So there's a lot here, I'm going to respond by subject rather than by person:
I don't see Apple using anything but the Mn Max SoCs for a dual/quad SoC configuration, they may offer a reduced GPU core variant (48-core GPU) but I doubt they would reduce the CPU core count...?
As for RAM, the minimum you can get with any M1 Max SoC is 32GB, so it stands that a dual M1 Max SoC configuration of the Mac Pro (Cube) would base line at 64GB total RAM (across both SoCs)...?
Disabling the CPU cores will do very little to lower the bandwidth demands of the GPU cores. The core infrastructure of the Max die variant is the likely the core for a duo/quad set up ( much of the design effort is being shared across all of these to control costs. ). The "floor" for a Max die on RAM is 32GB. So any "dual" (with twice as many GPU cores) would be at 64GB. Since trying to compete with dGPUs with GDDR6 (or better) the LPDDR5 has to go 'wide' to keep pace on throughput. That "go wide" drives up pragmatically the base RAM size. twice as many GPU cores leads to twice as many memory channels needed. Which leads to RAM attached to all the channels. ( economies of scale will lead to same semi-custom memory modules using on Max also used on "dual" version also. For a relatively much lower volume SoC there is little chance it can go off on its own and get the new semi-custom memory packages done cheaper at lower volumes. )
The catch-22 on Apple's. highly unified approach is that can't decouple wanting several more CPU cores from getting more GPU cores ( and Memory). If want biggest GPU core count with small number of CPU cores... that doesn't happen either. If want high core counts of either .... LPDDR5 module bandwidth means have to raise the floor of the memory bought directly from Apple. Want to max out any one of CPU cores, GPU cores, or Memory capacity ... then the others are also along for the ride. All coupled together with the shared die infrastructure.
None of that is going to lead to being the "low cost leader". ( The only way they could limbo down would be if they also build a M1 Pro (Jade-Chop) dual. Where they substantively cut the number GPU cores. That also seems doubtful since the dual/quad volumes are going to be much smaller. Just doing probably two variants ( Max mod , Max mod mirror with more general I/O ) is going to cost. Taking a Max sized die and turning off half GPU ... probably wouldn't lower the cost dramatically either. The package will be just as big. They will Likely charge the same price for the Max sized die as if only binned off a small number of GPU cores ( it isn't costing Apple anything less to have made. At modest scale throwing away extra wafers that could have been used to make significant amount of other product. ).
To get so something like $3.6K Apple would pretty much likely would do the slap a 512GB SSD in there. As soon as leveled back up to "Pro" > 1TB levels the system would be back over the $4K price point. [ like the MP 2019 that makes some sense for folks who either "hate" Apple SSD storage or virtual machine workloads where the most often used "macOS" drive is placed somewhere else (network or mainstream SSD ).
While I see the iMac Pro utilizing current chips, I still see the Mac Pro as being a custom beast that will, by default, be very low volume. BUT this will only be the case if they stick with the thinking that brought us the most recent Mac Pro… “anyone who wants something less powerful already has something they can buy, even if it may not be in their preferred form factor.”
In referring to cutting CPU cores, I was referring to similar for the M1 Pro, the base model dual die could turn off several CPU cores (using dies that would otherwise not pass the CPU tests) for cost reasons, not bandwidth. In actual fact, the floor remaining at 32 GB and 400GB/s would likely be fine for the dual die variant. The Anandtech reports were that the 400GBs was overkill for the Max chip and couldn't find a workload or even a set of workloads to saturate it. Nvidia 3070 which is 20TF (more than the 48 core GPU variant would be, about what the 64-core variant would be) with 448GB/s bandwidth off of 8GB of GDDR6. So Apple's solution has similar bandwidth and lower latency even at 32 GB. Further the memory is on-package but off-die. So there's no manufacturing (or performance) reason they couldn't make a 32 GB Max duo chip. Apple may still choose to make the floor 64GB. They just *don't have to*.
