MP All Models Introducing... The Mac?

[deconstruct60]

Hopefully it is the mythical xMac alongside a continuing full-size Mac Pro, possibly called Mac.

Doubtful this will be an xMac in terms of pricing. the xMac crow primarily wanted something in full fratricide zone with the iMac. That probably won't happen. The first Mac (with no suffixes or adjectives) was an all in one. If anything is getting a revival of just plain 'Mac' (or 'Macintosh' ) it will be an all-in-one model. [ there are folks who jumped into the Mac market in the 90's when Apple had a proliferation of "boxes with slots". Apple Silicon (ASi) is primarily so Apple does not have to build a box with slot. Laptops , iMac , and Mini are going to be sweet spot for ASi. The ASi SoC's will allow Apple over time to push further in the enclosure directions they have wanted to take the Mac but were limited to. Thinner laptops, Smaller volume iMacs, Minis with performance but the same size. etc ]

There is higher tension if they do a half-height, limited slot model Mac Pro with the space the iMac Pro is sitting in now. Apple just move the performance down to the "plain" iMac . This system with no-so-inexpensive MPX GPU models could/would sit above that iMac price zone , but below the 'full' Mac Pro zone. [ the iMac Pro may turn into effectively being the "large screen" iMac with the "extra deluxe" ( XDR-ish ) screen attached. More so gapped on graphics and screen than on application core counts.

Pretty good chance this first "half size" Mac is driven more so by the limited I/O capacity coming out of this first iteration ASi SoC ( probably aim more at the iMac range ) than with any deliberate objective based on lust to deliver an $1-2K priced xMac . If anything something back in the MP 2013 price range of $2,999 - $5,999 zone.
It will help Apple "spike the ball and do their early transition touchdown dance". See transitioned all of the Macs products lines (just with the 'full' Mac Pro still coasting before an even later SoC transition. )

I would love a Mac with more flexibility, but cannot justify the expense of a Mac Pro as it is beyond my performance needs (I could use the power, but it would make little difference apart from app builds taking a little less time), so a smaller and cheaper model would be excellent.

I think Apple knows they left as decent chunk of Mac Pro buyers behind when they increased the entry price of the system by pragmatically 100%. If they do a "half tall" system it would probably only to get those "Mac Pro" users that go left behind. It really won't be primarily targeted at the xMac folks the existed all along when there were Mac Pro 2009-2013 models.

Once has Apple full Mac Pro ASi SoC then these two will share SSD, RAM , MPX module all of which would probably set a higher baseline price. There would be some pricing shaving on the "half' tall model with lower power supply and less logic board signal routing and complexity. And "greener" to ship and stock.

 

[deconstruct60]

Wasn't that literally the point of having a big FPGA card - that it was supposed to be reconfigurable with software to suit the codec being used? I seem to recall apple execs talking about that specific aspect of it after one wwdc.

No, that really wasn't Apple's primary point. Nor was it RED's with their RED Rocket card. When Apple talked of codec flexibility it was more couched in what future ProRes variations might appear. Like ProRes RAW coming into higher adoption when Afterburner came out.

It was far more non-Apple folks hyping the "going to do every codec under the Sun". Apple did muddle/mangle/hype the message with the "end to proxies" mantra they used. But that was really the end of ProRes proxies if actually objectively looked at the examples they were talking about. Lots of folks 'ran' with 'end to proxies' to mean Apple was going to do everything native with the Afterburner. That is not how it is built or invoked.

Users solely get to Afterburner functionality by going through Apple system libraries. There is no SDK kit for this FGPA card at all. So it is only the codes that Apple is primary interested in. ( which exactly parallels what RED and lots of other FPGA buyers to also. They buy FPGA because it is cheaper than doing your own "raw down to the transistor" custom ASIC chip. Not because they want a card that flip-flops through 42 different modes. ) Apple was not going into the business of taking the 'burden' of doing a FPGA card off of RED (or any other specific camera vendor. )


Once Afterburner came out and the FAQ and technical info outlined that Afterburner only does ProRes variants decode ( not even encode) then bubble should have burst that it was a "do everything" card. As Apple gets more experience trying to cover more sensors with ProRes RAW there will likely be adjustments to the standard over time as the sensors either grow in resolution, size, and/or dynamic range. As a FPGA Afterburner can adapt to those incremental, evolutionary refinements. ( e.g. future 10K resolution with perhaps a step down in number of concurrent streams. )

Also in the follow up tech talk from Apple was that ProRes RAW was going to be 'native' to cameras. That has largely been a bit a hand waving as it is being done for many cameras being coupled to an external Atmos recorder. If it happens outside the camera it is huge stretch to label that native. It is "native" to what the primary recording (done outside the camera).

