Apple 2017 year in review: The 'Pro' desktop market is revisited with the iMac Pro, with m...
After years of no updates to Apple's "Pro" desktop line, the company uncharacteristically made some public statements in April about a forthcoming line refresh, which led to the iMac Pro's release in December. But, the remarks made may or may not ultimately lead to what Apple fans want to see.
During the conversation, the pair of Apple executives teased the iMac Pro which would see the light of day a few months later at WWDC, a new Pro display of some sort, and a new "modular" Mac Pro.
The iMac Pro was promised in 2017, and it made the deadline with about two weeks to spare. The Mac Pro was said to be "not this year" -- more on that in a bit.
The WWDC saw a brief mention of the iMac Pro. It heralded a return to the Xeon processor family, after four years of product development without them. Also promised for the machine was up to 4TB of flash storage, up to 128GB of RAM, and up to 18 cores.
Not much else was disclosed about the machine at the time, including price, or a specific shipping date, beyond before the end of the year.
Preorders for the machine have started to arrive in customers' hands, amidst some complaint that the iMac Pro is still not "Pro" enough for a given use case.
Not every "Pro" user is alike or has the same use case. Apple can't possibly make them all happy, and, frankly, shouldn't be expected to.
Like with "Pro" complaints about the 2016 MacBook Pro refresh, some users are lamenting Apple's choices. Some complain about the 5K screen, saying that they need something better. Others complain about a lack of expansion, and say that because they can't upgrade it easily, means the machine isn't "Pro."
It would have been better if Apple just stuck with a range of MacBooks of escalating performance back in the day, and just skipped the "Pro" tag on the gear entirely. Not because they aren't for professionals -- but because they are, and the people using them don't need to be reminded of such.
And before you get on a "Steve Jobs would have considered all this, and not left us in the lurch" kick, remember that Jobs himself axed the Pro-grade Xserve, because nobody was buying it. And, only 19 months had elapsed between last hardware revision and the death of the product.
Many user hopes are being put on the forthcoming Mac Pro, but there is no real focus of need from that user segment. So, Apple's got to choose what they want to do, lacking a clear consensus from the "Pro" segment who seem to like arguing about what the machines don't have that they need, more than they, or we, care to realize that there is no ultimate machine that will please everyone.
"We designed a system with the kind of GPUs that at the time we thought we needed, and that we thought we could well serve with a two GPU architecture," Federighi said in April. "That that was the thermal limit we needed, or the thermal capacity we needed. But workloads didn't materialize to fit that as broadly as we hoped."
Presumably, Apple has that figured out for the future, in what Federighi called a "modular" Mac Pro redesign. But, at the time the pair of Apple executives never said anything about upgradeability, and still haven't confirmed the presence of PCI-E slots in the new machine at all.
They were less specific about the Mac Pro both in April, and in the press release surrounding the iMac Pro. At the April event, the phrase "not this year" was said -- meaning not 2017.
Apple repeated April's message adding only slightly more data along the way in the iMac Pro press release. In the release, the company said:
The first chassis leaks for the 2016 MacBook Pro refresh surfaced in the summer of 2016, with the machine shipping in October. We've seen nothing at all from supply sources that could be a new Mac Pro.
But, as a reminder regarding the vaunted "upgradeability" -- Apple has never explicitly allowed CPU upgrades. Just because it's been possible with a very few 68030 and 68040 socketed chips, the slotted 601 PowerPC chip in some models, the G4 tower, and now common with Mac Pro tower users still clinging to the old machines, author included, doesn't mean that Apple has ever allowed it, much less encouraged it.
In fact, with some G4 firmware revisions, they actively took steps to discourage it.
PCI-E x4 is sufficient for most implementation. In fact, in our eGPU testing, it hasn't been much of a choke-point at all, versus a PCI-E 3.0 x16 expansion slot. Preliminary AppleInsider testing with a Vega 64 PCI-E card in our faithful Mantiz eGPU enclosure is delivering about 10 percent more performance than the down-clocked Vega 64 chipset standard to the 10-core iMac Pro, and is far faster than the Vega 56 in the $4999 configuration.
Apple may decide that external expansion is sufficient, and easier for users to execute. Given that Intel is loosening up licensing restrictions and costs for the technology in 2018, maybe costs will go down on Thunderbolt 3 peripherals as well.
