Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Except the M2 is faster.
Well, sort of. The majority of the speed bump comes from simply increasing the clock speed. A technique that came to the end of the Moore's Law road many years ago, due to the power draw and heat output issues. So basically, the M2 has used this old school technique to eek out a power gain without much in the way of genuine improvements. Not the technique one needs when using it for fan-less machines. Especially if the machines, such as the MBA, already have throttling issues with the M1.
 
Me too. I'm wanting to replace my late 2012 mini since it is stuck on Calalina and will stop getting security updates very soon. But since I keep my machines a long time I'm hoping if I can wait out the M2 mini it will get another year or two of updates before they stop supporting it.
Highly doubtful...security updates ending soon
 
  • Like
Reactions: Miat
I think what I'm reading is that they had the new case and internals ready to go for the new Mac Pro months ago, but were still working on the Extreme chip that it was designed for. They merely used the M1 for testing, but it was never going to be released without the Extreme chip.

I'm really looking forward to it. Not because I have any need for, or intention of ever owning one, but just for the excitement of seeing what it can do.

I'm particularly curious what they are going to do about the RAM in order to at least match the 1.5TB that the Intel Mac Pro can max out to.

My guess is that they will have both SoC RAM and plug-in, user-upgradeable, off-SoC RAM just like they do in the Intel.

l'm also guessing it will max out at 6TB RAM, partly because it will be running on LPDDR5 instead of 4, and partly because Apple will want the Mac Pro Extreme to blow the Intel Mac Pro out of the water.
 
Generally, workstations are bad long-term investments for individual customers. The performance edge is short-lived and the types of expandability and expansion reserved for the future don't really make economic sense as newer systems will offer better performance. A Mac Pro is something you really need to spec and buy specific to a current project (or type of projects) you are working on and move on to the next model when your new work requirements impact your ability to maintain your profit margins. I know a lot of people that spent small fortunes coddling cheese-grater Mac Pros for nearly a decade when they would have been better served by buying new high-end iMacs every few years. But sunk-cost reasoning would not let them let go of a machine they spent $5K+ on even when it's pointed out that the thousands of dollars in upgrades to keep them usable over the years still left them with an underperforming machine.
It depends on what your use case is. It doesn’t matter if the newest iMac has a processor that is ten times faster than in an old Mac Pro, if it doesn’t support the additional hardware requirements you need from the Mac Pro it doesn’t do you any good.
 
  • Like
Reactions: singhs.apps
Yes, that's my theory. The M1 Ultra had really bad scaling because of that bandwidth roadblock that was overlooked when they were originally designing the M1 family of chips.

(Chip engineers built the M1 Ultra/Extreme expecting the software to be properly optimized. Most developers took the easy route and never optimized their apps to properly manage the bandwidth and memory)


Apple ran a whole entire session at 2022 about optimizing and scaling GPU code in Applications.


It was not just the "chip engineers" that expect app developers to do their jobs well. It is closer to Apple expects developers to do their jobs well. Apple has rolled out more tooling to help with doing optimizations. With the tools and the tutorials it should more tractable for the less lazy to do something at this point. ( at least for the two die "Ultra" class solution. ). Apple expects developers to optimize their apps.

Apple is not particularly likely going to go do power bleeding, triple backward hardware somersaults trying to make badly optimized code run faster. The hardware is there. If the developers are using bubble sort where quick/merge would work better that it isn't Apple's job to 'fix' that.

Is Apple going to get perfect linear scaling with zero code optimizations across 2 and then 4 dies ? Probably not. AMD and Nvidia aren't with monolithic dies either.


Even if Apple made some improvements in "UltraFusion 2" to smooth out some very highly sensitive NUMA characteristics between 2 dies , it would likely pop back up at the 4 die coupling stage. So the "issue" isn't going to completely go away with hardware covering up dubious code assumptions.

CPU code there really isn't a huge problem for well crafted scaling algorithms. (e.g., the NASATruss benchmarks on Apple's Studio marketing page scale. the Adobe stuff doesn't. That is not an surprising at all. Not even in the slightest. Adobe is relatively very slow to optimization the bulk of their code base. That is not an hardware issue in the slightest. )




Because of that, I believe the M1 Extreme suffered from very bad scaling, so they scrapped it and are holding off until the M2 Ultra/Extreme are ready for the Mac Pro.
-Vadim from Max Tech

M2 or folk can use better Xcode tools and tutorials. Can't solve this issue solely with hardware. It time at least as much as hardware.

The M1 Extreme likely would have several other problematical issues besides GPU scaling. Economics ( four largish dies, multiple interposer fusion chips, more expensive packaging. etc. and yet much, much, much lower volumes. ) . Apple probably needs a die that isn't focused on being a MBP 16" chip. ( e.g. 4 TB controllers per die at the stage of a 4 die package is extremely likely at least 8 more TB controllers than you need. ) .