Apple's efforts to put ProRes RAW onto Windows 10 and get a broader set of cameras recording in ProRes variants are what Afterburner is about. Grow a large body of context in those formats and Afterburner can chew through that content faster than the non Mac systems. So it is an ecosystem growth enabler. Not a do everything to for everybody tool.

[ A side motivation for Afterburner may also have been the protracted legal dust up they were in with Red over compressed RAW codecs. If Apple had to detangle ProRes RAW from the RED patent they would have needed at new format. PreRes RAW-free or RAW2 or something. Afterburner would be a key tool to effectively taking the long way around those patents as the format substantively changed. ]


The current and future widespread standard codec H.264 , H.265 , AV1, etc. will probably be done inside the Apple Silicon (ASi) SoCs. Afterburner never did "have to" cover those since they were likely to get covered in the ARM transition anyway. ( and by the GPUs on the future MPX modules if Apple and the GPU vendor get the low level decode/encoder work done in future macOS iterations. Apple didn't completely avoid Intel's QuickSync. Going forward it probably won't be QuickSync as the major target. Intel would have to do some major work on add-in-cards to save that on macOS. Pretty unclear they would get that done. )

 

[deconstruct60]

MPX bays *only*, so 3 slots total, 1 with the cpu.

Given the CPU SoC probably has the securve enclave in it, then it is highly likely that it is soldered to the logicboard.
if the security keys for the system to work 'walk away" the system isn't going to be useful.

Weaving the T2 into the CPU is extremely likely going to bring T2 placement and mounting constraints to the main SoC also.

Probably would have to get the RAM , SSD , and SoC onto a single card in order to be even remotely viable as a solution. CPU and RAM have distance placement constraints so compled (e.g., the 2009-2012 "CPU" board. ). The T2 and NAND daughter cards are coupled cryptographically.

Also as the Apple Silicon (ASi) SoC have more I/O pins ( e.g., substantive double digit PCIe and USB lanes ) the package is going to be bigger. The more the package talks to the larger the package needs to get. [ that's why iPhone and iPad packages don't talk that much. One (maybe two) port wonder systems. GPU internal. One and only one system drive . etc. etc. There is lots being "throw overboard" there to keep the package smaller . That is going to be tough to do in a Mac Pro enclosure, even at 'half size' when the point of the enclosure is to talk to more stuff. ).

If Apple does a Mac Pro class SoC that can provision x64 PCI-e v4 lanes it isn't going to be "small". That is not because they need it to be bigger for enable a bigger heat sink. It is because pins can only get so small.

If Apple SoC eschews SATA and want to add that back into the new Mac Pro varaints... will need a discrete controller. Which again will consume logic board space ( where the current PCH is . It will be smaller. But has also thrown away 'half' of the logic board. )

 

[OkiRun]

Given the CPU SoC probably has the securve enclave in it, then it is highly likely that it is soldered to the logicboard.
if the security keys for the system to work 'walk away" the system isn't going to be useful.

Weaving the T2 into the CPU is extremely likely going to bring T2 placement and mounting constraints to the main SoC also.

Probably would have to get the RAM , SSD , and SoC onto a single card in order to be even remotely viable as a solution. CPU and RAM have distance placement constraints so compled (e.g., the 2009-2012 "CPU" board. ). The T2 and NAND daughter cards are coupled cryptographically.

Also as the Apple Silicon (ASi) SoC have more I/O pins ( e.g., substantive double digit PCIe and USB lanes ) the package is going to be bigger. The more the package talks to the larger the package needs to get. [ that's why iPhone and iPad packages don't talk that much. One (maybe two) port wonder systems. GPU internal. One and only one system drive . etc. etc. There is lots being "throw overboard" there to keep the package smaller . That is going to be tough to do in a Mac Pro enclosure, even at 'half size' when the point of the enclosure is to talk to more stuff. ).

If Apple does a Mac Pro class SoC that can provision x64 PCI-e v4 lanes it isn't going to be "small". That is not because they need it to be bigger for enable a bigger heat sink. It is because pins can only get so small.

If Apple SoC eschews SATA and want to add that back into the new Mac Pro varaints... will need a discrete controller. Which again will consume logic board space ( where the current PCH is . It will be smaller. But has also thrown away 'half' of the logic board. )

So in your mind ~ the first AS Mac Pro would have one MPX bay and one pci 16 slot; which could house an afterburner.
The cpu would be housed in a separate 'box' (top left) and 6 RAM slots next to it. If Apple has its own gpu - where would it be housed?
This looks like Mattspace design - but he has the Mac Station Covering over the CPU and RAM.
The Apple gpu would go in the same box as cpu?

 

[deconstruct60]

So in your mind ~ the first AS Mac Pro would have one MPX bay and one pci 16 slot; which could house an afterburner.
The cpu would be housed in a separate 'box' (top left) and 6 RAM slots next to it. If Apple has its own gpu - where would it be housed?
This looks like Mattspace design - but he has the Mac Station Covering over the CPU and RAM.
The Apple gpu would go in the same box as cpu?

