Whatever the new Mac Pro ends up being one has to questions Apple's commitment to it. It's approaching five years since the current Mac Pro was released and it has seen absolutely no changes since then.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP 7,1 Waiting for Mac Pro 7,1 A1991 (no more)
- Thread starter Mago
- Start date
-
- Tags
- macpro 2016
- Sort by reaction score
I definitely think this is frankly a more important question than the form that the machine takes.
The nMP wasn't perfect for my needs, but I would still have happily bought it if they'd ever updated the damn thing. As it was, it just wasn't enough of a performance boost for me to justify in 2014, and then it proceeded to never be upgraded. Apple shied away from their own investment into OpenCL and dual GPUs, so why should anyone else care about that?
Now they've got the iMac Pro, which basically covers the 6,1 Mac Pro cases although now you have a more expensive buy-in cost. But I'm mostly concerned that they slack on updates.
It's kind of weird seeing people getting upset at Apple for not coming out with new MBPs immediately after chips are out and instead two months later, when they've historically had far worse timings with their desktops. Either you update consistently with available chips or you update irregularly but clearly telegraph when new stuff is coming; Apple's problem is that they've been relatively opaque about releases on top of irregular ones.
There's evidence that this is shifting somewhat, but even if they're fully back into committing they've got to get the hardware out and then show support for it before people are likely to believe them.
And well noted, particularly when one looks a little further back in time to the PPC Power Macs, where their "Starting At" prices were as low as $1,599 (G3 "Blue & White" /JAN1999 and the G4 "Yikes"/AUG1999).
Which becomes particularly problematic as consumers shop beyond Apple. For example, on mainstream commodity SSD prices, Best Buy is having a sale right now on a SanDisk Ultra 3D 1024GB SSD for only $170 ... which is currently having me searching around right now trying to find what the deal was with 5,1 Mac Pro (in?)compatibility with 3D SSD's.
That's assuming customers consider the iMac/Pro a viable alternative to an MP, even when that iMac doesn't come cheap .
I don't see much cross shopping going on there ; some people want a convenient all-in-one solution - at a low price - other people want performance and flexibility - starting at a lowish price .
Zero flexibility at a high price, mid level performance on a cool day, it makes no sense for this computer user .
Agreed! If it’s another iteration of 6,1 there’s no point. Apple should show there is room for updates in the future-gpu upgrade path (either cards or modular option-although they could easily argue the solution is egpu boxes...) but overall they can’t ignore, what should be, the flagship pro machine.
While most people are buying iMacs (and laptops ), some people are not. The latter is the whole point of the Mac Pro at this point. They tried tweaking the Mac Pro into a very narrow corner, limited flexibility design and it didn't work out as well as expected. It wasn't the complete failure that some folks spin it to be as primary, priority response by Apple was the iMac Pro ... not a 180 retreat to old Mac Pro. However, that is one of the primary points though. The inclination to building something complete is what they have already done. So if adding to the Mac line up it would probably be something other than what they already have.
It isn't just about protecting high margins. The vast majority of customers just want something that just works. They don't want to fiddle with the hardware any more than they want to fiddle with the internals of their car engine. As system that doesn't break down and needs no maintenance is one of the primary wishful objectives ... not a feature to pop the hood and rattle around with their trusty screwdriver.
Apple's example of modularity in their April pow wow as plugging in a monitor. So no this isn't about stripped down the to core internal components modularity. However, it is probably not about zero access either (since they already have that. )
It probably has a heavier weighting on extermal modularity. Monitors and Apple is likely to point to Thunderbolt to some extent. How heavy they should lean on TB is where they haven't been transparently clear. It is probable that folks who need "lots" of slots will be pointed to TB and Apple push to eGPUs. However, there is little to indicate that Apple is going to exclusively rely on that ( they made comments about enabling leading edge bandwidth and i/O. TB has limitations at the top end of bandwidth demands. e.g., 8K RAW capture. extremely high data storage I/O ... etc. )
Unlikely. The explicitly referenced modularity of a monitor would most likely be accomplish through a Thunderbolt connection ( yet another Apple Display docking station. ). Third is sequence after Thunderbolt Display Docking station and the somewhat recent LG units ( with Apple design constraints of one and only one input connector). MP 2013 had a HDMI socket as an alternative. Not much to indicate Apple would skip that and/or DP v1.4 out also.
