References to Intel's Latest Ice Lake SP Processors in Xcode 13 Beta Hint at Forthcoming Mac Pro Update

[novagamer]

Just because the Memory channels are there on the die doesn't mean Intel has to hook them up to pins.
For the max sized die ( XCC ) with 38-40 cores they probably also need the larger package size for better thermals. However, for the 'low core count" ( LCC) dies they could just skip the 2 memory controllers being hooked up and just more to higher base clocks on the smaller boards space.

They didn't do it with memory channel pins before but I think they did this with UPI I/O in the past. Using the same die for multiple products. Downsiding the I/O to "lower" die into a more affordable price segment.

At the point where in the same 8-20 core count as the W3200 is the two extra memory controllers going to make a huge difference to their implementation given a bump in RAM clock speeds ?

Moving from 6 to 8 channels is so the additional cores don't 'starve' waiting on data when under max memory bisection bandwidth loads. If have 12+ more less cores that "max pressure" is going to be lower.

The problem with cranking up to the bigger package is that soak up more board space. ATX boards are only so big. 3497 is already heavily on the "large" size. More DIMMs also soak up more board space. At some point likely to end up with less general purpose PCI-e slots . In the "enthusiast" market that can be 'bad' becauase more slots is "more power".

I agree with you about the lower core count chips not being starved for memory bandwidth, I just don't think Intel is going to use the same socket for 5 years unless it's a really specific, profitable niche which this single socket product isn't.

The only comparable thing I can think of is when they released Kaby Lake-X on the same platform as Skylake-X which had to be some specific cost decision for the low core count parts, but I don't think Intel's current (much better) leadership would repeat that mess.

If an update does come, I don't expect XCC cpus to be compatible or at least sold by Apple. The talk about Dual CPUs etc. is also crazy, Apple is not going to go down that road with an x86 product that is probably the last of its kind. We're going to get the same case, same PSU, updated architecture with faster CPUs / memory and a couple new workstation cards from AMD - probably nothing else at all. Folks talking about holding back a product because it will make a new Apple CPU look bad are totally wrong-headed, we're not going to see an Apple CPU workstation that costs $15,000-$20,000+ soon, if ever, and even this updated Mac Pro will probably not hit the single core CPU geekbench that the current M1 in a MacBook Air does.. There will be M2 or M3 iMacs that destroy this potential Mac Pro that cost 1/5 the price within a few years at most, but for the people that need the expansion and full x86 compatibility this would be a great swan song.

I even think the Afterburner was probably some internal engineering experiment that they decided to release since it was so profitable for them - just looking at the refurb 2019 Mac Pros over the past year, most of them include it by default to pad the $ and increase profit margins on an already high margin product, so I don't expect any updates there either. It would have been really nice if they released Dev tools for it, but my hunch is they got a great deal on those ASICs and wanted to push creators further toward ProRes which M1 and certainly future M cpus will accelerate very well. Higher margins and funneling people toward their own ecosystem - win/win!

I really hope this update happens, even though almost no one will buy it but me. That's also why maybe it won't happen at all, we are almost a full year out from the last Intel Mac released after all. But, and this is a big but - Intel has already stopped production on a lot of 3647 socket Xeons, some as long ago as a year. Apple may be forced to switch to the newer line in the mean time if Intel is planning on killing off production of the current ones, which isn't out of the question. This is why the iMac Pro was totally discontinued, and I don't think they'll do that with the Mac Pro.

 

[Yebubbleman]

You’re contradicting yourself a bit. If you need to connect a $3k card, a $150 breakout box is not a huge obstacle.

It's not a huge obstacle relative to the cost of the card, but you're not giving any justification for the extra expenditure. You'll never win the spending $150 vs. not spending $150 debate with no justifiable reason for spending $150 other than to revive a design that was largely rejected by the target market audience it was aimed for DUE TO THE AFORMENTIONED REASONS.

The point was that manufacturers would build a Thunderbolt variant of the card.

Indeed! And we all saw how successful that approach turned out to be! 🤣

Thunderbolt is going to tack on that extra $150 no matter how you slice it and people aren't content to spend $150 extra unless they feel there's an actual benefit.

USB 4 will probably ease that burden a bit, but not enough to make it worth doing simply for being a radical alternative to how workstation towers have been all this time. There has to be a practical benefit too.

Yes, a panel or panels.

Not a standardized slot, but more of like Made for Mac Pro program that’s similar to what Apple is doing with Lightning. It wouldn’t make sense for nvidia to create a hot card that would harm the performance of other components. So Apple would have to set rules for cards.

The MFi program is successful as a CONSUMER platform. It would fail miserably as a PROFESSIONAL platform. Hell, the 2013 Mac Pro is evidence of it having already happened. Again, why do it again?

Well, exactly. The Mac Pro is a niche within a niche. Why have two models of that?

Workstations and Desktops don't share the same niche. They never have. Nor have they ever been marketed that way.

As for the answer to the question of "why have two models of that?": Apple NEEDS to treat Mac Pro customers and high-end Mac customers more carefully than they do customers of lower-end Macs. The 16" MacBook Pro and 2019 Mac Pro launches at the end of 2019 were basically apology tours catering to Pros who bought 2013 Mac Pros and butterfly-keyboard-based 15" MacBook Pros and had a horrible time with them. You can tell a home user with only a handful of apps that are not yet Apple Silicon native to just use Rosetta 2 far more easily than you can someone using a Mac for high end video applications with zillions of plug-ins (many of which won't be updated to be Apple Silicon native for a LONG time). That said, not every Mac Pro user's set of applications are going to be the same degree of not Apple Silicon Native. Maybe I need a Mac Pro for my Final Cut Pro and Motion workflows for the performance factor and all of the plugins I use are Apple Silicon native; why would I need to get the Intel version? Maybe I need a Mac Pro for my Adobe Premiere and After Effects workflows for the performance factor and none of the plug-ins I use are Apple Silicon native. Given the high cost of the Mac Pro, it is silly for Apple to only cater to one of those two hypotheticals, especially at the expense of the customer who would be shelling out the $6000-50000 for whichever Mac Pro was better for them.

(Maybe there's a lot of feedback that, actually, some people really did prefer the 2013 form factor, so now they want to please both crowds?)