Do a 4 die package with TSMC N3 ( or N4P ) would make lots more sense to manage the overall package size. It M2 isn't bring magic sprinkles but should/could be done with far more appropriate tech that is independent of the microarchitectural issues. Bringing the Extreme back under the 300W zone will help with the operational environment for the package.

Unless Apple had a major addition for PCI-e v4 provision the M1 Extreme was also likely weak in the area PCI-e provisioning for workstation class jobs.



.
 
Apple will fool everyone and it will have the M3 3nm chips:p
Or they could make a totally different chip with a different naming convention like X-series or something for a chip line that may not be updated year-over-year like the M and A-series.
 
  • Like
Reactions: JosephAW
I’m personally waiting for a Mini with either an M2 or M2 Pro. Also hopefully HDMI 2.1.
I have an M1 mini and will wait until the M3 which will be faster and run cooler than M2 … Pro … or … Extreme Pro Max Ultra. Actually i might wait a little longer until the M3 Max arrives so i can have an M3 Max mini
 
It depends on what your use case is. It doesn’t matter if the newest iMac has a processor that is ten times faster than in an old Mac Pro, if it doesn’t support the additional hardware requirements you need from the Mac Pro it doesn’t do you any good.
What old hardware are you dragging along 3, 4, or 5+ years down the line to meet your requirements? Sure you can put in a new GPU, or some niche PCI cards, but by that point your storage, memory and bus speeds are all generations behind and are going to diminish your expected performance gains by a noticeable margin. If you are beyond 18 months out from purchase and chasing performance upgrades, you are basically just tinkering and sinking money in a project computer for personal enjoyment. This doesn't account for the oddball use case where a workstation used for a specific task that doesn't change for years may just need more memory, more storage, or a newer video card to maintain compatibility with newer software. But the vast majority of consumers that personally buy workstations for long-term investments are wasting their money. Smart money is to lease them and swap them out for newer models after a few years.
 
I maintain that if Apple is going to make a tower, the scenarios they are going to go after are:

* Cinematic - Hollywood render farm capable graphics processing.
* Processing extremely large big-data sets for deep science applications.
* Advanced ML and AI processing.

Doubling the Ultra probably isn't enough for this. They'd need to develop a new backplane that could 10x or 20x and give insane memory bandwidth. I believe the next Pro will completely move the goalposts on what "workstation" means. Starting price, $10,999.

Fingers crossed.
Something like this is basically what Nvidia is doing now, so I could see Apple get into a similar market.
 
I think what I'm reading is that they had the new case and internals ready to go for the new Mac Pro months ago, but were still working on the Extreme chip that it was designed for. They merely used the M1 for testing, but it was never going to be released without the Extreme chip.

That's a little dubious of the M1 can't drive the same set of internals as the upcoming extreme chip, how would it e a complete test mule for the internal chassis , or the chassis overall if not really matching the operating thermals.


I'm really looking forward to it. Not because I have any need for, or intention of ever owning one, but just for the excitement of seeing what it can do.

I'm particularly curious what they are going to do about the RAM in order to at least match the 1.5TB that the Intel Mac Pro can max out to.

The M1 Mini max RAM capacity is so limited that Apple is still ( 2 years after announcing the transition) still selling the Intel model that has better ability. Not matching the Max capacity ram, but didn't stop them for that desktop.

The M1 iMac 24" max RAM capacity is half of the iMac 21.5 ( 16GB vs 32GB ). Apple backslid for this desktop class for the transition.

Similarly, the iMac Pro and iMac 27" are retired . The iMac pro maxed out at 256GB. The last gen iMac 27" did 128GB. The Max powered Studio can't touch either one of those. The Ultra manages to eek out a tie with the late regular iMac but mostly a backslide across more of the SKU. More backsliding.

So trying to match the 1.5TB capacity probably isn't a huge goal for Apple. Apple was quite happy to pile on top of the already high Intel "> 1TB " tax that was slapped on the W 32xxM processors. There was a thousands cheaper option in Intel's line up that Apple could have picked ; and did not. That "extra" money is likely a major contributing driving force. At the Ultra announce dog and show show Apple made a comment that the Studio's 128GB was a large amount. Without ECC, going deeper into the triple digits skates out onto thinner and thinner ice.

The Max and Ultra SoCs are largely GPUs with some CPU and other bits wrapped around them. The characteristics of the RAM are highly likely going to be skewed to keeping the GPUs cores feed with data. ( i.e., how the CPU core complexes throughout the M1 generation are capped on max bandwidth. At Max/Ultra still at same levels as the Pro. )


My guess is that they will have both SoC RAM and plug-in, user-upgradeable, off-SoC RAM just like they do in the Intel.

The Intel memory controllers that Apple uses don't work that way for the Intel Mac Pros. The "plug in " memory iGPUs are no where near as performant as the Apple iGPUs are. High end GPUs don't have "plug in" RAM for several good reasons.