No they likely would not do modules but Apple's SSD is largely evolving around the T-series serving as a controller. It is not likely that Apple would have more than one T-series chip in a Mac so just one SSD. If look at the rest of the Mac line up one SDD is all you get. The question would be why saddle the Mac Pro with that same limitation. ( if trying to pack something into a literal desktop and minimize desktop footprint perhaps. e.g., the MP 2013. .... but they already have that in the iMac Pro. ).
The NAND cards may show up in both the Mac Pro and iMac Pro ( and some iMacs ), but I doubt that is going to be "interchange" like you are using the term. It is modular in that if organization has a data destruction policy for computer systems that go into retirement then can yank the NAND cards and shred them. But it isn't third party commodity part modular.
If Apple has more than one SSD internal connector there isn't a good reason why it wouldn't be M.2 based. There are several external modules for M.2 appearing ( e.g., https://eshop.macsales.com/shop/express-4m2 ) so notion that Macs won't "see" a 3rd party M.2 is ridiculously narrow minded of Apple.
Apple doesn't need some extra-ordinary module for video but the likely do needs something that meets the functionality requirements of integrating well with Thunderbolt. However, for what was the 2nd, "Compute" GPU that was in the MP 2013 that doesn't need to be the exact same constraint. A x16 PCI-e standard slot could be used for x4 , x8 , or x16 cards. Apple could sell a narrow set of cards that fit that standard slot, but other 3rd parties could fill in the rest. [ again for folks who need 2-3 more slots they could point to the eGPU boxes, but they shouldn't be a requirement for relatively common user configurations. ]
There is a thread in this Mac Pro subform that asks what is in folks PCI-e slots. A very common configuration is a 'boot" GPU and another GPU (or high end i/O data/capture card). If the new Mac Pro doesn' enable what folks are commonly doing now then it will have problems. A free slot also gets substantially out of the way of the Nvidia vs AMD fan boy wars ... which aren't particularly rational, but are a significant part of the market segment. Leaving it out entraps Apple deeper into that tarpit.
Jony Ive should be able to design to what constraints he is given. The real core issue here isn't Ive but some apparently spineless marketing and product management that don't push back when the proposed designs don't meet market requirements. If Apple just scales back a on the design requirement to deal random PCI-e cards from 4 slots to just 1 (or maybe just 2 ) slots then that is a reasonable compromise that Ive (or any competent designer) should be able to handle.
the other core issue is that Ive and his limited sized team don't have the bandwidth to engage in the full set of products that Apple is actively pursuing at this point. The Mac team should have their own product design team (minimally for the the desktop line up and part of the laptop. )
However, it really isn't discount part from Newegg that are the core issue. It more so cards and equipment that folks have deep sunk costs into. Audio card+software that is $2-3K , or video capture + software that is same ballpark. sunk costs of $5-10K into CUDA. Once the sunk costs get up around 60+% of the cost of the new computer system they'd be going into then start to get into the "tag wagging the dog" is try to inflate the Mac Pro system costs very high. The Mac Pro will get dropped for a more cost effective alternative if push the price elasticity too hard in those contexts.
It is far easier to separate more folks from a $100 internal Blu-ray drive than from something an order of magnitude higher than that. The vast majority of the computer parts that Newegg sells aren't really the issue. Nor it solely just the latest tech spec porn gaming GPU. Professionals who have equipment that needs a slot is different from the Pros that don't have any slot use (which are better iMac Pro candidates). It is two different market segments.
A whole lot of hand waving. First, as soon as 27" iMacs got SSDs as BTO they jumped well into the $2k price zone.
The top end of the "good , better , best' standard configurations used to top at at $1,999 but clearing out for all iMacs never was an practice.
The top "best" of the 3 standards options tops out at 2,299 so it is still under $2,499 mark. If go to middle iMac 27" and bump the RAM to 16GB and the drive to 512GB you land at $2,499. So yeah that it is in the old Mac Pro price zone, but it is still lower than where the Mac Pro 2013 was ( at $2,999).
If look at BH iMac phot best sellers.
https://www.bhphotovideo.com/c/sear...PULARITY|1&srtclk=sort&N=0&Ntt=apple imac 27"
The top selling SSD iMac model is a $2699 model ( with 1TB drive) which is still lower than the $2,999 threshold. ( and yes a discounted iMac Pro is above that. but again that it is discounted is more illustrative that the $4,999 price has problems. ). $3,600 iMacs aren't unheard of, but they also are quite far from being the mainstream of iMac sales.