Apple also claims that in its heyday, that "Pro" desktops were "low single digits" of sales. This may be a Chicken or the egg situation -- which begat which, the poor sales, or the lack of new hardware? Again, Apple isn't talking.
But, what they did talk about is the iMac Pro and future Mac Pro. As the vanguard of the effort, the iMac Pro is clearly a test balloon to see how much of a demand there is for Apple hardware in the segment. Should the iMac Pro, and the consequent "not this year" Mac Pro not sell well, the "Pro" shouldn't expect to see more of it.
Be careful what you ask for. You might get it.
Apple's Pro hardware admission
In a wildly uncharacteristic move, Apple disclosed future product plans to a group of journalists in April. At the event, Apple marketing chief Phil Schiller and Senior Vice President of Software Engineering Craig Federighi dished on new Pro-branded hardware coming.During the conversation, the pair of Apple executives teased the iMac Pro which would see the light of day a few months later at WWDC, a new Pro display of some sort, and a new "modular" Mac Pro.
The iMac Pro was promised in 2017, and it made the deadline with about two weeks to spare. The Mac Pro was said to be "not this year" -- more on that in a bit.
The iMac Pro
The WWDC saw a brief mention of the iMac Pro. It heralded a return to the Xeon processor family, after four years of product development without them. Also promised for the machine was up to 4TB of flash storage, up to 128GB of RAM, and up to 18 cores.
Not much else was disclosed about the machine at the time, including price, or a specific shipping date, beyond before the end of the year.
Preorders for the machine have started to arrive in customers' hands, amidst some complaint that the iMac Pro is still not "Pro" enough for a given use case.
The "Professional" debate
But that's the real issue -- Apple has hemmed itself into a corner with the "Pro" nomenclature. The internet being what it is, fans and foes alike infer too much from the marketing term.Not every "Pro" user is alike or has the same use case. Apple can't possibly make them all happy, and, frankly, shouldn't be expected to.
Should the iMac Pro, and the consequent "not this year" Mac Pro not sell well, the "Pro" shouldn't expect to see more of it.
Like with "Pro" complaints about the 2016 MacBook Pro refresh, some users are lamenting Apple's choices. Some complain about the 5K screen, saying that they need something better. Others complain about a lack of expansion, and say that because they can't upgrade it easily, means the machine isn't "Pro."
It would have been better if Apple just stuck with a range of MacBooks of escalating performance back in the day, and just skipped the "Pro" tag on the gear entirely. Not because they aren't for professionals -- but because they are, and the people using them don't need to be reminded of such.
And before you get on a "Steve Jobs would have considered all this, and not left us in the lurch" kick, remember that Jobs himself axed the Pro-grade Xserve, because nobody was buying it. And, only 19 months had elapsed between last hardware revision and the death of the product.
Many user hopes are being put on the forthcoming Mac Pro, but there is no real focus of need from that user segment. So, Apple's got to choose what they want to do, lacking a clear consensus from the "Pro" segment who seem to like arguing about what the machines don't have that they need, more than they, or we, care to realize that there is no ultimate machine that will please everyone.
The Once and Future King
Apple isn't sticking with the "Coke can" design for the new Mac Pro. Federighi said that they had painted themselves into a corner with the thermal design for the product. Additionally, the architecture that Apple designed was forward-thinking, and they missed the target."We designed a system with the kind of GPUs that at the time we thought we needed, and that we thought we could well serve with a two GPU architecture," Federighi said in April. "That that was the thermal limit we needed, or the thermal capacity we needed. But workloads didn't materialize to fit that as broadly as we hoped."
Presumably, Apple has that figured out for the future, in what Federighi called a "modular" Mac Pro redesign. But, at the time the pair of Apple executives never said anything about upgradeability, and still haven't confirmed the presence of PCI-E slots in the new machine at all.
"Not this year" doesn't necessarily mean 2018
Schiller and Federighi were very specific about the iMac with "server-grade" components availability in 2017, and we got exactly that. Computers aren't designed in a vacuum, and it is pretty clear that given that there were prototypes at the WWDC announcement for attendees to try, that the hardware had been in development for some time. So the iMac Pro wasn't necessarily a knee-jerk response to anything.They were less specific about the Mac Pro both in April, and in the press release surrounding the iMac Pro. At the April event, the phrase "not this year" was said -- meaning not 2017.