However, even at 64GB minimum RAM, my analysis holds. The 1TB 14" Max model with 64GB of RAM is about $3.7K. Yes the duo chip will be more expensive and there will be more expensive IO for Mac Pro Mini but will also not be shipping with battery or extremely expensive screen. Apple is charging a minimum of $600 for the latter two. Apple again could choose to make it $4K, but it wouldn't be against their current pricing model to price it lower. In fact, I would view that $4K as an upper bound in expectation though I would agree that $3.5K is probably a lower bound if the minimum is 64GB of RAM, 1TB SSD, 48 GPU cores, and 20 CPU cores.
A dual M1 Pro is an interesting idea however I agree I don't think it likely though for different reasons (see below).
While I see the iMac Pro utilizing current chips, I still see the Mac Pro as being a custom beast that will, by default, be very low volume. BUT this will only be the case if they stick with the thinking that brought us the most recent Mac Pro… “anyone who wants something less powerful already has something they can buy, even if it may not be in their preferred form factor.”
While of course the Mac Pro will be lower volume than say the Air or mini, they are definitely not starting it at $6K and it will be expected to be higher volume than current Pro desktop models. While we disagree on how low it will go, I think even
@deconstruct60 would agree with me here that a $6K starting point is out. The difference between us in the above posts is about 10%. He's at $4k, I'm at $3.5. A $6K starting price for the As Mac Pro, only for the smallest slice of users, just doesn't make sense with how Apple has priced things so far or frankly how the chip is likely to be designed/perform. While still expensive at $3.5-4K, that's at least a 1/3 cut in costs aiming at a much bigger market than the current pros. Again their current model seems to be to stick a single chip in as a wide range of form factors as possible and let the form factor and thermals take care of the value/performance. We're entering a different product stack model than the old one.
I wouldn't expect anything beyond the M1 variants (M1 / M1 Pro / M1 Max / Dual M1 Max) for the initial two year transition period, with the M2 variants starting to release in the Fall of this year; so the tail-end of the transition...?
Yeah, it was a hope of mine that the 2-year transition would include M2 chips for more of it. It won't and that's okay (although again, rumors are that sourcing the mini-LED screens is the major issue rather than the chips themselves being unavailable). It's just somewhat disappointing.
I can see no real good reason for Apple to not have a M1 Max variant Mac mini; Mn Pro for those who need CPU over GPU, Mn Max for those who need the extra GPU power...?
One thing that is STARKLY different between then and now is that a fairly high end processor like the Max could actually easily go into a Mini. With the non-mobile market selling less and less year over year, though, it could be that their market research shows there’s not enough interest to make a product to serve the Max-in-a-mini market.
The M1 Max in a mini possibility is again, why I think the pricing will be tied down. I'm near certain that the mini will get a Pro chip and there's no technical reason why I can think of that it couldn't get a Max, even with a redesigned smaller chassis the thermal capacity should be good enough to outlay 90W of heat - that fits most SFF PCs smaller than the mini unless they go SUPER small and thin. So I think it likely. The only reason I'm not as certain as I am for the Pro is as
@Unregistered 4U says if Apple has done some research that this variant wouldn't sell for some reason. Equally as
@deconstruct60 says market research could easily have said it'll sell well and not putting it in would go against their current model of stick their chips into as many form factors as they can. So again, much more likely that they will put a Max in a mini than they won't - I'm just not as certain about it as I am for the Pro in a mini.
Errr what rumor said the Jade was the same thing as the Jade2C. The Bloomberg rumors was that there was a Jade-chop , Jade , Jade2C , and Jade4C being developer. Examining a Jade die doesn't necessarily tell you that that it is a Jade2C die.
The IRQ for multiple die interupt tracking is in both the Pro and the Max. That can easily be just a common utility dropped into the shared "top half" across all the variants. Most cost effective to lay it out once and just leave as a minor unused "blob" on the ones that don't use it. It isn't significantly skewing overall die size. Just like Jade-chop (Pro) looks alot like Jade (Max) that is probably true also of the Jade2C and Jade4C. Make relatively minor changes to Max as oppose to "chopping off" the GPU and second video+NPU complex. Could just drop the video+NPU complex and convert that to die-to-die connect for Jade2C. Jade4C would be the same size but would drop Thunderbolt Controllers for PCI-e v4 complex and also have the die-to-die connect (possible also mirror of Jade2C to ease memory layout) . 4 dies ... with a very substantial amount of work only have to do once. That saves Apple lots of money. Take all that "saved" money and work on a very fast, low latency chip-to-chip interconnect technology to remove most of the NUMA penalties with cross die data traffic.