Apple doesn't like admitting when they're wrong and they'll avoid it at every chance they can. The 2019 Mac Pro (with the iMac Pro as a stop-gap update geared at 2013 Mac Pro customers and those that were resorting to using 27" iMacs to cover the difference for those high-end workflows) was such an admission in response to the collective distaste of the 2013 form factor. Again, they did similarly with the 16" MacBook Pro in direct response to the butterfly keyboard equipped Touch Bar 15" MacBook Pros. They knew they had pissed off that segment of the Mac user base and that they were losing customers to Windows. Prior to the 2019 Mac Pro, it was far more sensible to get a Windows-based Workstation tower than it was a 2013 Mac Pro or iMac Pro, especially if expansion was a must. Similarly, pre-16" MacBook Pro, the Touch Bar MacBook Pros were a joke; I was in the market for a MacBook Pro then and I didn't buy it because they were so damn bad and unreliable.

So, no, while there may have been a lot of feedback from people that liked the 2013 Mac Pro design, it was FAR outweighed by people who hated it; otherwise Apple would not have reversed course.

Sure they are. For a lot of those use cases, smaller machines now have plenty performance, or you rent stuff in the cloud. Workstations are a much rarer breed than they used to be.

You're thinking of very specific workloads. And, while you're not at all wrong that those workloads are made easier by taking the same or better hardware and running it from a cloud vendor rather than in your own environment, there are still plenty of workloads and situations for which this isn't practical or even feasible. In fact, it's usually those workloads and situations that necessitate the higher-end Mac Pros that cost a large five-digit figure and are used in high profile industries such as media. In many cases, running those systems via a cloud instance is neither practical nor safe for the business.

Yes, I get that. But what about others? If you're, say, a developer, why would you buy a computer that has one CPU and two GPUs? It's a bizarre setup for that.

Mac Pros were largely overkill for those particular workloads to begin with. That's why Mac mini cloud instances on AWS and MacStadium are such a big deal. Macs with two GPUs are primarily geared for GPU-compute intensive tasks and/or high end content creation, not software development.

Next macOS ARM only most likely
So why blow 6k to 10k dollars?
Save for retirement in this very uncertain world.

First off, you're completely wrong. Apple only excluded 2013 non-Mac-Pro Macs and 2014 non-Mac mini Macs from the pool of Big Sur compatible Macs. Macs from 2015 with 5th Generation Intel and beyond are all still supported. Odds are extremely good that Apple will either not change this bar of entry next year or only change it so much (maybe requiring 2016 and newer Macs or 2018 and newer Macs or T2 and M1 Macs), but certainly they won't do M1 and newer. This is not 2009. This is not PowerPC to Intel. Nor is Tim Cook Steve Jobs.

Secondly, The Mac Pro isn't aimed at ordinary joes like you and me. It's aimed at either people for whom this is a tax write-off because it's an essential tool of the trade or it's for businesses that have staffed creatives that need the kind of performance that you won't get on a maxed out 2020 Intel 27" 5K iMac. Incidentally, where average joes are fine because they use common variety software from the likes of Adobe and Microsoft with minimal plug-in usage (and can therefore work satisfactorily on a maxed out M1 Mac mini), the kinds of people that need this kind of Mac have a TON of software on top of that. Many $3000+ plug-ins made by mom and pop software outfits that will probably need MUCH longer to port their stuff over to Apple Silicon. If you're rocking an iMac Pro or a 2013 Mac Pro, you're probably giving thought to an upgrade right around now, and an Apple Silicon machine at this stage of the developers end of the transition (there's still a TON of software that's not Apple Silicon native yet) would be a horrible idea for stability and dependability, which is more crucial for those Mac users than it is for ordinary joes like you and me.

Still. To spend this kinda money when you know in the future Intel support will end.

How is this different from buying any other Mac during periods of time wherein Apple ISN'T going through a processor architecture transition? Anyone who lept at the opportunity to buy the first 16" MacBook Pros in late 2019 knew that what they were buying would eventually lose support from Apple.

Furthermore, Apple is dropping Intel Macs from support for the same reasons they did before the Intel switch; either Apple doesn't have enough support from component manufacturers to produce updated drivers for a given Mac's component(s), or Apple has a fundamental OS change (e.g. Metal) that requires newer Intel hardware to support. For Big Sur, it was Intel ditching firmware level support for Ivy Bridge Macs and the 2013 iMacs using a Wi-Fi board the manufacturer of which wasn't providing Apple with what they needed to update the driver. For Catalina, it was that Apple didn't want to have to keep supporting 9 year old Mac Pros that HAD to have an aftermarket video card in order to work in the OS. For Mojave, it was that anything older than Ivy Bridge's Intel HD 4000 iGPU or anything with AMD Radeon HD 6000 series and older NVIDIA/AMD discrete GPUs wouldn't support Metal which was fundamental to the OS's architecture. To my knowledge, we don't yet know what caused Apple to allow things like the 2014 Mac mini and the 2015 Mid 2015 MacBook Pro, which are still Haswell Macs - but not any of the other previous Haswell Macs - into Monterey. Similarly, we don't yet know what about the 2013 Mac Pro allows it into the party that any other 2013 or 2014 Mac doesn't have also.

Also, we're not talking about a 21.5" iMac or a 13" 4-port MacBook Pro here, we're talking about a Mac that people depend on to get complex tasks done involving software programs and plug-ins that aren't just ported to a new Apple architecture overnight. Telling someone to just jump in on Apple Silicon and that they'll be fine with Rosetta 2 showcases a lack of understanding of the nature of those programs, plugins, and the developers that produce them (and how slowly they tend to operate).

I dont understand? Why push expensive HIGH POWER workstations when we all know the future is CLOUD COMPUTING and running applications from a web browser on say a fast 6G internet connection and monthly rental of software.

Because that doesn't work with high end video workflows. It never has. That being said, Apple makes a rack mount version of the Mac Pro that surely is being used in those ways. But it's also that cloud instances of macOS are relatively new (and mainly because Apple's business model with the Mac isn't structured to encourage this in the way that Microsoft's is with Windows).

If the PowerMac G5, machines limited to Mac OS 9, and the "granddaddy architecture transition of them all" to PowerPC from 68k are any indication, then there is no way in hell Apple has any intention to support Intel Macs, even with just security updates, 8 years into the future. That is a foolish bet to make given Apple's history. (And yes, I still feel burnt by my Quad G5 purchase.)