Xeon SP has some features where can plug in heterogenous RAM ( Optane RAM DIMMs and 'regular' DRAM DIMMS). Intel spent years on modifying Linux kernel updates to support that. it is not trivial. There is very little, to no, sign that Apple has been doing that kind of work. Unified Memory where the memory is not highly uniform causes issues. Issues a pretty good chance Apple isn't going to engage for a niche of niche of niche computer.

Plug in memory for a deadicated task like a RAM SSD would be much easier to tack on as the applications and most of the kernel largely would just see it as a device.

Decent chance Apple would take a 256 or 512GB max capacity as a 'win' in a half sized Mac pro and just move on. If that covers the 80+ percentile of the current Mac Pro popution that would still get them a viable set of user base.


l'm also guessing it will max out at 6TB RAM, partly because it will be running on LPDDR5 instead of 4, and partly because Apple will want the Mac Pro Extreme to blow the Intel Mac Pro out of the water.

LPDDRx shrinks capacity maximums versus DDRx ; not makes the higher. The M1 Max/Pro/Ultra are already on LPDDR5. There is not "new" 4th generation here with M2 .

At the lower half of the current Mac Pro performance coverage , probably yes (whip a 12-16 core with a single W6800 MPX module ) . At the top end, probably not. 16-28 Intel cores and two 6800Duos on embarrassingly parallel code.... the Intel Mac Pro will likely win. The 'problem' there is Apple charges around $27-32K for that. Likely Apple will point to a better Performance/$ ratio. (still as expensive but not quite as high completely maxed out. ). would not be surprising to see Apple bench against Pro Vega II Duos so can post a win.

If Apple+AMD do some driver product updates for W7000 class MPX modules even less likely will dominate if throw tons of money at it. Apple has a GPU driver problem for macOS on M-series. it doesn't scale-out to multiple instances. Two years in Apple has presented zero work that they are working on the issue. None. The biggers low level driver announcement at WWDC 2022 was that the new modern drivers work on iPad Pro also. Thats where Apple's focus has been.
 
Last edited:
View attachment 2034821

Just got an exclusive image from my contact in Apple of the upcoming Pro
Fake news! We all know it will be round.

Round Grater.jpg
 
  • Like
Reactions: xnu and visualseed
Was it really a replacement for the iMac Pro? Not really. the iMac Pro 2017 was mainly a replacement for the folks who thought the Mac Pro 2013 was "OK". About the same power supply limit. Swaps a better GPU and screen for a second GPU (which many didn't really get much utility out of. Some folks but not most).). There were a number of folks left "uncovered" by the iMac Pro ( hard core modular monitor folks), but mainly the iMac Pro was following the general trend of folks shifting over to upper end MBP and iMacs.

MP 2013 and iMac Pro 2017 left loads of folks still clinging to Mac Pro 2009-2012 systems still circling the airport.


The iMac Pro could be "faster" or " less expensive" than the Mac Pro 2019, but not both. A 16 Core MP with a Duo Vega II would be vastly more expensive than a iMac Pro , but would also 'smoke' it on anything that required a GPGPU workload that could scale to multiple GPUs. Anything that needed an over 512GB RAM footprint... again would smoke any iMac Pro that was 'swapping to disk' to keep up. Would 1TB RAM be cheaper than a iMac Pro? No. But it would have the performance mark.

At the top 25 percentile of the iMac Pro pricing scale with top end BTO options the Mac Pro was a better deal. Even more so if swap in 3rd party RAM and GPUs. It is no contest.



The entry-midrange iMac Pro versus the entry 8CPU/580X MP 2019 had problem with CPU/GPU bendmarks but in terms of doing doing high level aggregate I/O over bulky data, not as much if could put the data inside the box. A/V io inside the box has a value price point also that isn't capture by tech porn benchmarks. That latter is exactly the point that would move the folks squatting on Mac Pro 2009-2012 stuff.









That would be useful if Apple doesn't intend to cover the 8 internal slot territory anymore. Folks squatting from late 2012 to 2018 already was indicate that a decent chunk of folks have highly elongated update cycles. Some folks who bought in 2019 had waited 6 years or so to upgrade. 2019 --> 2025 who knows what Apple will have around at that point.

The primary customers for a 2022-23 Mac Pro are folks who bought the Mac Pro 2013 went it was stale (2016-2018 era ) or some used Mac Pro 2009-2012 in that timeframe. And also a small subset of the folks still holding out on pre 2013 models ( if Apple could slide back to $4999 then might get some of them. entry Ultra studio price + $1,000 gap ( Mac Pro with an "ultra" class SoC and $1K of modularity stuff the Studio doesn't have. ) .

Decent chance Apple sells both. Just chop off the 8-12 CPU core options for the Intel ( probably the 580/5700 options on the MPX modules. So 6000's series. ). That opens up the "bottom" range of current price range for another "Mac Pro" and the super duper , hardcore modularity folks can still buy a Mac where can swap out the CPU. Has even more legs of Apple releases a AMD 7000 series support later to limp the 2019 model into 23-24 timeframe.