View attachment 1569515

Apple likely isn't going to do a CPU on either this "half sized" Mac Pro or on a Full sized one. They are going to do a SoC ( system on a chip). The GPU is in that package. Same basic structure as it is now with the A-series. Just like is now with Intel's "mainstream" desktop & mobile series Core i3-i5 chips (and most of the i7 that are based on same microarchitecture and die combo variant.)

Apple Silicon is structure primarily aimed at laptops. I don't believe in the slightest that Apple is taking cues from the latest AMD Eypc or Intel Xeon SP or Ampere/Amazon/Marvell ARM server packages and using that as a baseline for Apple SoC at all. Nor are they likely to be chasing the Xeon W 3000 series or Threadripper. (or the desktop Ryzen )

Instead of going almost absolute maximum on application cores Apple is probably going to throw a hefty amount of transistor budget at ML/AI cores. At all the stuff the T2 covers ( video and audio codecs , SSD , security enclave ) ..... and a decent amount at GPU cores. It isn't going to be an "end all , be all" GPU. They'd only have to cover something a bit less than the W5500. Could be less if they want to control costs. Can still cover Audio workstation and no 3D GPU intensive markets with an iGPU. It doesn't have to be one large 5nm die, but if unified (shared / integrated ) memory it very probably is one large package.

5nm means their application (ARM) cores are smaller. So for about same size package can add more "other stuff". The whole point of a SoC is to add "other stuff".

If Apple wanted to do a discrete, non-uniform memory GPU then it could easily be on the MPX module just like an AMD GPU die would/could be. But that doesn't have to be the minimal default GPU. If there is a universal iGPU then pragmatically that basically frees up a slot.

That would make it hard to crank up to > 28 application (ARM) core counts, but Apple probably has very little interest in doing so. In Apple SoC so far the non ARM cores function units get the bulk of the transitor budgets. That is quite likely going to be true for the SoC that goes into future Mac Pro's also.

Apple is going to sell you a GPU that is soldered to the motherboard ( just like the ARM cores likely are). That will help put a floor on what the Mac Pro costs ( half or full sized) that is above the generic Mini ( or most of, if not all, of the iMac line up).


If the bottom side board heat from the SoC is low enough they could put the RAM slots on the bottom side ( if NAND blades back there too might as well ). That would allow space for the SATA cage if they haven't banished it.
But yes either SATA , cage , RAM , or NAND blades could be down stream of CPU package. .


If the MPX bay is provisioned with PCI-e v4 then could also flip those MPX and single width x16 placements. ( especially if the single with was stepped down to PCI-e v3 x16 ). Longer PCI-e v4 distances just 'buy' more complexity. The x16 PCI-e slot might go on this first "half sized" Mac Pro is the SoC isn't really Mac Pro class this first iteration.
[ I'm skeptical that in the first year SoC that Apple is going to do a high double digit PCI-e lane provisioning. They may have put the effort in but my guess is that they are primarily focused on the solution space that the MBP 16" presents in terms of external I/O. Maybe the current iMac 27' . That would put then in the 20-30 lane range. So this first 'half' Mac Pro would have quite 'thinned out' aggregate bandwidth. One of the motivators for chopping down the slots; no bandwidth to feed them. ]

 

[mattspace]

This looks like Mattspace design - but he has the Mac Station Covering over the CPU and RAM.

To clarify, my suspicion was that the CPU, RAM and IO would all be on a single PCB, and that making that a separate part from the backplane, like the CPU tray on the 4,1/5,1 could be done very elegantly if it was packaged as an MPX module - maybe that would be outside the capabilities of the MPX socket in terms of getting the CPU to communicate with the rest of the MPX slots and it might require a different connector, but i was thinking in terms of packaging utility - how to get the most reuse of existing manufacturing setups etc, and in terms of keeping the system modularly upgradable.

It's basically just the 5,1 CPU tray, with the built in IO on it, instead of the backplane.

MPXing the core functionality means you can put that on a cheaper, but faster upgrade track than the system as a whole, and then you can talk about the environmental benefits of shipping smaller boxes for more upgrade cycles.

Eventually, your MPX-core Mac Station becomes the subscription / rental product, on a similar lifecycle to the iPhone, and your MPX chassis, power supply and GPUs become like all your lightning accessories, that you keep through multiple generations of iOS device tradeins.

Apple likely isn't going to do a CPU on either this "half sized" Mac Pro or on a Full sized one. They are going to do a SoC ( system on a chip). The GPU is in that package.