The PowerPC to Intel transition happened during a very different era of personal computing. Macs would last an average of five years before losing support. The early Core Duo iMacs lasted exactly five years before losing support for a new OS (and then another two thereafter before losing security update support altogether). Now Macs are getting a good seven or eight years of being able to run the latest macOS release (with the two years of extended security update support thereafter). Apple is also a larger company now with many more Mac customers that they have to worry about pissing off than they had back then. Furthermore, your Power Mac G5 Quad purchase was probably a much safer buy in 2005 than the first release (1,1) of Mac Pro would've been in 2006 given that app compatibility with Intel would continue to be an issue for years to come thereafter. Apple has already shown that they're keeping support for Intel Macs in these new OSes. The only features Intel Macs are excluded from in macOS Monterey are features that leverage the neural engines that are not found in T2 based Intel Macs.

Regardless of future support (at most, I can envision 3 more years of OS support and maybe 5 years of security updates for Intel systems once a Mx-based Mac Pro is released), for the clients who need the most powerful Macs (i.e. not your typical customer and around just the 1%), then updating the Mac Pro to the latest Ice Lake chips makes a lot of sense. The real-world performance increase on Ice Lake is HUGE. It's really hard to state just how much of a performance increase we're seeing on Ice Lake chips under Linux at least, and if Apple doesn't update the CPUs in the Mac Pro, then it will be the trash can situation all over again – but it's also not like Apple always seems to care too much about that either, so who knows...

Three years is too conservative. They're only dropping support for Intel Macs because they either can't produce an updated component driver that works with the new OS (due to the manufacturer having ended support for the given component) or when a fundamental OS-wide feature requires hardware support that just isn't there on a given older Mac (e.g. Metal). Again, with Big Sur it was Intel dropping firmware level support for Ivy Bridge and Apple being unable to update the Wi-Fi driver on the Wi-Fi adapter used in the 2013 iMacs; with Catalina, Apple didn't want to have to keep supporting 9 year old Mac Pros with aftermarket video cards; with Mojave, Metal was fundamental to the OS's operation and therefore Macs needed a GPU that supported it. They're not going to get to the point of dropping Intel until they see that enough people no longer use Intel Macs regularly and when it becomes easier/better for the platform to jettison support for it and not a second sooner. That's how/why Apple drops support for things.

 

[Yebubbleman]

I agree with you about the lower core count chips not being starved for memory bandwidth, I just don't think Intel is going to use the same socket for 5 years unless it's a really specific, profitable niche which this single socket product isn't.

The kinds of sockets that Xeons and Extreme Edition Core i9 and Core i7 CPUs use are usually in use for AT LEAST 5 years. The kinds of consumer level sockets in use on standard Core i3, Core i5 Core i7 and Core i9 desktop CPUs (and in things like the Intel 27" iMacs) are usually in use for 2 years. Xeons are much longer term products than consumer-grade Core i5/i7/i9 processors.

The only comparable thing I can think of is when they released Kaby Lake-X on the same platform as Skylake-X which had to be some specific cost decision for the low core count parts, but I don't think Intel's current (much better) leadership would repeat that mess.

If an update does come, I don't expect XCC cpus to be compatible or at least sold by Apple. The talk about Dual CPUs etc. is also crazy, Apple is not going to go down that road with an x86 product that is probably the last of its kind. We're going to get the same case, same PSU, updated architecture with faster CPUs / memory and a couple new workstation cards from AMD - probably nothing else at all. Folks talking about holding back a product because it will make a new Apple CPU look bad are totally wrong-headed, we're not going to see an Apple CPU workstation that costs $15,000-$20,000+ soon, if ever, and even this updated Mac Pro will probably not hit the single core CPU geekbench that the current M1 in a MacBook Air does..

If what people are saying about Ice Lake-SP are to be believed, then it will definitely beat single-core performance M1 scores. It won't necessarily beat M2/M1X/M2X/M3/M3X scores, but that's not really the point. The point is to ensure compatibility for those that need it today

There will be M2 or M3 iMacs that destroy this potential Mac Pro that cost 1/5 the price within a few years at most, but for the people that need the expansion and full x86 compatibility this would be a great swan song.

Exactly. That's why an Ice Lake-SP Mac Pro is not a totally bad or even far-fetched idea.

I even think the Afterburner was probably some internal engineering experiment that they decided to release since it was so profitable for them - just looking at the refurb 2019 Mac Pros over the past year, most of them include it by default to pad the $ and increase profit margins on an already high margin product, so I don't expect any updates there either. It would have been really nice if they released Dev tools for it, but my hunch is they got a great deal on those ASICs and wanted to push creators further toward ProRes which M1 and certainly future M cpus will accelerate very well. Higher margins and funneling people toward their own ecosystem - win/win!

I really hope this update happens, even though almost no one will buy it but me.

I don't think you'd be the only one buying such a system. Just that anyone who isn't 100% gung ho about Apple Silicon (likely out of a professional necessity) is the target market audience of this would-be product and that those who either have all of their programs and plug-ins 100% Apple Silicon native or are completely okay with using Rosetta 2 to fill in the gaps are the ones who will be better served by an Apple Silicon Mac instead.

The vast majority of vocal commenters in these forums are consumer oriented and not pro-focused as is evidenced by the sheer volume of "why would anyone buy another Intel Mac in 2021" posts.

That's also why maybe it won't happen at all, we are almost a full year out from the last Intel Mac released after all. But, and this is a big but - Intel has already stopped production on a lot of 3647 socket Xeons, some as long ago as a year. Apple may be forced to switch to the newer line in the mean time if Intel is planning on killing off production of the current ones, which isn't out of the question. This is why the iMac Pro was totally discontinued, and I don't think they'll do that with the Mac Pro.

The 2017 iMac Pro was most likely discontinued because the Xeon was being EOLed by Intel. It was most likely not replaced with anything because it was largely redundant with the Mac Pro covering the high-end needs of iMac Pro customers while the 2020 27" iMac covered the low-end needs of iMac Pro customers. That being said, I wouldn't necessarily rule out the Apple Silicon replacement of the 2020 27" iMac being branded "iMac Pro" Certainly the two sizes of iMac have definitely diverged over the years in terms of under the hood performance and even intended use cases.

 

[deconstruct60]

I agree with you about the lower core count chips not being starved for memory bandwidth, I just don't think Intel is going to use the same socket for 5 years unless it's a really specific, profitable niche which this single socket product isn't.

Just because the pin count is the same don't means it is the same completely compatible Intel socket. Intel has moved the notch around in the past.

The microarchtectures + chipset pairs have shifted with the same pin count.

If the memory channels and UPI links stay the same then the pin count largely staid the same. That doesn't mean all the speeds on the pins stay the same or the data paths stay constant.