There are indications that Apple is done with 3rd party GPU cards with macOS on M-series. If so then the new Mac Pro is going to leave some folks unhappy even if it has 1-3 PCI-e slots that are useful for non-firebreathing GPU cards ( e.g., 75W-max cards that are useful to many others.)

For the folks who needed 2-3 double wide cards that used lots of AUX power the MP 2019 is still going to have lots of utility even after the "half sized" Mac Pro comes out.
Yes, I think people will be unhappy with the next Mac Pro if they are expecting the last generation modular type of thinking...

I think apple might take a risk again like they did with Mac Pro 2013...but this time, due to the over whelming power of Apple Silicon, buyers might buy in. It depends on the numbers of sales of Mac Pro 2019. If they saw a trend in buying and potential of modular again, they might work on again some modular hybrid thing...but they might switch to an all "Apple Modular" accessories instead due to the complexities of the Apple Silicone architecture. But inclusive Apple modular would be too much work for Apple to keep up as Mac Pro is now a niche, so probably not.

If I bought Mac Pro 2013 and then Mac Pro 2019 I would be totally jaded by now. Apple gave the impression during the Mac Pro 2019 launch that they would be continuing developing modulars or add-ins for Mac Pro 2019 so you could continue to keep it relevant as the years go bye (like 2009-2012 era). Just like when Apple said that Mac Pro 2013 would be the next form for 10 years..(it seem it was close), they also gave that impression with Mac Pro 2019....

Then came Apple silicon...changed the game field. All we can do is wait and see. Predictions are futile and click-bate.
 
Apple hasn't been able to reproduce the expandability and overall capabilities of the current Mac Pro with an ARM based option.

The least they could do is Intel and AMD with Zen 4 EPYC processors on-board as an option while waiting for their own systems to evolve.

Showing comparisons to Mac Pro 2019 that used a several years old Xeon family is weak sauce.

Love to see a 64 core Zen 4 each with up to 2 TB of LPDDR5 options on PCIe 5.0 this fall. That would hold me over for five plus years in audio production. As it is I'll invest in more studio production hardware from Rupert Neve Designs and rely less on any sort of plugins for production work, while using a Mac Mini w/ 64GB. It's amazing how unnecessary the DAW is for performance when the studio is going through analog/solid state hardware doing 99% of the heavy lifting.
 
  • Like
Reactions: Kenjutsuz
I maintain that if Apple is going to make a tower, the scenarios they are going to go after are:

* Cinematic - Hollywood render farm capable graphics processing.
* Processing extremely large big-data sets for deep science applications.
* Advanced ML and AI processing.


that is largely not what they went after for the Mac Pro 2013 and MP 2019. So why would they go there now?
They have a feud with Nvidia so the last one is at a huge deficit. Also AMD and Intel are far bigger players there than Apple.

Data set visualization on a Mac? Yes. Hosting huge data set on a mac how? Most Macs have a one and only one storage device mindset. Even the current Mac Pro ships with zero spinning HDD drives. While you can add HDDs as an afterthought, that isn't what the chassis design revolves around. APFS is relatively dubious for HDDs and extremely dubious for large data sets where trying to accurately keep track of data over time (no data integrity protection there) . Apple is spending very little to no effort there. Apple Silicon version of a Mac pro isn't going to change that.


Apple has not enabled 3 party or dGPUs in two years since the announce of the transition. WWDC 2020 , 2021 , and 2022 have come and gone and no dGPU support. None. So now Apple is going to now turn around and do something that focuses almost solely on scale-out GPU support? Probably not.


Here is where Apple's ground work has been laid.


Cinematic -- ProResRAW Hollywood edting. reasonable motion graphics, and focused video effects. The visualization node on the front of the render farm. Something that involves a GUI interface for a single users. (dozens of headless and non inidivdual users. no. )

Large (not hyper large) data set -- enabling more than just one internal drive. That's the hurdle ( multiple internal SSD drives. M.2 . Maybe more direct U.2 support in a Mac Pro. )

Ai/ML -- Inference ? Yes. ( 64 Neural cores and AI/ML support in the GPU. probalby going to be respectable if leverage Unified memory well. ). New 'King Kong' of training? No.


Apple likely would do a rack version of their half sized tower that is appromatively just as 5U non-dense as the current rack mount is. (e.g., take the tower board and just turn it horizontal). Mini and Studios will likely be the dense rack solution. (as the Mini is now. and Mac Pro 2013 had/has some luck in also. )


Doubling the Ultra probably isn't enough for this. They'd need to develop a new backplane that could 10x or 20x and give insane memory bandwidth.

UltraFusion is already faster than most folks hyper short distance interconnects. Bandwidth wise , they aren't behind the curve there. It is so fast it has to be shorter distance to hit Apple's power consumptions constraints.