No Apple Silicon GPU has been demonstrated as capable of driving more than a single display, which isn't even good enough for a laptop, let alone a desktop machine. "competitive with games consoles", which apart from anything is a bald-faced lie on Apple's part, isn't good enough for a desktop system's GPU. Has anyone seen a AS devkit driving multiple displays under macOS?

Apple tried "pro" computers with built-in GPUs. It was a failure. It is a failed paradigm, and changing from built-in AMD GPUs to built-in Apple GPUs isn't going to change that.

The trashcan wasn't a good idea, implemented badly. It was a bad idea, implemented badly. Any attempt to revisit any similarity to that system, will end the same way.

 

[deconstruct60]

No Apple Silicon GPU has been demonstrated as capable of driving more than a single display, which isn't even good enough for a laptop, let alone a desktop machine. "competitive with games consoles", which apart from anything is a bald-faced lie on Apple's part, isn't good enough for a desktop system's GPU. Has anyone seen a AS devkit driving multiple displays under macOS?

Given that Apple is about to at least "soft release" a Mac Laptop on Tuesday, that highly likely means they have solved the multiple display issue with whatever iGPU they have attacked to the A-14X (or better) class SoC they have assigned to the first Mac laptops. Given at this point Intel is on Xe graphics on their laptops and AMD is minmally on Ryzen 4000 series APU ( and posed to be on RNDA mobile APUs inside of six months).... Apple will need to roll out with something that is decent for general usage.

There is pretty good chance that Apple's Tuesday "dog and pony show" will show a Apple Silicon Mac laptop driving a XDR display.

The DTK has a GPU from (at this point) two generation back. Having an iGPU goes a long way to "solving' the Thunderbolt 4 requirement that all the TB ports be feed an DIsplayPort video stream. If someone yanks the Apple MPX GPU module out of the system then the TB ports are simply just feed the output from the iGPU. If MPX present then have an optional feed . ( basically the "alt DP feed' switch have on full size Mac Pro 2019 now only the feed from MPX Connector 2 gets feed from the SoC)

The iGPU also allows more seamless running of unmodifled iOS/iPadOS binaries. Better or worse that will be a feature checkbox for ASi Macs. Including the desktiops. Thinking that the Mac Pro isn't going to do that probably is in the same zone as the folks who thought the Mac Pro 7,1 wasn't going to have native support for Thunderbolt (something like the MPX connector). That unmodified binary expects to see a unified memory , Apple GPU . (sure Apple could stick an emulator on top, but why bother if a physical one is there. Just copy the finished frame buffers over. )

Apple tried "pro" computers with built-in GPUs. It was a failure. It is a failed paradigm, and changing from built-in AMD GPUs to built-in Apple GPUs isn't going to change that.

Failed for you perhaps. Failed over all ... not really. iMac sales are high. iMac Pro sold better than some iMac configurations. As much as folks want to hand wave at the Mac Pro 2013 as failed.... Apple's priority response was with the iMac Pro ( not the Mac Pro 2019). The several billion dollar a year Mac business overwhelming consists of embedded GPUs in the systems.

If the "half sized" Mac Pro has some PCI-e slots then it isn't the Mac Pro ... so it really isn't the same response. Will an iGPU piss off the total configuration control freaks. Probably. Does the T2 SSD already do that? For many yeah .... and yet the "Pro" labeled systems are all selling at profitable enough levels for Apple right now.

If Apple intended the iGPU to be the primary usage for all of the "ASi Mac Pro" users that would be an issue. But it probably isn't.

The trashcan wasn't a good idea, implemented badly. It was a bad idea, implemented badly. Any attempt to revisit any similarity to that system, will end the same way.

This isn't going back to the Mac Pro 2013 at all other than the really extremely broad category of having two GPUs.
The system can have just one GPU. If the user wants to primary use just one single "big, high powered" GPU then the two GPU (and the CPU) are not thermally coupled to each other. The CPU and iGPU are coupled but they are balanced by Apple is very same way it is balanced by them in the iPhone, iPad Pro , etc. Which if you want to label that a technical failure you are smoking something strong. It hasn't failed.

The everyone gets a "Dual GPU" supposedly failed in part because Apple said not enough folks had software that found a use for it. So the question in this new context is how many Mac Pro users do not have any iPhone/iPadOS software at all to run. That is extremely likely under the 50% range. It is pretty likely under the 25% range. It is actually the opposite probably they had before. In part, precisely because they are not identical twin GPUs. [ In the significan't "Pro" space of developers of Apple eco system software the number of folks not using an Apple GPU is going to be about zero percentage zone in 1-2 years. ] So again not even close to repeating the Mac Pro 2013 at all.

The Mac Pro 2013 CPU was socketed. The ASi Mac Pro CPU probably will not be socketed. The CPU on a card thing probably isn't going to happen. Apple has soldered down the Secure Enclave contained in a SoC in every system they have built. That is probably coming to the Mac Pro also. Going to make a narrow subset of folks unhappy? Probably. Going to lead to failed overall sales in target market. Very probably not.