The GPU is probably more of an issue of not being able to hit the $2,499 price point agin more so than any CPU overlap with the iMacs. Apple would have to hit 4 core , strip down the RAM , SSD , and GPU to hit that point. The GPU probably would be difficult as I don't think it would be an off the shelf product with some minor firmware bump.
Pricing the Mac Pro so that extremely few want to buy it isn't going to significantly "Protect" the iMac Pro at all. Most of those folks are either going to walk away from MacOS or get a hackintosh. Neither of which is going to provide Apple with substantive revenue as opposed to actually trying to meet the needs of the market. Apple doesn't have to folks who demand that they just 'clone' Windows Workstations, but they also can't purely make the Mac Pro into a "saving" the iMac Pro. The iMac Pro doesn't really need saving that bad. It is not costs so much that Apple is not providing in this space but functionality ( e.g., enable the sunk cost hardware that customers have ). Moving the Mac Pro higher in price isn't going to solve the function mix match that the iMac Pro has with their requirements.
Apple can't get 100% market separation but in the 2.5+K price zone the more highly price sensitive folks are gone and the issue gets more grounded in functionality.
In the xMac space it is more about price than it is the higher priced equipment. It is more so about riding the commodity parts supply chain in a 'race to the bottom'. Or at least much closer to the bottom. Systems should always get cheaper over extended amount of time .. blah .. blah blah. That's not what the Mac Pro vs. iMac Pro should be primarily differentiated on.
But will the customers buy (literally) the compromise, or decide that Linux and Windows are their future?
Last edited:
While probably it probably has some user demographics selection bias drift in the the thread in this forum that asked for real world slot usage indicates that it isn't much of a compromise. Most folks have a "boot" GPU and one other x16 type of card usage.
The x4 slots are mainly used for SSD and to keep up with beyond ancient ports on the older Mac Pro ( USB 2 -> USB 3.x ). Direct M.2 slots would enable SSDs and Thunderbolt merged with USB in terms of physical socket. USB 4 (etc. isn't going to top TB anytime soon.). You have ports that can adapt to the future nominal mainstream I/O.
Most Mac Pro users probably aren't using all the slots. They'd loose some edge case folks, but it probably isn't the mainstream workstation market. ( z4, z6 and first two Dell workstation enclosures have limits on "big hungry" add-in cards too. )
Apple probably has better data than that forum questionnaire.
Many folks were not using optical drives ( not all but a significant fraction that was growing ). Apple tossed ODDs from the line up and has done just fine. Did they loose rabid optical fans, Yes? Was that critical for the overall Mac ecosystem? No.
The folks who will be unsatisfied until Apple builds spec-for-spec clones of the HP/Dell workstations are never going to satisfied until Apple churned out a commodity system. Those folks are going to leave eventually because that path doesn't line up with Apple's path for rather long into the foreseeable future.
[doublepost=1532725042][/doublepost]
Shouldn't have to wait more than 2-3 years to see clear direction. If Apple uses late '18 - early '19 tech then there should be some iteration, if just speed bumps in 2 years. For example, if using Xeon W then by '20-'21 there should be some 10nm alternative. ( if used AMD then again there should be at least a bump in two years). Similar on the GPU side. Especially, if the entry level GPU is Polaris/GDDR5 based. That's has hiccuped with HBM prices staying high and AMD not moving Vega across the line up. That is exactly the issue. Being non-responsive to design lock ins with the original design.
I wouldn't expect every year. The whole Mac line up has year gaps over the last 2-4 years. That is probably going to continue. But at 3 years even if skipping every other (and there is a minor hiccup) something should happen.
The "modular" aspect that is a discrete monitor/docking station. yeah that wouldn't be surprising to be throw into the "do nothing" status for 4-5 years at all.
Don't you remember the Lord God Jobs exhorting y'all to "Think Differently"? Who mentioned spec-for-spec clones?
Think about mating modules, where the interconnect is the QPI! Back to the picture:
This system could be split in two, rather easily. The closer processor P1 (square heat sink tower) has its twelve DIMM slots and the four PCIe slots on the right. (Some of the P1 PCIe lanes go to the onboard SAS RAID controller and NICs, so P1 only has four slots.)