Lots of folks seem to think Apple is going to throw Perf/Watt completely out the windows for the Mac Pro. That is probably delusional. They probably are not. Decent chance they will relax it enough to provision a couple of x16 PCIe-v4 bundles out of the SoC package. However, throw the constraint into the trash can and go for data center and biggest workstations "King Kong" crowns. Probably not.


I believe the next Pro will completely move the goalposts on what "workstation" means. Starting price, $10,999.

Likely starts with something "Ultra" class and closer to current $5,999 or perhaps $4,999 if stripped down a bit ( e.g., half sized so smaller power supply , less slots, PLEX switch complexity , slot complexity (MPX connectors and AUX power ) , etc. etc. )

The $5,999 starting price already doubled the entry price for Mac Pros and left a decent sized number of classic Mac Pro users out in the code. Doing another double is just a death spiral pricing. Fewer users so rasie prices , gets fewer users , rinse and repeat. If going to try to kill it, just kill it. if anything Apple needs to look for a way to pull back from the doubling they did in 2019 ; not double down on that.


Fingers crossed.

If Nvidia , Intel , and/or AMD perhaps. Doubling the price is good way for Apple to remove themselves from the table. Over at Ampere and Amazon they would be clapping ecstatically.
 
* Cinematic - Hollywood render farm capable graphics processing.
* Processing extremely large big-data sets for deep science applications.
* Advanced ML and AI processing.
Something like this is basically what Nvidia is doing now, so I could see Apple get into a similar market.

Except Apple doesn't have a software foundation to support that. macOS tops out at 64 threads. Linux doesn't. Nor does it limit the discrete GPU options. 100+ Gb Ethernet on macOS? Nope. Nvidia is shooting for the data center, super computers , and all the lengthily checklist and standards compliance places that Apple doesn't like to go. Apple dumped XServe years ago. They have shown zero intention in going back.

Apple is doing some stuff with XCode Cloud. Amazon, Azure, MacStadium are doing Mac CI/CB cloud services, but that is mainly a "make do" with Apple systems that are sold individually in the mainstream. macOS is mainly licensed as a single user (or at best single organization ) system.
 
I think what I'm reading is that they had the new case and internals ready to go for the new Mac Pro months ago, but were still working on the Extreme chip that it was designed for. They merely used the M1 for testing, but it was never going to be released without the Extreme chip.

I'm really looking forward to it. Not because I have any need for, or intention of ever owning one, but just for the excitement of seeing what it can do.

I'm particularly curious what they are going to do about the RAM in order to at least match the 1.5TB that the Intel Mac Pro can max out to.

My guess is that they will have both SoC RAM and plug-in, user-upgradeable, off-SoC RAM just like they do in the Intel.

l'm also guessing it will max out at 6TB RAM, partly because it will be running on LPDDR5 instead of 4, and partly because Apple will want the Mac Pro Extreme to blow the Intel Mac Pro out of the water.
The problem with mixing current technology with the new Apple Silicon is that the speed they are getting is the combination of onboard RAM sync'd with the cpu & storage. It is almost one and the same. If you separate it, speed will probably drop dramatically.

The example is with the current Silicon combination SSD storage. Since the RAM, CPU and Storage (SSD) is all inclusive and basically on one architecture or really one chip (so to speak) concept, you get the insane speeds, but if you bring in current modular RAM or SSDs as an add-on (like the top of the market Samsung blades), speeds and power will still drop.

I notice a major difference when using external thunderbolt SSD to house and directly work from the file with my video projects due to storage internal limitations (cannot afford 4-8TB SSD internal Apple pricing). Though the external SSD is fast (getting 28000+ MB/s on bench mark scores) it bottle-necks in contrasts with the speed of "all-in-one" Apple Silicon architecture (getting 54000+ MB/s on bench mark scores). Sometimes my M1 Max works so fast compared to the external RAID SSD (or single external unit) that it glitches or slows down trying to keep up. So mixing current PCie or SSD RAM NVMe add-on blades etc. along with Silicon UNLESS Apple comes up with another miracle technology to integrate, I highly doubt we will see modular or the ability to add-on and get the same results.

Again... we can only see what happens.
 
i don't care about Mac Pro and i don't care about a Mac Mini redesign. just refresh the current Mac Mini with M2 and 24-32GB RAM and you can have my money Apple.

edit: unless, of course, a M2 Mac Mini runs so hot it thermal throttles even with the current cooling solution. 😁

That's the problem, Apple: We have Studio for ya (smiley face).
 
If I bought Mac Pro 2013 and then Mac Pro 2019 I would be totally jaded by now. Apple gave the impression during the Mac Pro 2019 launch that they would be continuing developing modulars or add-ins for Mac Pro 2019 so you could continue to keep it relevant as the years go bye (like 2009-2012 era). Just like when Apple said that Mac Pro 2013 would be the next form for 10 years..(it seem it was close), they also gave that impression with Mac Pro 2019....