If an update does come, I don't expect XCC cpus to be compatible or at least sold by Apple. The talk about Dual CPUs etc. is also crazy, Apple is not going to go down that road with an x86 product that is probably the last of its kind.

Nevermind that it really doesn't fit inside t he basic Mac Pro chassis. The higher than average number of slots squeezes out board space for the another addition set of DIMMs. The DIMMs are already on the back side of the board.

there isn't room for another socket if keep the same basic board lay out.

We're going to get the same case, same PSU, updated architecture with faster CPUs / memory and a couple new workstation cards from AMD - probably nothing else at all. Folks talking about holding back a product because it will make a new Apple CPU look bad are totally wrong-headed, we're not going to see an Apple CPU workstation that costs $15,000-$20,000+ soon, if ever,

Errr. I think the "half sized" Mac Pro could climb close to that high. If pick every whiz-bang BTO selection for the iMac 27" can creep up to $8K. I think the iMac Pro got pretty close to $11K. If it is all soldered to SoC RAM it won't be cheap to jump up 128GB. Ditto for the storage on maximum.
Go big on iGPU and card cost will wiggle into view. If there is one open slot for Afterburner that's another $2K. Get two that's another $4k. Some $400 wheels. etc. [ don't even have to weave any AMD MPX modules into the mix and still can go quite high.]

The leaks/rumors about the Jade4C are either a relatively very large monolithic singe die or somewhat exotic multiple chip package with 2.5D or 3D features. And the few of the 4C options Apple makes the more likely they'll throw a "low volume tax" on it.

The Mac Pro 2013 got smaller and more expensive than the Mac Pro 2010 at the same time.

The entry level price around the old iMac Pro entry price there would be far more affordable options to pick from than $15K.

and even this updated Mac Pro will probably not hit the single core CPU geekbench that the current M1 in a MacBook Air does.. There will be M2 or M3 iMacs that destroy this potential Mac Pro that cost 1/5 the price within a few years at most, but for the people that need the expansion and full x86 compatibility this would be a great swan song.

As Apple is likely to drop the lower core counts ( 8 & 12 ) probably won't be a single threaded drag racing oriented system.

So far Apple has done exceeding little to reduce the Mac System prices with any M-series conversion. They have taken the cost savings fokls have imagined for dropping Intel and either applied it to other more costly subcomponents or just pocketed it.

If talking about how M3 powered Mini several years out will be faster than the new entry-mid range systems. Perhaps on a few things. With 3rd party GPU driver updates on apps that do GPGPU computations... probably not. if insert an upgrade GPU in 2-3 years time.

Apple fumbling the ball on eGPU support for the next 2-3 years mean the M-series Macs will be cut off ( natively run iPhone apps just great but on a non progressive path on compute ).

I even think the Afterburner was probably some internal engineering experiment that they decided to release since it was so profitable for them - just looking at the refurb 2019 Mac Pros over the past year, most of them include it by default to pad the $ and increase profit margins on an already high margin product, so I don't expect any updates there either. It would have been really nice if they released Dev tools for it, but my hunch is they got a great deal on those ASICs and wanted to push creators further toward ProRes which M1 and certainly future M cpus will accelerate very well. Higher margins and funneling people toward their own ecosystem - win/win!

Shouldn't really be a guess. Afterburner was all about ProRes. Some of the Afterburner logic implementation may go into fixed function media support engines for M series in 2-3 generations down the road if there is transitor budget for that.

I doubt it was some random experiment. The lawsuit with RED over patent rights to aspects of ProRes RAW probably meant that the format was in flux. An FPGA meant they could adjust later if legal stuff went into a direction they really didn't like to get stuck in. But once the logic is settled and has a proven track record then it can go ASIC (and possibly swallowed up into the main SoC for a lower number of concurrent streams. )

Lots of folks on the outside suggested that Afterbuner was going to be about covering old RED Rocket for RED and other niche codecs. Apple never really said anything remotely close to that. They said there was some flexibility, but that was always in the context of ProRes updates. ( e.g., if pick up Apple camera RAW or Apple did something so that new ProRes variant came out of Apple cameras. )

I really hope this update happens, even though almost no one will buy it but me. That's also why maybe it won't happen at all, we are almost a full year out from the last Intel Mac released after all.

Are they? If Apple does a "half sized" Mac Pro M-series model they could just declare transition victory.
It is buried on the 27" intel iMac buy button , but Apple is still selling the non-Retina 21.5" model powered by a now quite ancient MBA processor. The Mini has a "Intel Sidecar" model . ( probably until the SoC for the MBP 16" comes out).

With a PCI-e v4 Mac Pro backplane they could release "refreshed" models with better MPX modules for a while. (e.g., the 2012 model that had some basic motherboard as the 2009 model )

But, and this is a big but - Intel has already stopped production on a lot of 3647 socket Xeons, some as long ago as a year. Apple may be forced to switch to the newer line in the mean time if Intel is planning on killing off production of the current ones, which isn't out of the question. This is why the iMac Pro was totally discontinued, and I don't think they'll do that with the Mac Pro.

The iMac Pro dead-ended because who would by more expensive W-2100 options when there are substantive cheaper W-2200 options that Apple never made the jump to. Plus there was lots of performance overlap with the max BTO iMac 27". There was about as big of a hiccup of what to jump to for a GPU (due to AMD supply slips ) on iMac Pro as there was the Intel CPU aspect. But intel doing TDP creep ( both W-2200 and W-3200 options would have run even hotter) made them a bad match for the iMac Pro case limitations.

Pretty good chance Apple is waiting on something of their own to fill that similar zone.

 

[deconstruct60]

If what people are saying about Ice Lake-SP are to be believed, then it will definitely beat single-core performance M1 scores. It won't necessarily beat M2/M1X/M2X/M3/M3X scores, but that's not really the point. The point is to ensure compatibility for those that need it today

if Ice lake baseline were so great at single threaded why did Intel do Rocket lake. I think the "jump in performance" is more gear to mixed and more pure multithreaded loads than the highly focused single theaded Geeky benchmark races.

Single thread.
https://www.anandtech.com/show/16594/intel-3rd-gen-xeon-scalable-review/7

on some single thread spec test cases there are losses with Ice Lake. Multhreaded there are more consistent wins.
If have something that neatly flows into a solid AVX workload there are bigger wins. But 'dick , jane , spot' single threaded drag racing code there are not very huge wins there.