Apple repeated April's message adding only slightly more data along the way in the iMac Pro press release. In the release, the company said:
No year again. "Modular" again -- but at least this time it says upgradeable, so maybe unlike the iMac Pro, we'll get a RAM door.In addition to the new iMac Pro, Apple is working on a completely redesigned, next-generation Mac Pro architected for pro customers who need the highest performance, high-throughput system in a modular, upgradeable design, as well as a new high-end pro display.
The first chassis leaks for the 2016 MacBook Pro refresh surfaced in the summer of 2016, with the machine shipping in October. We've seen nothing at all from supply sources that could be a new Mac Pro.
But, as a reminder regarding the vaunted "upgradeability" -- Apple has never explicitly allowed CPU upgrades. Just because it's been possible with a very few 68030 and 68040 socketed chips, the slotted 601 PowerPC chip in some models, the G4 tower, and now common with Mac Pro tower users still clinging to the old machines, author included, doesn't mean that Apple has ever allowed it, much less encouraged it.
In fact, with some G4 firmware revisions, they actively took steps to discourage it.
The eGPU and the Pro
Modular may very well mean "upgradeable with Thunderbolt 3" peripherals. In essence, Thunderbolt 3 provides PCI-E x4 bandwidth, used to good effect in external GPU enclosures. Apple is officially launching the feature in the "spring" with High Sierra -- but again, there is no specific release date for it.PCI-E x4 is sufficient for most implementation. In fact, in our eGPU testing, it hasn't been much of a choke-point at all, versus a PCI-E 3.0 x16 expansion slot. Preliminary AppleInsider testing with a Vega 64 PCI-E card in our faithful Mantiz eGPU enclosure is delivering about 10 percent more performance than the down-clocked Vega 64 chipset standard to the 10-core iMac Pro, and is far faster than the Vega 56 in the $4999 configuration.
Apple may decide that external expansion is sufficient, and easier for users to execute. Given that Intel is loosening up licensing restrictions and costs for the technology in 2018, maybe costs will go down on Thunderbolt 3 peripherals as well.
2017 and 2018
Apple has made its first move towards the "Pro" market in 2017, more than it has in previous years. Maybe it was Apple deciding that like the Xserve, the market was stagnant, and maybe it didn't like what Intel was offering for chips. Who knows, and they don't talk about it much.Apple also claims that in its heyday, that "Pro" desktops were "low single digits" of sales. This may be a Chicken or the egg situation -- which begat which, the poor sales, or the lack of new hardware? Again, Apple isn't talking.
But, what they did talk about is the iMac Pro and future Mac Pro. As the vanguard of the effort, the iMac Pro is clearly a test balloon to see how much of a demand there is for Apple hardware in the segment. Should the iMac Pro, and the consequent "not this year" Mac Pro not sell well, the "Pro" shouldn't expect to see more of it.
Be careful what you ask for. You might get it.
Comments
Regarding your thoughts on the future of "Pro" hardware at Apple... I see the Apple universe in two segments: content consumption (think iPhone, iPad, etc...) and content creation (think MacBook Pro, high en iMacs, iMac Pro and the future Mac Pro). High performance Macs are essential to create hot new games, VR experiences, mobile apps, etc... Developers need good tools. Why should Apple abandon this segment causing it to drift to other platforms? That seems insane to me. Ok, so the pro market isn't a cash rich proposition. Look at it as a cost of doing business and as a means to an end for the future of Apple apps and experiences. Software is still king.
BTW: I started a new Facebook iMac Owners Group so that current and prospective owners can exchange info on this new device. I used to run the widely used Twentieth Anniversary Mac web site that launched in 1997. I have experience with niche machines.
Just set this up today: https://www.facebook.com/groups/1769374126698240/
It's very difficult for Apple to compete with Windows/Linux workstations in terms of variety and versatility. I also think Apple is chose the wrong partner with AMD for graphics, NVIDIA would have been a better option.
An additional excerpt from the Q3 2017 report:
Tim has announced that AR will be the next big thing, and that will require a whole new tool kit (hardware & software) to create apps and content. Apple HAS to own this space if it wants to dominate this opportunity. Going out past five years AR will transition to VR which will take even more computing and software power. So regardless of sales now, I see Apple staying in the Pro space going forward. This means not only Pro hardware, but Pro AR software development programs. This will also merge into entertaiment and allow Apple’s consumptive devices to dominate.