The actual Max (Jade) die would never be used in a 2 or 4 chiplet/tile package at all (that's would explain why 2C and 4C have a 'C' in their name. They are designed to be chiplets). Just like the Pro die (Jade-Chop ) wouldn't be used either.
Apple could have done the Jade-2C first. Then modified it with a "chop" and gotten the Jade-Chop with very little extra work. In parallel swap out the chip-to-chip (because know the ProRes thing is working) and put in the second video complex and got Jade (Max). Do another set of focused mods of the same working chip into a 4C. [ and if chip-to-chip communication system doesn't work to expectation can abort 4C. If have trouble scaling to two then four isn't worth doing now. And finding that out sooner rather than later is much less expensive. ]
Just looking at a M1 Pro ( Jade-Chop ) die wouldn't necessarily would have gotten an completely accurate picture of what a Max was. Probably the GPU doubling. However, the second video/ProRes/NPU complex ... that doesn't necessarily follow from the top. So if Jade/Max is actually a redacted/moded die than not really going to get the full picture of a Jade2C from it.
Actually the IRQ multiple die interrupt is in both, but only the Max actually supports multiple dies and while it does support 2, but it doesn't support more:
Now you're entering an area past my expertise level. So I'm just going off of what Hector is saying here. You seem to have a difference of opinion.
Basically it looks like the Max either is or is intimately tied to the die that will be used in Jade 2C and it isn't built for 4 dies. Could they have another one basically identical to the Max that can be used in 4 dies? Sure. To me, that seems possible. But the evidence so far is that the Max Jade die is at least what they are basically planning on using for the 2C and is not what they have planned for the 4C.
Agreed, but if someone’s happy with their current laptop AND a new laptop from their current vendor (running the OS they’ve become accustomed to) would be less expensive and have higher performance than an MBP, then I see folks on old Macs upgrading to newer Macs most assuredly, but I don’t see anyone pros coming back from Dell, Acer, etc. You’re absolutely right that future numbers will show if it’s just a long pent up Mac user upgrade cycle.
Well on the CPU side, they're still waiting for such laptops as Alder Lake mobile devices technically haven't shipped yet and not clear where the price/value proposition will be for it. I know people initially thought ... "wow desktop Alder Lake looks so price competitive!" Then they saw the increase in costs for (good) motherboards and DDR5 RAM
(which you definitely need when you are competing against the M1 Max and upper-end AMD laptop CPUs). And AMD 6000 series are still not here yet either.
On the GPU side I'll grant you, it's a 10TF GPU with no ray tracing. It can compete with a bigger in GPU in certain scenarios due to its TBDR design, but the application has to be written with that in mind. But again, software is coming along and that's really the primary driver right now that I see of interest. People want to port their software to Metal in a way that they really didn't for years before this. That's a *very good* sign. It also helps that Apple is taking a more active interest as well. Hopefully that continues since this is a marathon not a sprint.
I’ll have to look into that rumor as wishful thinking plus Xcode references previously indicated Apple’s transition to AMD I’m still of the opinion that Apple said they would still ship Intel machines after their announcement and they’ve already shipped it. I don’t think there’s more Intel.
What new Intel Mac have they shipped since the transition? The only I one I know of that was even rumored was the Mac Pro. It's also been a rumor from several sources and that combined with the Xcode is usually a pretty good sign. It could still be wrong. And it's less wishful thinking (at least on my part), honestly I think it muddies the transition. However I can think of a few reasons to release such a device:
1) it can have far more memory than what the new MPs will be able to offer and Apple doesn't want to lose those customers until it comes up with a solution
2) customer investment into the peripherals/internal units that may not work with the new towers (especially GPUs which definitely won't)
3) Apple only getting basically one generation out of the tower form factor and wanting to recoup design/research costs
4) admission that some Pro software will take longer than others to come over and wanting to give some Pros more time.
Some of these are not very Apple like to care about. So I dunno, but the rumors are pretty sure it's coming.