The 2017 iMac Pro was most likely discontinued because the Xeon was being EOLed by Intel. It was most likely not replaced with anything because it was largely redundant with the Mac Pro covering the high-end needs of iMac Pro customers while the 2020 27" iMac covered the low-end needs of iMac Pro customers. That being said, I wouldn't necessarily rule out the Apple Silicon replacement of the 2020 27" iMac being branded "iMac Pro" Certainly the two sizes of iMac have definitely diverged over the years in terms of under the hood performance and even intended use cases.

Even more so if Apple doesn't "thin out" the iMac 27" by putting the constraint that all the electronics have to fit into the chin and Ethernet jack be banished to the power brick because too thin to plug in a cable. ( and earphone jac move to the side because of the same reason; plug thicker than the machine).

I think Apple has less of a problem with processor performance overlap with the Mini and iMac on the same M1 at this point. A shift coming to where same M-series SoC in more products.

A 27" iMac with Jade2C and jade4C could overlap with a "half sized" Jade4C Mac Pro. As long as these are Apple iGPU only and no 3rd party GPU drivers the performance variation is going to be more uniform. If "thin out" the Jade2C enclosure then perhaps iMac Pro and Mac Pro share the jade4C.

 

[Stephen.R]

will spin up dedicated GPUs for higher end workflows.

I'm not involved in pro video, or 3d or anything that uses a high end video card (besides, you know, driving high resolution displays) so take this with a grain of salt.... but this just sounds like it'd be more akin to the Afterburner card in the current Mac Pro.

Not related at all to driving displays, just there as specialised hardware for specialised tasks.

 

[farewelwilliams]

It's not a huge obstacle relative to the cost of the card, but you're not giving any justification for the extra expenditure. You'll never win the spending $150 vs. not spending $150 debate with no justifiable reason for spending $150 other than to revive a design that was largely rejected by the target market audience it was aimed for DUE TO THE AFORMENTIONED REASONS.

- The ability to use the accessory in other Apple products like iMac Pro or an on-location shoot with a MacBook Pro
- The ability to provide more than 1500 watts to a graphics card in the future without buying a new Mac Pro with a bigger PSU (should they make one)
- Potentially better cooling, keeping heat away from the main tower
- In an environment where floating software licenses are used a lot, it's very handed to just unplug the card and move it to another edit bay without turning off or on the computer.
- 6 external PCI-e devices via Thunderbolt ports, each with daisy chaining means you're never out of ports to connect professional equipment. Right now if you use 2x Vega II MPX modules, you're stuck with 3 PCI-e ports left which then you have to buy external PCI-e enclosures anyways.

Plenty of reasons.

Indeed! And we all saw how successful that approach turned out to be! 🤣

The fact that Mac Pro wasn't updated for several years and the fact that it was Thunderbolt 2 were major factors of lack of adoption. Not that an external PCIe via Thunderbolt 3 was fundamentally flawed.

Thunderbolt is going to tack on that extra $150 no matter how you slice it and people aren't content to spend $150 extra unless they feel there's an actual benefit.

Again, $150 isn't a factor due to benefits listed above.

USB 4 will probably ease that burden a bit, but not enough to make it worth doing simply for being a radical alternative to how workstation towers have been all this time. There has to be a practical benefit too.

Not sure what burden USB4 is going to ease with respect to external PCI-e considering it's just a confusing Thunderbolt port.

The MFi program is successful as a CONSUMER platform. It would fail miserably as a PROFESSIONAL platform.

Agree to disagree.

Hell, the 2013 Mac Pro is evidence of it having already happened. Again, why do it again?

We are talking about what Apple intended with the 2013 Mac Pro. I don't know what you mean by do it again. No where did I suggest that.

 

[Stephen.R]

Not sure what burden USB4 is going to ease with respect to external PCI-e considering it's just a confusing Thunderbolt port.

I don't know what the original poster was referring to, but the biggest change I can see relating to this, is that USB4 allows for docks/hubs that have multiple downstream USB4 ports (as opposed to the limit of just daisy chaining, in TB3). It doesn't increase your bandwidth obviously, but it would allow for adding more external devices.

 

[Yebubbleman]

if Ice lake baseline were so great at single threaded why did Intel do Rocket lake.

We're not talking about standard Ice Lake (the likes of which we see in the still-sold 2020 Intel 4-port 13" MacBook Pro). We're talking about Ice Lake-SP. And comparing Ice Lake-SP to Rocket Lake, if my understanding is even remotely correct, is like comparing a Toyota Tacoma to an 18-wheeler.

I think the "jump in performance" is more gear to mixed and more pure multithreaded loads than the highly focused single theaded Geeky benchmark races.

Single thread.
https://www.anandtech.com/show/16594/intel-3rd-gen-xeon-scalable-review/7

on some single thread spec test cases there are losses with Ice Lake. Multhreaded there are more consistent wins.
If have something that neatly flows into a solid AVX workload there are bigger wins. But 'dick , jane , spot' single threaded drag racing code there are not very huge wins there.

The gains relative to M1 are, again, a moot point. Someone who can better be served by an Apple Silicon Mac Pro, if given the choice between that or an Intel one, will gravitate toward that direction, while someone who can better be served by an Intel Mac Pro today will, out of necessity, go that route instead.

Even more so if Apple doesn't "thin out" the iMac 27" by putting the constraint that all the electronics have to fit into the chin and Ethernet jack be banished to the power brick because too thin to plug in a cable. ( and earphone jac move to the side because of the same reason; plug thicker than the machine).

I think Apple has less of a problem with processor performance overlap with the Mini and iMac on the same M1 at this point. A shift coming to where same M-series SoC in more products.

I see there being a consumer grade SoC (M1, M2, etc.) for the MacBook Air, Mac mini, smaller iMac, and maybe a smaller MacBook Pro (depending on how Apple decides to handle replacements to both of the currently sold 13" MacBook Pros with respect to what they do with the 16" MacBook Pro's Apple Silicon replacement), a prosumer grade of SoC (M1X, M2X, and/or a different letter etc.) for the 16" MacBook Pro (if not the 14" as well) and a professional workstation-class grade of SoC (P1, P2, and/or a different letter etc.) for the Mac Pro. They could close the gaps on the consumer vs. prosumer disparity, but it's not like they don't have a similar ability to do so with the iPad Air and opt not to in favor of only giving the iPad Pro models the beefier SoC.

A 27" iMac with Jade2C and jade4C could overlap with a "half sized" Jade4C Mac Pro. As long as these are Apple iGPU only and no 3rd party GPU drivers the performance variation is going to be more uniform. If "thin out" the Jade2C enclosure then perhaps iMac Pro and Mac Pro share the jade4C.