The back processor P2 has its twelve DIMM slots and the five PCIe slots on the left. (All slots are single width, so the dual width Titans block a PCIe slot.) Note that the P2 is optional - but without it the twelve DIMMs and five left PCIe slots are dead.
The only real connection between the two halves is the QPI, and there's no need to have the QPI connected to another processor - P2 is optional.
Imagine a modular Mac Pro that has a mini/midi tower with P1, its DIMMs, four PCIe slots, and onboard IO.
The expander chassis has P2, its DIMM slots, and PCIe slots. The side panel of the P1 chassis comes off so that the expander can connect directly through the QPI. (Firm latched or screwed mating - no chance of accidental disconnect.) A few extra control signals - for example so that the power supply in the expander will be controlled by the main power switch.)
we will know instantly if they release something that will be hard to redesign in the future. The nMP's biggest issue is to update it they need to design a custom motherboard and gpus. Who thought this nonsense was a good idea when computers need to update every 18 months? We will know.
It's very likely that if you need > 2 graphics cards, Apple will push you into Thunderbolt enclosures.
There is just no way to push stuff like processors and RAM out to enclosures. The performance penalty of just moving the electrical signals that far, even if you had the bandwidth, is just too severe. Latency is the issue even if bandwidth isn't. Unless Tim Cook can change the constants of physics that's just not happening.
There is just no way to push stuff like processors and RAM out to enclosures. The performance penalty of just moving the electrical signals that far, even if you had the bandwidth, is just too severe. Latency is the issue even if bandwidth isn't. Unless Tim Cook can change the constants of physics that's just not happening.
Not necessarily, because my suggestion doesn't put the enclosure at the end of a metre long cable. It basically uses a two-piece mobo.
Look again at the picture I posted. The two processors are 10cm to 15cm apart, connected by the QPI. It could easily be done with a tightly coupled secondary motherboard that mates with the QPI of the primary motherboard.
There is little to support this issue. Every Mac in the the product line up has a custom motherboard. The 2008-2010 designs were custom either. For example, 2009-2010 model has a CPU+RAM tray ( daughter card ). The mainstream competitors of those Mac Pro did not .
It is not about whether it is custom it is far more so likely whether Apple has people and resources allocated to it. The MP 2013 could have been updated. Bumping the CPU and PCH would not have been some "moon shot" complexity level project. Once AMD got to Polaris tech the GPUs could have been upgraded. ( If Nvidia and Apple could have come to terms there could have been options there too. The design itself was the primary issue there. )
Where is it universally required that computers update every 18 months? The notion that Moore's law is the primary driver of user computer buying cycles is deeply flawed at this point. More than a few folks are buying new computers at a substantively slower rate than they were a decade ago. Apple's multiple year gaps in the product line updates is probably too slow, but that is Apple "dialing in" to the effort they want to apply to the product line. It is not outside the scope of capability if resources were applied.
There was bad timing in the Mac Pro 2013 design ( jumping on the tail end of the Intel tick-tock sequence for the class of processors that went into the Mac Pro. AMD getting stuck betting on a fab process shrink to work their way out of some issues. ) .
Apple only works on a fixed number of Macs at a time. ( same with the other major product lines also. Fixed number of iPhones, fixed number of iPads , etc. ). Apple could release a "box with slots" but if they design resources on the product goes down to zero for 4 years, you want see anything. The "box with slots" is form over function here.
with AMD epyc you can have an 1 cpu tray and an 2 cpu tray. And with both you have the same number of pci-e lanes
TB3 is too slow to allow eGPUs on a mMP. Unless they come out with a TB4, bandwidth can be easily saturated.
With the official delay for Cannon Lake, Intel will very likely not come out with new Xeon before 2020.
So... what now? A longer wait or an outdated machine?
With the official delay for Cannon Lake, Intel will very likely not come out with new Xeon before 2020.
So... what now? A longer wait or an outdated machine?
I wholly agree that we'll see a LOT of Thunderbolt 3 on the new Mac Pro... The iMac Pro has two TB3 buses (4 ports, but only two separate buses) - the Mac Pro may match that, or have 4 buses (8 ports). Especially if it has 4 buses, it would be possible to double-connect an eGPU enclosure for extra bandwidth (of course, MacOS and the enclosure would both have to support it).