High End/Pro level: being expanded with extreme power in the MacBook Pro, iMacPro and MacPro
Middle level: Mostly ignored. User needs will be met by the lower end of the Pro level machines or the high end of the consumer grade machines.
Low End: Evolving from the A series chip and overtaking the low end MacBooks and Mini's. The only question being: Will Apple put the A series into its MacBooks or add a cursor to the iPad to enable it to become a true laptop replacement instead of a marketing slogan.
And, I think the biggest variable will be how Apple utilizes eGPU's? I can see eGPU's being embedded into external monitors to expand the functionality of MacBooks (or maybe even iPads at some point?) for consumer grade machines.
I suspect that really sticks in the craw of the types of people who mindlessly defend Apple. They now have to admit Apple was wrong, because Apple has admitted it. And in so doing, they have to admit that they were wrong in their mindless defense of Apple’s past pro screwups.
And so this leads me to wonder — is it just me, or is there a really snotty, sour grapes tone to this article?
Just reality. People are getting worked up for slotted everything in 2017, with industry-standard connectors in a Mac Pro. I'd like that, as I have modified a 5,1 to within an inch of its life, but given Apple's repeated messaging on it and very specific words used, I just don't think that's what we're going to get.
If you listen to what they said, instead of projecting your own BS onto it, I think that’s the most reasonable interpretation.
In other words, I don’t think they meant “you can attach TB peripherals to it” — that’s what the 2013 Mac Pro was. (Although of course it will support peripherals — I just mean that’s not what they have in mind when they say “modular”)
Nor do I think they mean “user-upgradable” in the sense that DIY PCs are user upgradable. It’s not like it will now be easy for users to swap out old Xeons for new.
Heck, Apple could not ship a Siri Speaker before Christmas.
And Apple was once rumored to be building a car?
LMAO
My recommendation is for Apple to develop a half a 1U unit with basic high core count and optional high end graphics. Keep a smaller high speed SSD in that for boot and enough room to do basic needs (256 GB). Then sell other 1/2 rack units for RAID SSD storage. Interconnect it all with high-speed TB3 and 10Gb Ethernet. Great thermal management front to back and a sexy face (dark anodized aluminum with a glowing logo). Heck throw an OLED on the front with configurable display (CPU, RAM, Net, etc)
You ship it with rails and mounts. The 1/2 unit should be able to stand/lay on a desk or be mounted in a rack by itself or coupled with another one to make a full unit.
That’s it! It’s really simple. Make blocks. Make lots of them. Make the blocks work together seamlessly. Add hardware-level security like in the iMac Pro. Open up developer kits for PCI hardware developers. It’s pretty simple!
That’s what I think is silly.
It's interesting as an idea, but in reality I don't think the modular concept would work very well. You'd likely end up with higher costs and compromise to the performance of the components.
I think you would end up with more failure in the end as well. The more things are removable, the more potential failure you have. Pro's don't need failure and downtime. This costs them money, and lots of it. I know the 90's techies hate everything integrated, but its a safer way to go as far as reliability.
These are basically like the dongles of a Mac Pro and the Mac Pro sells so low I doubt Apple would ever recoup its costs on something like this. So its okay for Apple to do this to the Mac Pro, but not the MacBook Pro?
I doubt we'll see a MacPro in 2018 either, but only because it does take time to totally redesign a computer from the ground up. It seems that it takes a good 2-3yrs to totally redesign a product. If Apple just wanted to pull an HP and just slap a bunch of shit together with a shitty heatsink on it then they could have had a new Mac Pro out before the end of 2017, but obviously Apple isn't going to do that. If that's what makes you happy then by all means go buy the HP. Nearly everything you see on and inside a Mac (or any Apple product for that matter) is custom engineered and built by Apple. Its not as simple as slap a bunch of shit together and call it a day like PC manufacturers do.
I'd rather Apple get it right than just rush a product to market to appease people like you. So if it takes a few extra months to get a product to market then so be it.