 

[Yebubbleman]

- The ability to use the accessory in other Apple products like iMac Pro or an on-location shoot with a MacBook Pro

I'm not saying that breakout boxes are bad. I'm saying that removing slots from a Mac Pro is bad. I'm also saying that, specifically in the context of the Mac Pro, the argument of breakout boxes over slots is bad since there's no benefit and you spend money you wouldn't otherwise need to spend.

- The ability to provide more than 1500 watts to a graphics card in the future without buying a new Mac Pro with a bigger PSU (should they make one)

Moot for a variety of reasons. One being that non-Apple GPUs are not likely to be a thing for Apple Silicon Mac Pros. But another being that a graphics card requiring 1500 watts in and of itself likely will require way more bandwidth than USB 4 or Thunderbolt 3/4 can realistically provide. MPX modules already use greater bandwidth than PCIe provides naturally which is how Apple can hang Thunderbolt 3 ports off of those graphics cards.

- Potentially better cooling, keeping heat away from the main tower

That's never really been a problem for Mac Pro towers.

- In an environment where floating software licenses are used a lot, it's very handed to just unplug the card and move it to another edit bay without turning off or on the computer.

Floating licenses makes sense. Floating cards, especially when you typically have more than one person needing to use that card at any given time, doesn't. You either will have a dedicated machine that people share or you have a traveling licensing dongle that travels around. But floating cards is not a super common use case in the media companies that I've been an IT admin at.

- 6 external PCI-e devices via Thunderbolt ports, each with daisy chaining means you're never out of ports to connect professional equipment. Right now if you use 2x Vega II MPX modules, you're stuck with 3 PCI-e ports left which then you have to buy external PCI-e enclosures anyways.

Right, but, to your point, you still get those Thunderbolt ports as part of the MPX modules, so really you get the best of both worlds. Again, I'm not saying that breakout boxes are bad. I'm saying that advocating for them in a 2013 Mac Pro style system would cause more problems than it would solve as is evidenced by the 2013 Mac Pro's reception and prolonged infamy.

Plenty of reasons.

Right, and I've poked holes in most of them.

The fact that Mac Pro wasn't updated for several years and the fact that it was Thunderbolt 2 were major factors of lack of adoption. Not that an external PCIe via Thunderbolt 3 was fundamentally flawed.

There's nothing fundamentally flawed about external PCIe expansion via Thunderbolt 3. For those with MacBook Pros, iMacs, and Mac minis, as you've said, it's a fantastic ability.

There is A LOT fundamentally flawed about removing PCIe slot expandability from a Mac Pro. The move from Thunderbolt 2 to Thunderbolt 3 doesn't fix that problem. If it did, Apple could've easily released a 2016 or 2017 Mac Pro with the same cylinder form factor but with six Thunderbolt 3 ports instead of six Thunderbolt 2 ports and that would've resolved everything. Except that's not how it all went down. The iMac Pro was the closest thing that we got to this and it was seen by Apple as a stop-gap measure to bide them the time that they needed to release the 2019 Mac Pro.

Again, $150 isn't a factor due to benefits listed above.

It is when you could've otherwise had that same expandability without spending that money. Again, I'm not sure where that point is lost on you. Unless you're talking about PCIe to Thunderbolt bridging outside of the context of the Mac Pro, which I'm not. I'm only talking about it in the context of the Mac Pro.

Not sure what burden USB4 is going to ease with respect to external PCI-e considering it's just a confusing Thunderbolt port.

Licensing. That was a big part of why Thunderbolt 1/2/3 Mac accessories were so much more costly than their USB 3 and USB-C counterparts.

Agree to disagree.

Again, the 2013 Mac Pro proves as much.

We are talking about what Apple intended with the 2013 Mac Pro. I don't know what you mean by do it again. No where did I suggest that.

Are you not advocating in favor of Apple producing a Mac Pro that lacks PCIe and asserting that the 2013 Mac Pro would not have been as much of a failure if Thunderbolt 3 was in tow instead of Thunderbolt 2? Because most of my debate with you here on $150 breakout boxes is predicated on that notion.

I don't know what the original poster was referring to, but the biggest change I can see relating to this, is that USB4 allows for docks/hubs that have multiple downstream USB4 ports (as opposed to the limit of just daisy chaining, in TB3). It doesn't increase your bandwidth obviously, but it would allow for adding more external devices.

Again, licensing. USB4 doesn't require that anyone pay licensing or royalty fees to Intel. Thunderbolt 3 and Thunderbolt 4 both do.

 

[MacBH928]

I agree with the first part and disagree with the second. There are FAR more Intel Macs in the world today that are still running the most current software compared to the last time Apple switched architectures, and I'd be surprised if the support of both Apple and developers doesn't continue to be strong.

Apple already drops support for 2-3 year old OS. So all Apple computers today will lose support within 5 years or so, and once Apple pulls the plug so will the developers.

Some developes will choose to continue to support older hardware and OS. You can always install Windows or Linux on it too so maximise the use of the hardware. or keep it as a dedicated machine for like view videos or as a server.

 

[Yebubbleman]

Apple already drops support for 2-3 year old OS. So all Apple computers today will lose support within 5 years or so, and once Apple pulls the plug so will the developers.

That logic doesn't follow at all. Apple drops support for three year old OSes. You are correct about this. But that doesn't have anything to do with the decision Apple makes to drop or not drop support for a given Mac from that next macOS release and therefore your five year figure for current Macs makes no sense, especially since there are seven year old Intel Mac minis and eight year old Mac Pros that are going to be able to run macOS Monterey.

 

[chucker23n1]

(Ice Lake-W 3300 is LGA 4189, when Cascade Lake-W 3200 in the 2019 Mac Pro was LGA 3647.)

 

[chucker23n1]

if Ice lake baseline were so great at single threaded why did Intel do Rocket lake.

Mainly because Intel had trouble producing 10nm in volume. They slowly ramped it up, and started with the biggest pain point: laptops.

Rocket Lake is a stopgap. As were Kaby Lake and Coffee Lake and Whiskey Lake and Amber Lake and I’m sure I’ve forgotten one before it.

 

[mode11]

For high powered desktop graphics, GPU would likely not be on the same chip as the thermal envelope would be too limiting.

M1 uses DDR4X and is shared among CPU/GPU/etc...likely a Mac Pro graphics by Apple would not tap into the unified memory, but instead have a separate DDR6 pool of memory.