The earlier mention of commodity SSDs as cheap as $170/TB is confusing cheap SATA SSDs with the PCIe SSDs (at least twice as expensive, really fast ones are more) that are all Apple has used in recent years. Apple might put some standard NVMe slots in a new Mac Pro, or they might do something PCIe based, but proprietary (perhaps a combination if the boot drive is something really fast running off a custom controller on the T2 (or T3?), but they also allow a couple of NVMe slots).
I'd be VERY surprised to see any internal provision for SATA SSDs. The only two current Macs that support any kind of SATA drives internally are the Mac Mini and the iMac, both of which are older enclosures (2011 and 2012, respectively), introduced before "PCIe SSDs only" became viable. There's a lot of speculation that the iMac will lose its SATA capability in the next case revision, as the iMac Pro already has.
As for other expansion, I'd put RAM in the "probably" category - maybe rising to "almost certainly". The RAM architecture will certainly be standard - probably nearly identical to the iMac Pro, likely with two slots per channel for a total of eight. The only question is how accessible the slots will be. Apple has certainly been known to bury RAM slots behind things and on the back of motherboards...
There will be some provision for PCIe cards, probably internal - but there is very likely to be a catch of some sort . It seems very unlikely that they'll let users plunk a standard PC graphics card in there and have it work (there has never been a Mac that didn't have some graphics barrier). They will probably be trying to accommodate Mac-specific audio interfaces and the like, while keeping GeForces out. Even "any old Vega will work without a custom ROM" is more than the Mac Pro has ever offered.
The earlier mention of commodity SSDs as cheap as $170/TB is confusing cheap SATA SSDs with the PCIe SSDs (at least twice as expensive, really fast ones are more) that are all Apple has used in recent years. Apple might put some standard NVMe slots in a new Mac Pro, or they might do something PCIe based, but proprietary (perhaps a combination if the boot drive is something really fast running off a custom controller on the T2 (or T3?), but they also allow a couple of NVMe slots).
I'd be VERY surprised to see any internal provision for SATA SSDs. The only two current Macs that support any kind of SATA drives internally are the Mac Mini and the iMac, both of which are older enclosures (2011 and 2012, respectively), introduced before "PCIe SSDs only" became viable. There's a lot of speculation that the iMac will lose its SATA capability in the next case revision, as the iMac Pro already has.
As for other expansion, I'd put RAM in the "probably" category - maybe rising to "almost certainly". The RAM architecture will certainly be standard - probably nearly identical to the iMac Pro, likely with two slots per channel for a total of eight. The only question is how accessible the slots will be. Apple has certainly been known to bury RAM slots behind things and on the back of motherboards...
There will be some provision for PCIe cards, probably internal - but there is very likely to be a catch of some sort . It seems very unlikely that they'll let users plunk a standard PC graphics card in there and have it work (there has never been a Mac that didn't have some graphics barrier). They will probably be trying to accommodate Mac-specific audio interfaces and the like, while keeping GeForces out. Even "any old Vega will work without a custom ROM" is more than the Mac Pro has ever offered.
TB3 is too slow to allow eGPUs on a mMP. Unless they come out with a TB4, bandwidth can be easily saturated.
With the official delay for Cannon Lake, Intel will very likely not come out with new Xeon before 2020.
So... what now? A longer wait or an outdated machine?
eGPUs don't have that much of a performance penalty in real-world usage. If you're talking about a Mac Pro, then the only time it'd particularly make sense to have eGPUs is if you're running software that can take advantage of multiple GPUs, at which point you're still getting additive gains from an eGPU, albeit not as much as if it were inside the computer. Given that the previous tower only had room for two full-size GPUs anyhow, if the new one has the same limitations you're still gaining something for your money versus the old towers, although you're still paying $200+ for 75-85% of the performance.
The old macpro is a btx form factor case and parts for it are more readily known. Regardless of the daughterbords when you are building into a standard you have some flexibility...
https://en.wikipedia.org/wiki/BTX_(form_factor)
Except when they do - ie a performance penalty in VR can drop you below the Barf limit, and there's plenty of VR stuff that flirts dangerously close on a 1080ti in a motherboard slot. The other major one, is that a TB3 eGPU can't feed all the displays it's capable of supporting. On a RX580 you've got 2x5k60 capable DP1.4 ports, 2x4k60 capable HDMI ports, and a 2560x1600 capable DVI port - I'm happy to be corrected, but I don't think you can actually push all that display data down a single TB connection.