Apple Silicon Mac isn't going to boot other OS images. It is going to have the equivalent of the T2 run SSD. ( first couple of transition systems may stumble out the gate without Thunderbolt, but over time all the systems will probably have it ... as they do now). The Apple GPU is likely get added to the list of "if it is a Mac then it has 'x' " kind of thing.




P.S. If Apple eventually got their hypervisior virtualization software up to the point that can assign a second GPU to a specific virtual machine folks could run the Mac off the iGPU and the virtualize instance on a full speed second GPU. ( that would blunt the blowback on the lack of native RAW boot. )

 

[deconstruct60]

So in your mind ~ the first AS Mac Pro would have one MPX bay and one pci 16 slot; which could house an afterburner.
The cpu would be housed in a separate 'box' (top left) and 6 RAM slots next to it. If Apple has its own gpu - where would it

View attachment 1569515

If Apple is going after the more budget constrained crowd with this "half height" Mac Pro then after putting more thought on it I don't think Apple is going to do explicitly target another slot for Afterburner. I suspect that more folks on a budget will punt on MPX modules. Apple will want to enable it. But if Apple can sell a working system without a MPX module at a lower price that too will boost volume sales to those that don't want to buy Apple designs.

If many of the users will just buy a double-wide 3rd party GPU card then the second MPX bay slot could host the Afterburner card for those contexts. ( also for anybody buying a "half" MPX GPU module. ) . That also pushes the folks who need "Full size" MPX GPU + Afterburner into the "Full size" camp. Given Afterburner costs 2K and is in a narrow area... I suspect Apple won't loose sleep over that "push to higher priced system".

Even more use cases where users use a M.2 drive caddy card in Slot 2 along with some other card of choice ( GPU , i/O capture, Network , etc. ) in slot 1.

Just 3 slots covers lots of use cases the rest of the Mac line up won't.


So something like

[ The 2.5" Drive Caddy is a bit different than the J2 bracket in that the drives would stack vertical and consume less space so lots of board height clearance above the DIMMs (which are partially obscured in picture) and upstream from CPU which presumably runs relatively cooler than Xeon W 3000 series top end . Still a large radiator used as it wouldn't be getting as "clean" a flow of air as when directly behind fans. The DIMMs behind the CPU isn't a good idea without some substantive ducting around the CPU. Probably could put a "pull" fan behind CPU radiator here is needed some enhancement in the smaller enclosure (and back to four fans in system). )

That would being in the "half MP height" range. Putting in another x16 slot starts to push substantively larger than just 1/2 reduction in height. It also probably increases the complexity of the PCI-e switching. If Apple's SoC could handle two x16 (PCI-e v4) , seven x4 lanes then could handle those two full sizse slots, the half I/O slot , four TB controllers ( two built in on system and up to two on the MPX connector) , 1-2 10GbE ports , and Wi-fi/Bluetooth + discrete SATA + misc. then would need next to no discrete PCI-e switching. ( the DisplayPort switching isn't any more complicated than the current MP 2019... so can just largely reuse that design effort. )

If this is a limited SoC with only on x16 PCI-e v4 feed than can split that into two x16 PCI-e v3 worth of bandwidth and not any particularly worse than the current Mac Pro. And a Full height MPX GPU module would get full x16 PCI-e v4 bandwidth. Even if switched would be simpler set up than on the current MP 2019 (no A and B zones or PCI-e management for user to do. Part of the A/B thing is so that the Afterburner doesn't get "starved" in more complicated slot usages. Fewer slots , less juggling needed/possible. )


If the DIMM & SoC assembly is less tall/wide as I think they are then there could be room for a x16 single width in there too . The x4 slot , "Lock USB" structure, and 8-pin aux power all shift up higher on diagram on slot width and x16 slides in.

The layout here doesn't "bury" the 8 pin aux power because probably will get used more often.


Several groups can leverage this.

A. "Hate Apple MPX GPU modules and T2 managed SSD drives".

---> buy iGPU only configuration and fiill slot with GPU and SSD . Add a 2.5 caddy and some STAT storage and .. done.


B. "Just want mid range GPU and non T2 managed SSD drives"

--> get a Half- MPX GPU module and SSD drive.

C. " own a mainstream AMD GPU Card in Mac Pro 2019 and Afterburner "

--> get iGPU + Afterburner configuration and add 3rd party GPU

D. " high end MPX GPU module and 10GbE Shared bulk, archival storage "

--> MPX module and some local SATA SSD/HDD storage. Chuck I/O card (thunderbolt slots) and put in a simple x4 M.2 SSD caddy.