If we assume half of workstations are premium workstations (>$2k) and Apple gets 1/3 of the market like HP/Dell, they are aiming for a best-case of ~200k units per quarter. Apple used to sell this many in 2004 when the PowerMac started at $1800 and the PowerMac line was 20% of Mac sales. Once the price is near doubled, the unit volume drops more than half. These prices are largely due to Intel. AMD's latest chips currently offer much better value, maybe they will help drive Intel prices down if they gain some traction in servers.
When Intel delivered 4-core/8-thread chips in laptops and iMacs in 2011 with 16GB RAM support, there was much less need to pay ~$1000 extra (plus display) to get roughly the same performance from the entry level workstations and a lot of the pro market migrated to iMacs and MBPs.
The numbers show that people aren't buying expensive workstations in large numbers so options and specs don't factor in much. Having more options would help of course but not to the point that it makes the segment overall worth investing in beyond it being a passion project.
It would be nice if Apple offered NVidia options but NVidia workstation GPUs are really expensive. They've been neutering compute features in their consumer cards, including Titans so they can push Teslas and Quadros on the high-end. The worthwhile models of those GPUs cost thousands each. Gaming cards like the 1080/ti aren't ideal for workstations. Titans are the only realistic option from NVidia and still ~$1200 each.
The cylinder Mac Pro has the processing boards plugged into a circular daughterboard at the bottom:
Those boards are made by Apple. If they had a system that allowed for 4 modules containing boards to be connected like this and more easily removed, they can not only offer more options but easier upgrades. If someone wanted 3 CPU boards and 1 GPU board, they'd just install that config. If someone wanted 1 CPU and 3 GPUs (likely 2 high-end for compute, 1 low-end for displays), they'd just switch the boards and position for the best cooling, e.g if they are in a row, high-end GPUs at either end. If they get a return on a machine, they can easily reconfigure it. Mid-cycle, they'd be able to offer board upgrades.
The boards would still be stuck to a different arrangement of heatsinks, that's the best way to keep them cool but the modules could be designed so that there's a copper bar on them and the chip is stuck to that. Then they stick the copper bar to the chassis heatsink, which makes it easier to remove things without damage.
The main demand is for GPUs rather than CPUs. CPU processing can be done on servers if the bandwidth is there:
https://aws.amazon.com/ec2/pricing/on-demand/
Just spool up 96 cores for an hour or two and it's under $10 to get the job done. This can be done from any hardware. What most people seem to want is an inexpensive box that they can keep upgrading with the latest single GPUs. This is what people did with the Mac Pros ~$2.5k, just throw in a new $500 gaming GPU every 3 years and keep it running for 8-10 years. This setup can be covered with eGPUs.
The cylinder Mac Pro has been obsoleted by the iMac Pro as it has the same thermal capacity and the iMac Pro is even more compact. There's no point in Apple selling both options so they have no choice but to make the Mac Pro more powerful in order to avoid the overlap. This will have to be around double the performance of the iMac Pro to be worthwhile making at all but it also means more expensive. If the Mac Pro wasn't selling at $3k, it's not going to sell better at $5k regardless of the options it has. By the time the Mac Pro comes out, the iMac Pro will be on its second revision. GPUs will be at the point where nobody cares about upgrades, same with RAM. People get stuck thinking that they've needed to upgrade before so this need will come again but this isn't going to be true forever. Every computer user will reach a comfortable RAM/storage/GPU limit and for 99.99% of people this will be no more than 64GB/4TB/10TFLOPs. There are laptops that have this now but it will be in sensibly designed laptops in about 5 years.
Given that the iMac Pro will cover so much of the pro hardware base, Apple could just build custom Macs for the remaining users that need more power. Just like how they gave away $10k+ Apple Watches, they'd say to the people who need the fastest workstations to send in their component requirements (48-core Xeon, dual 16GB Quadro, 128GB RAM, 8TB SSD, any custom chips like RED decoder) and they'd just custom build it for the cost of the parts plus markup. Ultimately even they will be satisfied with the lower Mac options eventually.
Apple could also help the pro Mac community by offering compute services in their data centers. They can have it bundled with the developer program or have a creative subscription at $99/year that includes Final Cut and cloud syncing for sharing movie edits. If we are talking about a few tens of thousands of users who each need some amount of CPU power every day, they can have 100,000 cores shared (~4000 servers, $40m). If a user needed to render something from After Effects, the Mac would upload the assets and Apple would have a deal with Adobe (and others) to run the processing across the server nodes and then download the result.