No you can't. You can do the 4K resolutions and below fine, but the DP version in TB3 means you can't do multiple 5K.
Again, TB3 has limitations. Some people need performance for VR applications, some people don't give a hoot about VR. Some people have work or needs that bottlenecks on a single GPU and thus eGPU are a less performant option versus highly scalable workflows, etc. But those limitations aren't going to matter to everyone, especially when you're talking about huge performance gains versus the status quo (and given that Apple sells a lot more MacBook Pros than pro desktops, then focusing on it so heavily makes sense.)
I don't really envision Apple making a 6+ TB3 Mac Pro because a desktop has fewer reasons to invest so heavily in external expansion and they were already burned by that in the previous revision. Not to mention I'm pretty sure at least with Xeon-W that you couldn't have the same number of slots as the previous Mac Pro *and* 4TB3 ports without running into bandwidth limitations; in the real world I doubt most people would max out their boxes and their Thunderbolt ports at the same time, but there are only so many PCIe lanes.
Given the depressing news out of Intel recently, I kind of hope Apple isn't trying to wait for the next gen Xeons for the Mac Pro, because it seems like it's going to be a long wait for 10nm...
It's not the displayport version limitation over TB3 though that I'm thinking of. A standard desktop GPU in an eGPU box, has displayport between itself and the display, but the data connection between it and the motherboard over the TB3 cable, that won't get sufficient bandwidth over the TB3 connection, by virtue of being an x4 slot, right?
mMP´19 should come with 14nm++ Cascade Lake-W, I figure.
Apple won't wait till 10nm comes for sure, that would be suicide.
Apple won't wait till 10nm comes for sure, that would be suicide.
Errr. No. For the primary GPU yes it is too slow. But for a supplementary GPU it works. There are already plenty of demos with the iMac Pro that shows that it does. Apple did some demos at WWDC also.
I haven't seen any creditable rumors about TBv4. It seems doubtful it is coming any time soon. Unless, some traction starts up again on affordable fiber cables for for TB then there isn't much room for it to go anyway ( without the copper cables shrinking in length).
With the official delay for Cannon Lake, Intel will very likely not come out with new Xeon before 2020.
So... what now? A longer wait or an outdated machine?
err no. Intel probably has one or two interim bumps for Intel W deviates off of the -SP baseline infrastructure. Scalable has two.
https://www.anandtech.com/show/1311...ap-leaks-cooper-lakesp-ice-lakesp-due-in-2020
Cascade Lake and Cooper Lake. Cooper Lake in the -SP space looks to be a socket bump to match the Cannon Lake that might still trickle in even later. However, the "extra' 2 channels of memory ( to 8) , Omni Path upgrades , and probably Optane DIMMs compatibility probably won't trickle down to the -W packaging. There is a decent chance that even if the Cooper Lake -W socket changes it will on be in minor tweak (e.g.. 2011-2 --> 2011-3 ). They'll bump PCH chipset (also probably about same footprint size) so that compatible with Ice Lake later and tweak to sockets slightly so the upgrades line up.
In that context, Apple would have to do some minor board work in next year to move up, but that would pay off later because changes are coming for Ice lake anyway (probably year after that. the cost was coming anyway). The -W TDP probably isn't changing much. Apple could be exceedingly Scrooge McDuck cheap and wait for Cooper Lake, but would be extraordinarily dumb. Really dumb. The better move if super cheap ( *cough* frugal) move would be to skip Cooper Lake and just do a Ice lake in early 2021 if doing a minor board tweak inside of 12-18 months "scares' them too much financially. The speed bump probably coming at Cooper Lake -W is probably not much to write home about.
However, skipping both Cascade and Cooper is a severely brain damaged move. Few sane folks are going to 'circle the airport' while Apple does something that extremely dubious.
Cascade lake -W will probably pick up "Meltdown" and "Spectre" fixes which are probably worth waiting for given how super late Apple was anyway. It would not be surprising for Apple to bump the iMac Pro and Mac Pro around the same time (so can explain the differences) if both are moving onto Cascade Lake -W.
There is no 'magic bullet' in waiting for Ice Lake. The only thing Ice Lake might bring is more cores in a single die ( which will fit into the -W package easier. ) . The 10nm SP products will all be EMIB based where scale up core counts with multiple dies.
However, "Core Count" is the primary pain point right now though. It is lack of a current system period.