E " audio workstation "

--> iGPU only and add on audio card and one m.2 caddy and a SATA SSD for bulky sound libraries (and/or in system backup)


etc.

 

[deconstruct60]

Apple just shrank the Max RAM capacity on the Mini from 64GB down to just 16GB . M1 capped at just two video output streams too.

That really doesn't bode well for some "Mac Pro" class SoC coming out of Apple any time soon.

They actually need to get so something desktop class in the Mini iMac area, before they get to the Mac Pro.

At least, they managed to choke out a connection to just one Thunderbolt controller.

This looks even more getting a "half sized" Mac Pro model because getting a "half sized" SoC to power it.

The MIni went backwards in terms of I/O. The Mac Pro probably is on track to go backwards on I/O also in 2021. 2022 looking more likely of when they'll get eventually get their act together.

 

[fuchsdh]

Apple just shrank the Max RAM capacity on the Mini from 64GB down to just 16GB . M1 capped at just two video output streams too.

That really doesn't bode well for some "Mac Pro" class SoC coming out of Apple any time soon.

They actually need to get so something desktop class in the Mini iMac area, before they get to the Mac Pro.

At least, they managed to choke out a connection to just one Thunderbolt controller.

This looks even more getting a "half sized" Mac Pro model because getting a "half sized" SoC to power it.

The MIni went backwards in terms of I/O. The Mac Pro probably is on track to go backwards on I/O also in 2021. 2022 looking more likely of when they'll get eventually get their act together.

I dunno, I actually think it bodes well for a possible "Mac Pro mini".

Apple decided to jigger with the Mac mini in 2018, positioning it less as "the cheapest Mac" as it had been conceived and into something more about "the edge case Mac serving lots of people". With the notable exception of its graphics performance, it was an excellent machine for those uses.

Today, the Mac mini reverts back to its previous role—I'm guessing the pivot ultimately didn't work. Its lost the "pro" space grey paint job, the two extra Thunderbolt ports, the RAM capacity, the 10gE option, and $100 off its price. The last bit won't be solace to the people who wish for a $599 or even $499 model (although I can't imagine with Apple's margins such a machine would be something you'd actually want, like a return of a 128GB SKU) but there's now much more of a gap that can be filled.

Now, I still don't think it's gonna' be the xMac as some people want. I think it's much more likely something that's going to appeal to the people who bought $2500 or $3500 Mac Pros in the past and felt left out in the cold by Apple pivoting the Mac Pro to a higher-end workstation rather than the midrange it had been for its lifetime before that.

 

[bsbeamer]

Apple could simply release a 27" iMac without the screen and it would satisfy 70% of the xMac/MacMiniPro crowd.

 

[edanuff]

It will be very interesting where this goes. My guess is that the next rev of the M1 chip has some sort of external bus to companion silicon lets them drive more RAM for the 16" pros and iMacs, and then for the Half Pro model there's maybe a bigger version of that companion chip that handles more I/O. Sort of kind of a northbridge, but I'd be careful not to take that analogy too far. I think the important takeaway is that the Apple Silicon strategy is not about just substituting an ARM core for Intel while the rest of the chipset architecture stays the same. I think we might see an interesting situation where Apple keeps spec-bumping and supporting the current Pro for an extended period as a pressure relief valve and lets the Half Pro do it's thing without needing to step into those shoes.

 

[OkiRun]

It will be very interesting where this goes. My guess is that the next rev of the M1 chip has some sort of external bus to companion silicon lets them drive more RAM for the 16" pros and iMacs, and then for the Half Pro model there's maybe a bigger version of that companion chip that handles more I/O. Sort of kind of a northbridge, but I'd be careful not to take that analogy too far. I think the important takeaway is that the Apple Silicon strategy is not about just substituting an ARM core for Intel while the rest of the chipset architecture stays the same. I think we might see an interesting situation where Apple keeps spec-bumping and supporting the current Pro for an extended period as a pressure relief valve and lets the Half Pro do it's thing without needing to step into those shoes.

Yes ~ after information from new AS line, thinking that Mac Pro Midi will be 7,1 runt little brother at first. Less ram, less cores, less ports, less Gpu power. First 8,1 is going half size because not a competition for 7,1 so can be small to accommogdate components AS can handle at first. Yes, half size makes sense now..also, now thinking 7,1 life cycle much longer.

 

[OkiRun]

If Apple is going after the more budget constrained crowd with this "half height" Mac Pro then after putting more thought on it I don't think Apple is going to do explicitly target another slot for Afterburner. I suspect that more folks on a budget will punt on MPX modules. Apple will want to enable it. But if Apple can sell a working system without a MPX module at a lower price that too will boost volume sales to those that don't want to buy Apple designs.