Then came Apple silicon...changed the game field. All we can do is wait and see. Predictions are futile and click-bate.

Apple Silicon triggered a set of changes but it doesn't eliminate modularity from a future Mac Pro across the board. 32-bit apps were axed before Apple Silicon arrive but likely were coupled to that move ( to make transition cleaner and overall support matrix complexity cleaner. ). Likely the move to the new driverKit ( and deprecateing kernel extensions). those came to the Intel side also though.

New cards for macOS on M-series have come during the transition. For example, Sonnet Technology introduced this card over a month ago.

https://www.sonnettech.com/product/mcfiver-pcie-card/overview.html

It works in a Thunderbolt external PCI-e expansion box, so it is extremely likely it would work in a PCI-e slot in a new M2 <insert adjective here> powered Mac Pro that had a decent x8 (or better) PCI-e v4 slot. The drivers are already there. (just need a plug-in-play driver extension to basic PCI-e driver for the Thunderbolt aspect. So basic driver is likely there. )

If Apple had done a Mac Pro M2 generation Mac Pro at WWDC 2022 then that card would been a perfect merge to put on stage that .... yes we are still doing a sizable set of PCI-e cards support. If go back to 2019 Apple was *NOT* doing dog and pony demos for off-the-shelf from Fry's/Microcenter/Newegg demos for "Windows" GPU cards. It was stuff like Sonnet Tech storage/network cards, AVID HDX , Aja, Afterburner , etc. here is a picture of what Apple flashed up on the screen WWDC 2019




20190604125238_AppleWWDC2019-MacProPresentationOCIeCardsWeb.png







Pretty sure all of that stuff works on macOS on M-series now. So would work in a M2 generation Mac Pro. What dramatically changed there? Relatively little.


Now the AMD vs Nvidia GPU fan clubs would say the sky has fallen but there dozens of PCI-e cards that do work with macOS on M-series. And the number has gotten steadly bigger over the last two years. Those cards have worked on the MP 2019. The MP 2019 can cross boot into Windows so can even kludge the Nvidia GPUs if temporary swap out of macOS.

The whole "bare iron" boot Windows is likely gone for the next Mac Pro but the 2019 still has it. The Mac Pro came with a T2 chip so the days of "hack the system firmware" were over at that point. Apple Silicon walking away from UEFI should have been a surprise. ( Apple grumbled for years before about UEFI security issues. the iPHone/iPads didn't use it. ).

Apple has not been working toward zero PCI-e card support. Myopically viewed through discrete GPUs may give that view, but that is by no means the whole PCI-e card ecosystem.


For non critical boot environment (e.g., not the primary display driver) they groundwork is already there in the future path of Apple Silicon systems.

Apple committed in that April 2017 meeting to high bandwidth at least as much as to modularity. (modularity mentioned there was about the screen). Going to SoC mounted RAM is getting them high bandwidth. ( effectively a "poor man's " HBM implementation). The presumption has been that they would sacrifice bandwidth for modulatory. That is probably flawed. Even more so given Apple has droned on repetitively since 2020 about Pref/Watt priorities.

I think some folks presumed that hyper expensive MPX modules would be futured proofed into future machines. That is probably not true. It probably would have helped if Intel had previewed their Gen 11 (TigerLake) SoCs with integrated TB controllers earlier, but Apple simplifying the Thunderbolt solution in the Mac Pro shouldn't have been hard to see coming. (well perhaps not for Mac Pro folks as 100's of posts from 2013 to 2019 about how Apple should dump TB from future Mac Pro... didn't happen. Those folks have missed the boat for a decade now... probably won't change. ) . The MP has seen GPU MPX modules from 580X/Polaris , 5000 (RDNA ) , 6000 (RDNA2) series. Three generations (yeah Apple started off with something 'old' at the low end). So there have been updates over time.


The other parts from the April 2017 meeting on state of future Mac Pro was that:

1. Many users wanted one big GPU rather than multiple GPUs. Well, careful for what you wish for. A four die, mega package GPU is going to be 'big'. They said some users wanted dual ( or more) GPUs. That is somewhat of a break. Although I suspect they will counter that this is really big single.

[ Apple probably does need a multiple GPGPU ability. Multiple GPUs for driving display probably not. ]


2. "leaned to much on Thunderbolt" for storage. Got slots for PCI-e SSDs and some SATA sockets and 3rd party brackets for HDDs. I would not bet against those for an Apple Silicon Mac Pro. A x16 PCI-e v4 slot for a SSD carrier card needs work (not supported in current M1 generation), but a x4 SSD could be done with a M1 Ultra now. There is just no room in a Studio for a M.2 slot.

The internal USB socket for a iLok dongle has a more than good chance too.