Or rather, there will still be a GPU on the SOC, but that would be "low powered" GPU and will spin up dedicated GPUs for higher end workflows.

That all makes a lot of sense, but isn't all the buzz about "Jade 4C" and the lack of references to PCIe graphics in recent Apple developer docs? Normally, extrapolating the MacBook Air's M1 to the whole range would be silly (lower end machines have long used integrated graphics). But the low TDP and high performance of the M1 suggests a version the size of an high-end Nvidia GPU die might be 'fast enough' for a Mac Pro.

Does seem a bit weird though - it's a very large, custom part for a tiny market, and a long way from the iPhone SoC it's leveraging. And cutting off the possibility of PCIe GPUs also seems odd. Apple would also need to iterate on their MP SoC pretty often to stay competitive with PC GPUs, something they would be loathe to do given the costs. OTOH, if the Mac Pro is perennially behind in GPU, it won't be an attractive platform.

 

[mode11]

The 2019 Mac Pro (with the iMac Pro as a stop-gap update geared at 2013 Mac Pro customers and those that were resorting to using 27" iMacs to cover the difference for those high-end workflows) was such an admission in response to the collective distaste of the 2013 form factor.

Personally, I think the iMac Pro was intended to be the Mac Pro's replacement. Perhaps after the 2013 MP's low sales were taken as 'proof' the demand wasn't there. Stick a Xeon in an iMac and call it a day. The decision to go back to a tower seemed unplanned, hence the emergency meeting with cherrypicked journalists, to tell pro Mac users a tower + monitor was on the way (keep the faith!). Perhaps the push-back over the 2016 MBP's lack of ports / very unreliable and hard to service keyboard / pointless emoji bar / lack of Escape key / loss of MagSafe / thermal throttling gave Apple a wake up call that there was a limit to how much pro customers could be dictated to.

To my knowledge, we don't yet know what caused Apple to allow things like the 2014 Mac mini and the 2015 Mid 2015 MacBook Pro, which are still Haswell Macs - but not any of the other previous Haswell Macs - into Monterey. Similarly, we don't yet know what about the 2013 Mac Pro allows it into the party that any other 2013 or 2014 Mac doesn't have also.

Yes, it's funny isn't it? Almost like they somehow can make exceptions for machines that sold in high volumes, or at high cost. Perhaps those technical challenges didn't prove insurmountable for Apple after all.

 

[Yebubbleman]

Personally, I think the iMac Pro was intended to be the 2013 Mac Pro's replacement.

I have modified your statement with the text in bold. I don't believe it was intended to be a wholesale Mac Pro replacement, especially since there were press events with Apple wherein they outright told the media that they were working on a Mac Pro that would suffice for Mac Pro customers' needs that weren't addressed with the 2013 model (2019 Mac Pro) and that they'd be putting out a high-end iMac in the then-nearer future to act as a stop-gap (2017 iMac Pro). But I definitely believe that they figured that for those that didn't have an issue with the 2013 Mac Pro's expansion situation, the iMac Pro would probably be a suitable upgrade.

Perhaps after the 2013 MP's low sales were taken as 'proof' the demand wasn't there. Stick a Xeon in an iMac and call it a day. The decision to go back to a tower seemed unplanned, hence the emergency meeting with cherrypicked journalists, to tell pro Mac users a tower + monitor was on the way (keep the faith!).

Oh, for sure. The switch back to a tower was something that didn't have a long-term plan on doing. It was basically a course reversal on what they had done in 2013. And, both due to design architecture limitations and user complaints, they had no choice but to start planning for it. I'm sure it was much easier to design the iMac Pro than it was the 2019 Mac Pro.

Perhaps the push-back over the 2016 MBP's lack of ports / very unreliable and hard to service keyboard / pointless emoji bar / lack of Escape key / loss of MagSafe / thermal throttling gave Apple a wake up call that there was a limit to how much pro customers could be dictated to.

Oh, most definitely. The 16" MacBook Pro was absolutely an apology tour, as was the 2019 Mac Pro and both came out at around the same time. Apple had five different models of Mac; their one workstation system and every notebook in their lineup, all with poor customer satisfaction. The iMac and the Mac mini were probably the only systems exempt from this. They had to start turning things around.

Yes, it's funny isn't it? Almost like they somehow can make exceptions for machines that sold in high volumes, or at high cost. Perhaps those technical challenges didn't prove insurmountable for Apple after all.

Honestly, I'd bet that it comes down to a Wi-Fi card, a storage controller, or something of that nature. It would make far more sense to just say "Broadwell CPUs or newer" otherwise (despite that doing so would cut out the 2015 15" MacBook Pro, but allow the 2015 13" MacBook Pro). Plus, aside from Force Touch trackpads and maybe Apple wanting to finally drop NVIDIA drivers from macOS once and for all (alongside whatever that AMD GPU used in the first 27" 5K Retina iMacs from 2014), there doesn't seem to be much rhyme or reason. I have a Mid 2014 15" MacBook Pro in storage that only has Iris Pro graphics and, aside from Force Touch on the trackpad and a different SSD, I'm not at all sure what the Iris Plus Mid 2015 variant has that allows it into the Monterey party that mine doesn't. (Not that I mind terribly; I've already moved on from it.)

 

[jonnysods]

True x86 support would probably be the predominate reason. Apple may be able to best it with raw CPU performance but there may be other first-gen limitations like memory, ports, and gpu options. - Which gives more credibility to a smaller Mac Pro being in the works that may be a peace offering to those that want more than an Apple Silicon iMac while leaving the full size Intel Mac Pro for those that need it.

Its probably the best enclosure Apple has to keep an Intel processor cool too - the iMac and laptops just struggle IMO so you can't get enough out of the processors.

I'm a little disappointed because I want to see the Xeon killer that Apple would need to conjure up to improve the Mac Pro. Or to see how they are going to tackle the discrete GPUs while using a SOC. Very curious!

 

[Yebubbleman]

Its probably the best enclosure Apple has to keep an Intel processor cool too - the iMac and laptops just struggle IMO so you can't get enough out of the processors.

I've never heard of these issues with the iMacs. Just the USB-C era MacBook Pros.

I'm a little disappointed because I want to see the Xeon killer that Apple would need to conjure up to improve the Mac Pro. Or to see how they are going to tackle the discrete GPUs while using a SOC. Very curious!

You ARE going to see it. The potential for an impending Intel Mac Pro refresh doesn't mean you won't still see the Apple Silicon Mac Pro SoC. Hell, you might even see both at the same time because they both might launch at the same time! But even if they don't, this is very clearly a stopgap meant to ensure that Mac Pro customers aren't left feeling like Apple is playing a nasty game of Lucy and the football while more software and hardware products adopt native Apple Silicon support.