If many of the users will just buy a double-wide 3rd party GPU card then the second MPX bay slot could host the Afterburner card for those contexts. ( also for anybody buying a "half" MPX GPU module. ) . That also pushes the folks who need "Full size" MPX GPU + Afterburner into the "Full size" camp. Given Afterburner costs 2K and is in a narrow area... I suspect Apple won't loose sleep over that "push to higher priced system".

Even more use cases where users use a M.2 drive caddy card in Slot 2 along with some other card of choice ( GPU , i/O capture, Network , etc. ) in slot 1.

Just 3 slots covers lots of use cases the rest of the Mac line up won't.


So something like


View attachment 1569584
View attachment 1569585
[ The 2.5" Drive Caddy is a bit different than the J2 bracket in that the drives would stack vertical and consume less space so lots of board height clearance above the DIMMs (which are partially obscured in picture) and upstream from CPU which presumably runs relatively cooler than Xeon W 3000 series top end . Still a large radiator used as it wouldn't be getting as "clean" a flow of air as when directly behind fans. The DIMMs behind the CPU isn't a good idea without some substantive ducting around the CPU. Probably could put a "pull" fan behind CPU radiator here is needed some enhancement in the smaller enclosure (and back to four fans in system). )

That would being in the "half MP height" range. Putting in another x16 slot starts to push substantively larger than just 1/2 reduction in height. It also probably increases the complexity of the PCI-e switching. If Apple's SoC could handle two x16 (PCI-e v4) , seven x4 lanes then could handle those two full sizse slots, the half I/O slot , four TB controllers ( two built in on system and up to two on the MPX connector) , 1-2 10GbE ports , and Wi-fi/Bluetooth + discrete SATA + misc. then would need next to no discrete PCI-e switching. ( the DisplayPort switching isn't any more complicated than the current MP 2019... so can just largely reuse that design effort. )

If this is a limited SoC with only on x16 PCI-e v4 feed than can split that into two x16 PCI-e v3 worth of bandwidth and not any particularly worse than the current Mac Pro. And a Full height MPX GPU module would get full x16 PCI-e v4 bandwidth. Even if switched would be simpler set up than on the current MP 2019 (no A and B zones or PCI-e management for user to do. Part of the A/B thing is so that the Afterburner doesn't get "starved" in more complicated slot usages. Fewer slots , less juggling needed/possible. )


If the DIMM & SoC assembly is less tall/wide as I think they are then there could be room for a x16 single width in there too . The x4 slot , "Lock USB" structure, and 8-pin aux power all shift up higher on diagram on slot width and x16 slides in.

The layout here doesn't "bury" the 8 pin aux power because probably will get used more often.


Several groups can leverage this.

A. "Hate Apple MPX GPU modules and T2 managed SSD drives".

---> buy iGPU only configuration and fiill slot with GPU and SSD . Add a 2.5 caddy and some STAT storage and .. done.


B. "Just want mid range GPU and non T2 managed SSD drives"

--> get a Half- MPX GPU module and SSD drive.

C. " own a mainstream AMD GPU Card in Mac Pro 2019 and Afterburner "

--> get iGPU + Afterburner configuration and add 3rd party GPU

D. " high end MPX GPU module and 10GbE Shared bulk, archival storage "

--> MPX module and some local SATA SSD/HDD storage. Chuck I/O card (thunderbolt slots) and put in a simple x4 M.2 SSD caddy.


E " audio workstation "

--> iGPU only and add on audio card and one m.2 caddy and a SATA SSD for bulky sound libraries (and/or in system backup)


etc.

Now that you have more AS information, what do you think the I/O port side will look like.

 

[wallah]

I'm shocked there's no thread speculating on the smaller Mac Pro that's in development. I'm betting it won't have Intel inside. Glad they finally listened.

Its getting a little discussion at the end of this thread

AMD RDNA 2, 6800, 6800 XT, 6900 XT MPX Modules

... and 19 pages over in News: https://forums.macrumors.com/thread...r-form-factor-and-apple-silicon-chip.2265185/

Are these two threads candidates for a merge?

 

[deconstruct60]

Now that you have more AS information, what do you think the I/O port side will look like.

Had to work on getting a couple other projects completed today so only got to watch the video and read some of the relatively lightweight marketing material.

I don't know what the desktop/workstation would like because I'm not so sure that this really is a "Mac SoC". It smells far more like the A14X ( destine for next iPad Pro) with a "M1" label slapped on it . Not quite the same as the A12X and A12Z but the same in the sense that I will be not shocked at all if the iPad Pro rolls out with "16 billion transistor , 7 GPU cores. blah blah" which essentially map to the same die as the M1.

I think the "only design specifically for Macs" is fairly decent chance to turning into the "can't innovate my a**" marketing spin of 2020 once long term info sheds more objective light on it. The Metal GPU optimizations they have here probably work just as well on iPad OS apps too. Ditto for the ARM core traces.