[ e.g. If Apple chopped 4-6 slots from current MP 2019 there still would be room for a PCI-e M.2 carrier car and/or the same bracket ( if the SoC socket and heat sink to creep into the HDD bracket zone. ). If just half the volume of current Mac Pro there is still lots of internal space left relative to the rest of the Mac line up. Lower chance, Apple could see Apple pushing folks to a U.2 drives is a push do dump SATA. Or just provisiong 2.5" SSD as cheap storage if the socket/cooler get bigger. ]


3. Modular screens. Mini , Studio already do this . New Mac Pro wouldn't be a problem.

4. "... we’re committed to making it our highest-end, high throughput desktop system, designed for our demanding pro customers. ..."

That isn't promising modular everything. Throughput is not modularity. Happens to coincide in the past , but really the same thing.

" ... As part of doing a new Mac Pro — it is, by definition, a modular system — we will be doing a pro display as well. Now you won’t see any of those products this year; we’re in the process of that. We think it’s really important to create something great for our pro customers who want a Mac Pro modular system ..."

The explicit modularity was display (also least revealing since already there on previous Mac Pros, but in contrast to heavily hinted future iMac Pro in same conversation. ). Got more with the Mac Pro 2019. but largely that was driven from constraints that Intel did; not Apple. Xeon W3200 series has no GPU , thunderbolt ( hence there complexity of routing video to various different places on the logic board. ). Intel picked the memory format. That whole package is a "repurpose" of a die primarily constructed for data center servers. That is highly unlikely Apple basic guidelines when they start their design. Apple doesn't operate deeply in the data center server market. Apple is going to start off with the intent of making a single person system as the primary end goal. That's what they sell.



For another example, the Afterburner card. The Ultra (and an Extreme) basically 'smoke' that card. Apple should have thrown that out of SoC because it didn't meet some modular design constraint. That is kind of goofy. A modern , bigger FPGA won't beat the Ultra/Extreme either because there is bandwidth bottleneck with a external card.


The disconnect comes from folks who read "high throughput desktop system" as being a euphemism for top end Nvidia GPU card support.


P.S. Grrr. audio site image presents in editor but not inline. Similar picture different angle.


Screen-Shot-2019-06-03-at-14.23.16-PM.png


 
Last edited:
  • Love
Reactions: loby
Apple Silicon triggered a set of changes but it doesn't eliminate modularity from a future Mac Pro across the board. 32-bit apps were axed before Apple Silicon arrive but likely were coupled to that move ( to make transition cleaner and overall support matrix complexity cleaner. ). Likely the move to the new driverKit ( and deprecateing kernel extensions). those came to the Intel side also though.

New cards for macOS on M-series have come during the transition. For example, Sonnet Technology introduced this card over a month ago.

https://www.sonnettech.com/product/mcfiver-pcie-card/overview.html

It works in a Thunderbolt external PCI-e expansion box, so it is extremely likely it would work in a PCI-e slot in a new M2 <insert adjective here> powered Mac Pro that had a decent x8 (or better) PCI-e v4 slot. The drivers are already there. (just need a plug-in-play driver extension to basic PCI-e driver for the Thunderbolt aspect. So basic driver is likely there. )

If Apple had done a Mac Pro M2 generation Mac Pro at WWDC 2022 then that card would been a perfect merge to put on stage that .... yes we are still doing a sizable set of PCI-e cards support. If go back to 2019 Apple was *NOT* doing dog and pony demos for off-the-shelf from Fry's/Microcenter/Newegg demos for "Windows" GPU cards. It was stuff like Sonnet Tech storage/network cards, AVID HDX , Aja, Afterburner , etc. here is a picture of what Apple flashed up on the screen WWDC 2019




20190604125238_AppleWWDC2019-MacProPresentationOCIeCardsWeb.png







Pretty sure all of that stuff works on macOS on M-series now. So would work in a M2 generation Mac Pro. What dramatically changed there? Relatively little.


Now the AMD vs Nvidia GPU fan clubs would say the sky has fallen but there dozens of PCI-e cards that do work with macOS on M-series. And the number has gotten steadly bigger over the last two years. Those cards have worked on the MP 2019. The MP 2019 can cross boot into Windows so can even kludge the Nvidia GPUs if temporary swap out of macOS.

The whole "bare iron" boot Windows is likely gone for the next Mac Pro but the 2019 still has it. The Mac Pro came with a T2 chip so the days of "hack the system firmware" were over at that point. Apple Silicon walking away from UEFI should have been a surprise. ( Apple grumbled for years before about UEFI security issues. the iPHone/iPads didn't use it. ).

Apple has not been working toward zero PCI-e card support. Myopically viewed through discrete GPUs may give that view, but that is by no means the whole PCI-e card ecosystem.


For non critical boot environment (e.g., not the primary display driver) they groundwork is already there in the future path of Apple Silicon systems.

Apple committed in that April 2017 meeting to high bandwidth at least as much as to modularity. (modularity mentioned there was about the screen). Going to SoC mounted RAM is getting them high bandwidth. ( effectively a "poor man's " HBM implementation). The presumption has been that they would sacrifice bandwidth for modulatory. That is probably flawed. Even more so given Apple has droned on repetitively since 2020 about Pref/Watt priorities.

