system.
The Mac Pro 6,1 was widely panned basically since it was announced as being a bad successor to the Mac Pro 5,1, which was a large tower computer with room for two processors, four hard disk drives, two optical drives, and four PCI Express slots. Years after its introduction, the Mac Pro 5,1 was still lauded as the best system for creative professionals. (Note: the linked article appears to be written by a business specializing in selling customized Mac Pros and aftermarket options specifically for the Mac Pro 5,1.)
I won't belabor the point too much, but it's worth noting for all intents and purposes the 6,1 is a fine computer. The scalability problem allegedly keeping Apple from updating is the dual GPU design was done under the presumption multiple midrange GPUs would become the standard for creative professionals, mainly because the AMD GPUs available in 2012 weren't particularly great at running general purpose workloads at the same time they were operating as graphics cards. As such, Apple built a system meant for one CPU, but two GPUs. It created a system approximately 5% faster than the old one, at a great increase in cost, and at the expense of internal flexibility. So, to use a car analogy, it was like giving someone a sports car when what they asked for was a pickup truck or a four-door sedan.
Apple's solution is two-fold. The first step in the plan is the iMac Pro. The second is an upcoming "modular" computer, about which we know nothing, other than it will allegedly be a good successor to the 5,1.
The iMac Pro, as the first part of Apple's plan to replace the Mac Pro 6,1 builds on the idea of the 6,1 by integrating a Skylake Xeon W processor and Radeon Vega RX graphics chip (new parts at the start of their lifecycles!) into the body of a 27-inch iMac and adding a bunch of Thunderbolt 3 controllers. The idea is part of the problem of the Mac Pro 6,1 was finding good displays for it, due to its Mini DisplayPort/Thunderbolt 2 outputs. The design should be much more scalable, as Apple works in the future to keep the system updated.
The announcement of the iMac Pro's availability was met with a lot of interesting commentary, almost exclusively about the price. The baseline configuration is $5,000 for a system with an 8-core CPU, 32 gigs of RAM, 1TB of SSD storage, an 8-gig Radeon Vega 56 graphics chip, and what's likely the best 27-inch display you can buy. The top configuration is around $13,999.
Immediately, most of the commentary was (as it always is) about how you can build an equivalent PC Workstation for a lot less money. This is technically untrue and the argument almost always relies on a user not needing most of what's available in a configuration, and the fact that Apple often intentionally chooses only high-end parts out of a range. For example, there are desktop workstations from PC OEMs with Xeon W processors, but there are options to configure those system with quad-core CPUs and 16 gigabytes of RAM, which Apple does not allow. The other thing to consider as part of the iMac Pro's cost is its 27", P3-capable 5k display, which isn't available inexpensively anywhere else. The nearest configurations from Dell can match the iMac Pro's CPU, memory, and storage configurations, and then use a low-end GPU and low end display. This is, of course, the beauty of the generic PC market – not everybody needs Apple's 5k display or a Radeon Vega GPU, and building a different system allows you to put that money into different things.
The interesting comparison I heard from a lot of people, and it was surprising to see this from some Mac power users, was a "workstation" built using high end desktop enthusiast parts would be well received. This is where we start to get into some interesting discussions about what makes something a "workstation."
Traditionally, something was a workstation if the vendor called it as such. Computer vendors have traditionally been careful with the w-word, because it meant claiming you believe your product is a step or two above the competition. For most of the 1990s, this meant while the Mac and PC markets contained professional computers they didn't contain workstations, because for every Power Macintosh 9600, there's a much better equipped SGI Octane or Sun Ultra or Compaq AlphaStation with 64-bit processors, a better operating system, faster networking, properly implemented and much faster SCSI subsystems, and so on.
In the 2000s, workstation-class hardware started becoming less expensive and the money needed to rev up all the different platforms wasn't quite so available as it had previously been. Intel had been building chips suitable for low end technical workstations and Windows NT was up to the task of being a workstation OS. AMD also produced the 64-bit extensions to x86 and licensed them to Intel. Apple had just acquired a workstation vendor, NeXT, and started coupling the better OS with some of its newest hardware.
In the early 2000s, Apple started advertising its hardware and software to classical RISC UNIX workstation users who were looking for a modern platform, especially as some of the RISC UNIX vendors failed to commit to building new workstations based around their old UNIX operating systems, either on the old processors or on other hardware.
I've had a lot of discussions with people on the finer details of this point. It is argued that because Apple was putting its foot almost exactly up to the "workstation" line with ads such as Sends Other UNIX boxes to /dev/null and efforts such as the Xserve G4 it's safe to say all Macs running OS X at that point were workstations because of Mac OS X. I tend not to agree with this point because Apple had built UNIX systems before, none of which it tried to classify as UNIX workstations in the sense that, for example, a Macintosh IIfx could compete against a SPARCstation. By 1990 when the IIfx came out, Sun had moved onto the SPARC architecture and in raw compute numbers, a SPARCstation was a couple of times faster than the IIfx.
The change in the mid-to-late 2000s is "workstation" as a term went from meaning a machine that existed in a different performance class from normal desktop computers to something that is qualified to run specific applications, or is qualified by particular hardware features, regardless of what performance class a machine is. (For example: error correcting memory.)
I have been talking about The Plateau for a few years now. I should probably start a page or category for it. The relevance here is leading into the 2010s, new workstation products started to use Intel Xeon processors aligning very closely with mainstream desktop platforms. This ended up representing a new low end in workstations in particular, positioned for "entry level" work and certified often for tasks such as viewing CAD files, 2d CAD, software development, general purpose UNIX chores, and so on.
This is speculation, but my theory is Apple, upon starting to use the term "workstation", decided its workstations should in the traditional sense be a step or two above normal Macs. There was a period of crossover in 2009-2010 where quad-core iMacs were starting to exist, at a time when the baseline Mac Pro was a quad-core configuration, for just $200 over the top end iMac. Then the two lines diverged again, with the Mac Pro clearly demanding a premium for its performance and reliability improvements over the iMac
Over the course of a few generations of iMacs getting new processors and graphics and with the Mac Pro 5,1 then 6,1 standing still, the gap narrowed again, but Apple has reopened it with the iMac Pro. Keeping that gap open requires that Apple keeps the machine updated, but that should be easier to do with the new thermal configuration.
The internal upgradeability of the iMac Pro has been discussed a lot as well. The memory can be replaced, but only at a service center. Tear-downs reveal that the solid-state storage is on modules, although the modules are unique, and that the processor can be replaced.
Storage flexibility was a prominent criticism of the Mac Pro 6,1, particularly as it pertained to the machine's video editing credentials. (Important, given that Apple pretty much designed it explicitly for 4k video editing) I don't know the state of the ultra high end video editing today, but up to that point, it was common for the highest end video editing systems to feature not multiple internal hard disks, but FibreChannel or SAS cards to connect to disk arrays. They fit in well with the large video tape recorders, other import/export equipment, sound boards, and interconnect boards, plus program monitors that often end up in the highest end systems.
I think criticisms regarding upgradeability are a little misplaced. Primarily, Mac users remember (or hear about) a time in the distant past when the machines could be upgraded with new parts and processor generations well beyond what is considered reasonable or necessary today. These upgrades often didn't deliver anywhere near the potential performance of a whole new computer, but I understand the appeal as a way to get a little more life out of a machine in an era when the bare minimum baseline price for a fast new machine is over $3000.
Upgrades for capability and capacity, like RAM or storage, would be nice to be easier to do, but they appear to be doable and storage upgrades housed externally are in a better position than they have ever been, so it's not particularly worrying to see, say, a machine where the primary way to add storage is via ThunderBolt 3.