I think Apple's "High Bandwidth , low latency" RAM and the caches ( at multiple levels L1 , L2 , large system cache) of the M1 are a dual edged sword. It is goosing their performance much higher than many of their competitors. The back edge though is that there is decent change it doesn't scale without some major pain points.
Their memory controller is doing exotic stuff and getting speed. But if it is "chewing" on exotic memory configurations then there is a chance won't scale well capacity wise. Folks with data working sets in the > 128GB range may be looking for a new platform. (e.g, inference does OK on example photograph/video while large scale training has problematical issues. )

The SSD speeds they are making a big deal about are suggestive that still on PCI-e v3. ( That is bit detached from the bleeding edge GPU cards that can push PCI-e v4 like bandwidths. ) If the micro-architectural version is still stuck on v3 then they'll need something else to be competitive in 2022.

I suspect talk about fastest iGPU will peter out when AMD uncorks their Zen3+RDNA2+DDR5 APU next year. ( van Gogh ? ). [ and since Apple may have been playing around with those on macOS I think they know they are behind eight-ball there. ]

Just managing to do two x16 slots seems more likely now for the "half" Mac Pro. They can't even manage x8 lanes on this M1. So getting to 32 (plus some TB provisioning) is going to be work.

Until Apple rolls out something that is desktop respectable , there isn't much path to see to something higher into the current Mac Pro class I/O.

I wouldn't be surprised at all if the "half" Mac Pro dumped the top/front two Thunderbolt ports. Perhaps replace with USB-C or just nothing but the power button. Even more I/O port compromised that what I had above.

The ports on the M1 are not Thunderbolt 4. Not sure Apple bought Intel controllers and skipping paying for certification or off on some tangent. The teardowns next week should clear that up.

 

[deconstruct60]

I dunno, I actually think it bodes well for a possible "Mac Pro mini".

"bodes well" is nice if don't need the ports or I/O of the current Mac Pro.

If means Apple is going to have to "shrink" the desktop Mac's I/O to fit their upcoming SoCs ? Well yeah. That kind of what I was pointing at.

In the 68K -> PPC transition did Apple have to gimp I/O ports because the new archiecture implementations didn't provision the same level of I/O? No. PPC -> x86 (Intel) ? Again, No. Broad scope outlook for the Mac line up ( from way at the bottom to way at the top of the line up) this isn't a "bodes well". Folks in the middle of the range. Yesh sure that looks good. But from the historic Mac Pro end of the line up, this isn't a good first move.

The range of SoC is going to shrink. And that is going to take Mac system features with it for some Macs.

Apple decided to jigger with the Mac mini in 2018, positioning it less as "the cheapest Mac" as it had been conceived and into something more about "the edge case Mac serving lots of people". With the notable exception of its graphics performance, it was an excellent machine for those uses.

I don't think there is some "role reversal" mode going on here with the Mini. This Mini costs less because you get less. Two less TB ports. fewer screens. Dramatic "drop" in RAM capacity . no enough PCI-e juice to run a 10GbE.
Apple doesn't have another SoC to offer. Either an M1 or nothing. so if the Mini was "nothing" or M1 then hobbled it down and went with the M1. If the iMac is getting an M2 then possibly the Mini could also. I doubt the Mini is pressed at thermal load capacity by the M1. If there was something better they could use it but Apple doesn't have a complete line up ... so they don't.

Today, the Mac mini reverts back to its previous role—I'm guessing the pivot ultimately didn't work. Its lost the "pro" space grey paint job, the two extra Thunderbolt ports, the RAM capacity, the 10gE option, and $100 off its price. The last bit won't be solace to the people who wish for a $599 or even $499 model (although I can't imagine with Apple's margins such a machine would be something you'd actually want, like a return of a 128GB SKU) but there's now much more of a gap that can be filled.

If Mini is suppose to be playing "economy entry" role why does Apple's marketing page go on and on about being more powerful than a much bigger Windows PC. I suspect Apple in part reduced the price to do a comparison with weaker mainstream PC " In the same price range". $7990899 Ryzen 7 5800X beige box system. Do they really want to go "toe to toe" with that?


That is probably because there is no desktop SoC available so they are just chucking something out the door. Says more about "don't want to make a wide variety of SoC" than says something specific about the Mini.

The "Pro" Mini users for web services and clusters I think this still works. And by the time got a Mini up to 16GB RAM and 512GB SSD you have cracked the $1K price barrier. That isn't some "discount" system.

Now, I still don't think it's gonna' be the xMac as some people want. I think it's much more likely something that's going to appeal to the people who bought $2500 or $3500 Mac Pros in the past and felt left out in the cold by Apple pivoting the Mac Pro to a higher-end workstation rather than the midrange it had been for its lifetime before that.