 

[chucker23n1]

If what people are saying about Ice Lake-SP are to be believed, then it will definitely beat single-core performance M1 scores.

Intel promises a 20% IPC improvement. That still puts it well behind M1 on single-threaded tasks. It roughly competes with an A12 or A13.

Multi Thread? No.

Mac Pro 28 core Multithread 19552 https://browser.geekbench.com/macs/249
Mac Mini M1 Multhread 7412

that is 2.6x as fast advantage for Intel. ( -62% drop from the 28 core score ).

Threadripper 3990X 25016 https://browser.geekbench.com/processors/amd-ryzen-threadripper-3990x

An even bigger beat down (3.4x) . Couple more months and will have a next gen of that from both Intel and AMD.

Your point notwithstanding (an M1 is impressive at single-threaded tasks, not multi-threaded ones), think about how weak that is: that Threadripper needs 64 cores to achieve that performance. The M1 only has four performance cores.

There is lots of "it will scale perfectly linearly" from folks projecting that a "Much bigger M1 die " will catch the 28 core.

It won't scale linearly, but you can easily see that the M1 would need far fewer than 28 cores to meet the Cascade Lake-W, and certainly far far fewer compared to the Threadripper.

 

[mode11]

especially since there were press events with Apple wherein they outright told the media that they were working on a Mac Pro that would suffice for Mac Pro customers' needs that weren't addressed with the 2013 model (2019 Mac Pro) and that they'd be putting out a high-end iMac in the then-nearer future to act as a stop-gap (2017 iMac Pro).

There was only one press event that discussed the Mac Pro - the 2017 apology event that said a tower MP was coming and essentially asked people not to defect to Windows. The iMac Pro came out a few months later, so was surely already slated for release at that point.

Honestly, I'd bet that it comes down to a Wi-Fi card, a storage controller, or something of that nature.

Does Windows or Linux have this problem?

 

[chucker23n1]

Does Windows or Linux have this problem?

Not wanting to backport new features to old hardware? Yes, of course. That's always a cost-benefit thing.

 

[farewelwilliams]

I'm not saying that breakout boxes are bad. I'm saying that removing slots from a Mac Pro is bad.

You're changing the context. I was responding to "but you're not giving any justification for the extra expenditure." and "no justifiable reason for spending $150".

I'm giving reasons for the $150 expenditure.

Moot for a variety of reasons. One being that non-Apple GPUs are not likely to be a thing for Apple Silicon Mac Pros.

Talking about in the context for what Apple's initial intentions were for the 2013 Mac Pro.

That's never really been a problem for Mac Pro towers.

GPUs back in 2013 could absolutely throttle due to too much heat

Floating cards, especially when you typically have more than one person needing to use that card at any given time, doesn't. You either will have a dedicated machine that people share or you have a traveling licensing dongle that travels around. But floating cards is not a super common use case in the media companies that I've been an IT admin at.

Like I said, lack of Mac Pro updates and Thunderbolt 2 restricted the concept of floating cards. Therefore of course it's not a super common use case because it was never given a shot.

Right, but, to your point, you still get those Thunderbolt ports as part of the MPX modules

Which means you're buying PCI-E enclosures to plug it in.

Again, I'm not saying that breakout boxes are bad.

You literally said " not spending $150 debate with no justifiable reason ".

Right, and I've poked holes in most of them.

Nope. See above.

The move from Thunderbolt 2 to Thunderbolt 3 doesn't fix that problem.

Disagreed.

If it did, Apple could've easily released a 2016 or 2017 Mac Pro with the same cylinder form factor but with six Thunderbolt 3 ports instead of six Thunderbolt 2 ports and that would've resolved everything.

They didn't because of the thermal design for the GPUs. Not sure how you expect Apple to sell an updated Mac Pro four years later with only 1 new feature: "Thunderbolt 3".

Except that's not how it all went down.

See above.

It is when you could've otherwise had that same expandability without spending that money.

Except base price of the tower Mac Pro is $5999 while trash can was $2999. You literally are paying extra for the PCI-e slots.

Again, I'm not sure where that point is lost on you. Unless you're talking about PCIe to Thunderbolt bridging outside of the context of the Mac Pro, which I'm not. I'm only talking about it in the context of the Mac Pro.

See above.

Licensing. That was a big part of why Thunderbolt 1/2/3 Mac accessories were so much more costly than their USB 3 and USB-C counterparts.

Well no, Thunderbolt 3 became royalty free regardless of whether or not your accessory adopts USB4. This happened a couple of years before USB 4 was even implemented in Apple products. It's not "thanks to USB4" like you're framing it.

Again, the 2013 Mac Pro proves as much.

Mac Pro 2013 didn't implement a professional program, so no.

Are you not advocating in favor of Apple producing a Mac Pro that lacks PCIe and asserting that the 2013 Mac Pro would not have been as much of a failure if Thunderbolt 3 was in tow instead of Thunderbolt 2? Because most of my debate with you here on $150 breakout boxes is predicated on that notion.

I'm asserting that since Thunderbolt 3 was not given a chance due to lack of updates, you can't say for sure whether or not it would have failed.

 

[mode11]

Not wanting to backport new features to old hardware? Yes, of course. That's always a cost-benefit thing.

No, I was meant do Windows or Linux still support this 2014 wifi card. Not talking about supporting a Voodoo graphics card or something.

 

[chucker23n1]

No, I was meant do Windows or Linux still support this 2014 wifi card. Not talking about supporting a Voodoo graphics card or something.

What @Yebubbleman is saying that it could be some Wi-Fi chipset. Or another hardware component.

It might that:

• the company that originally did the chipset no longer exists, or isn't willing to expend the effort to offer new drivers
• new drivers exist, but they're a poor fit for Apple's use case, and Apple wasn't willing to expend the effort to adapt them
• they weren't allowed, by license, to adapt them, so they would've had to create a wrapper that would've been inefficient, making the user experience miserable

And all sorts of other reasons.

Ultimately, it usually boils down to costs and benefits. What Apple came down on, apparently, is: some people with an older Mac don't get some of the new features. Their computer will continue to run fine, and with the latest OS, just not with all of its whiz-bang stuff.

It's not like they're suddenly disabling Wi-Fi or something like that.

(OTOH, they did decide a few releases back that killing off subpixel rendering was A-OK.)