I think some folks presumed that hyper expensive MPX modules would be futured proofed into future machines. That is probably not true. It probably would have helped if Intel had previewed their Gen 11 (TigerLake) SoCs with integrated TB controllers earlier, but Apple simplifying the Thunderbolt solution in the Mac Pro shouldn't have been hard to see coming. (well perhaps not for Mac Pro folks as 100's of posts from 2013 to 2019 about how Apple should dump TB from future Mac Pro... didn't happen. Those folks have missed the boat for a decade now... probably won't change. ) . The MP has seen GPU MPX modules from 580X/Polaris , 5000 (RDNA ) , 6000 (RDNA2) series. Three generations (yeah Apple started off with something 'old' at the low end). So there have been updates over time.


The other parts from the April 2017 meeting on state of future Mac Pro was that:

1. Many users wanted one big GPU rather than multiple GPUs. Well, careful for what you wish for. A four die, mega package GPU is going to be 'big'. They said some users wanted dual ( or more) GPUs. That is somewhat of a break. Although I suspect they will counter that this is really big single.

[ Apple probably does need a multiple GPGPU ability. Multiple GPUs for driving display probably not. ]


2. "leaned to much on Thunderbolt" for storage. Got slots for PCI-e SSDs and some SATA sockets and 3rd party brackets for HDDs. I would not bet against those for an Apple Silicon Mac Pro. A x16 PCI-e v4 slot for a SSD carrier card needs work (not supported in current M1 generation), but a x4 SSD could be done with a M1 Ultra now. There is just no room in a Studio for a M.2 slot.

The internal USB socket for a iLok dongle has a more than good chance too.

[ e.g. If Apple chopped 4-6 slots from current MP 2019 there still would be room for a PCI-e M.2 carrier car and/or the same bracket ( if the SoC socket and heat sink to creep into the HDD bracket zone. ). If just half the volume of current Mac Pro there is still lots of internal space left relative to the rest of the Mac line up. Lower chance, Apple could see Apple pushing folks to a U.2 drives is a push do dump SATA. Or just provisiong 2.5" SSD as cheap storage if the socket/cooler get bigger. ]


3. Modular screens. Mini , Studio already do this . New Mac Pro wouldn't be a problem.

4. "... we’re committed to making it our highest-end, high throughput desktop system, designed for our demanding pro customers. ..."

That isn't promising modular everything. Throughput is not modularity. Happens to coincide in the past , but really the same thing.

" ... As part of doing a new Mac Pro — it is, by definition, a modular system — we will be doing a pro display as well. Now you won’t see any of those products this year; we’re in the process of that. We think it’s really important to create something great for our pro customers who want a Mac Pro modular system ..."

The explicit modularity was display (also least revealing since already there on previous Mac Pros, but in contrast to heavily hinted future iMac Pro in same conversation. ). Got more with the Mac Pro 2019. but largely that was driven from constraints that Intel did; not Apple. Xeon W3200 series has no GPU , thunderbolt ( hence there complexity of routing video to various different places on the logic board. ). Intel picked the memory format. That whole package is a "repurpose" of a die primarily constructed for data center servers. That is highly unlikely Apple basic guidelines when they start their design. Apple doesn't operate deeply in the data center server market. Apple is going to start off with the intent of making a single person system as the primary end goal. That's what they sell.



For another example, the Afterburner card. The Ultra (and an Extreme) basically 'smoke' that card. Apple should have thrown that out of SoC because it didn't meet some modular design constraint. That is kind of goofy. A modern , bigger FPGA won't beat the Ultra/Extreme either because there is bandwidth bottleneck with a external card.


The disconnect comes from folks who read "high throughput desktop system" as being a euphemism for top end Nvidia GPU card support.


P.S. Grrr. audio site image presents in editor but not inline. Similar picture different angle.


Screen-Shot-2019-06-03-at-14.23.16-PM.png


Thanks for the in-depth explanation. Understandable.

Not owning the current Mac Pro 2019 I am not aware of third-party add-ons or if third party bought into creating accessories for it with expansion cards etc. I own Mac Pro 2013 and it had limited support with third party accessories over the years unfortunately (i.e. thunderbolt 2 short lived issues). Even with the Thunderbolt 3 to 2 adapter from Apple (much appreciated from Apple), the market did not really have too much expansion to offer, especially the GPU issues/options.

Maybe I am wrong, but my impression from Apple was that they were going to make regular expansion modulars and other options for Mac Pro 2019, but we did not see much after.

If Apple takes the stream of thought you presented (even quoting Apple's own words - which I do not always trust and take it with a grain of salt - snake oil "marketing" company, but like their products)...then if they create Mac Pro 2022-23 like you said...it will be an incredible Mac Pro! Let's hope that happens.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.