Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Asahi Creator: 1 monitor support is because of hardware limitation
- Thread starter senttoschool
- Start date
- Sort by reaction score
Yes and the people that need two displays put one document on the external monitor and the other on the laptop display.
Most people with a laptop only use the laptop display. A subset of people buy an external monitor for their laptop. A much smaller group use two external monitors. “Pro” is just a shorthand to designate a smaller market that has non-mainstream needs.
It works just fine with an M1 MBA. The sidecar is some kind of virtual display so it doesn’t count against the count of display controllers.
There's always workarounds to get things done in lesser and smaller screens, but comprising productivity and ergonomics. "Most people with a laptop only use the laptop display." Wow you are so quick to dismiss legitimate use cases with your subjective opinion.
You are missing the point that it's not about market segmentation between "pro" and non-pro markets, but a regression in a key functionality. MacBook Air from a decade ago (Mid-2012) supported dual external display officially by Apple (https://arstechnica.com/gadgets/2012/06/new-macbooks-can-manage-many-many-monitors/). This has been discussed since the day M1 MBA was unveiled and I can't believe people are still finding excuse for Apple.
I don't think that you will find anyone here who denies these facts. Yes, in terms of number of displays supported M-series have been a regression (on the other hand they support higher external resolutions). And it's not about finding excuses, but about finding rational explanations for why Apple made these choices.
Pragmatically, I think you are over-counting that 'second' DisplayPort connection. If it cannot be decoupled then it really isn't a second 'connection'. If Apple is hyper optimizing this Display controller complex , there is a decent chance they build that "multi-tile" construct as effective one display connection manager. The 'two' streams are not handled at all in a decoupled fashion. If they are always holistically processed in a synchronized fashion with primarily fixed function hardware then there is zero overhead in synchronizing them back up later for transmission or rendering. ( dropping 24Hz and maybe queuing up going past 60Hz on very high pixel count screens. Maybe even keeping two 4K monitors for eyeballs synched up more efficiently also. )
Apple is just trying to make something that works for their own products ( defacto the LG UltraFine is just Apple 'back seat driving' the product specs. One and only one video input; LG did not think that up. ). It will "happen to work" for a narrow set of other displays, but Apple's objective is expand total addressable market for their own display docking stations. ( so even lowest end of the line up can 'drive' a display that costs 4x the cost of the system. ). There would be just some vestigial reporting of a 'second' DP connection as necessary for downstream drivers or standardized/legacy reporting.
If it was actually a truly independent second DP connection they could have easily gotten TBv4 certification. And yet they skipped it even though the hardware is all there? That would be odd. More likely they skipped the certification because the hardware is not there and they saved some Perf/Watt and/or costs by skipping it. Same trade-offs in making the whole Display Controller complex larger so can hit objectives almost no other general PC market GPU maker is trying to do.
When Apple covers DisplayPort v2 probably that internal "multi-tile as one" concept will just map over to just the one abstract stream they could have now. There could be some reuse of the current system (if implemented this way) for backward compatible to same subset of mult-tile screens have now. Apple skating to where the puck is going as opposed bound to constructs of the past (and willing to insert a stop-gap but it is just incrementally too soon). Short term the Thunderbolt protocols and implementation means the transport between system and monitor goes out as 2 coupled (same destination) streams , but just how late is that more concretely composed?
Again, if macOS doesn't have the concept why build it into the custom hardware building just for macOS? Apple isn't out to build the optimal hardware for Windows 11 and Nvidia/AMD GPUs display controllers for a range of monitors most often not used with Macs.
However, if that experiment worked then it isn't as custom as it looks from the outside.
As good as dated Intel Mac.
Mobile Gen 11 Intel SoC (with Xe-LP GPU tech) offerings went to 4 displays back in 2021.
"...
# of Displays Supported‡ 4
...."
Intel® Core™ i5-1155G7 Processor (8M Cache, up to 4.50 GHz, with IPU) Product Specifications
Intel® Core™ i5-1155G7 Processor (8M Cache, up to 4.50 GHz, with IPU) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.
same in Gen 12 (and HDMI 2.1 were exposed by implementor )
"...
# of Displays Supported ‡ 4
..."
Intel® Core™ i7-1260U Processor (12M Cache, up to 4.70 GHz) Product Specifications
Intel® Core™ i7-1260U Processor (12M Cache, up to 4.70 GHz) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.
even desktop Gen 12 is 4 and HDMI 2.1
Product Specifications
quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.
For lastest Intel stuff you have to drop down the the Pentium/Celeron products to get constrained to just three displays supported.
Current AMD Ryzen 6000 series toward lower end iGPU... A Ryzen 5 6600 (lower resolution cap than Intel in specs , but probably more realistic)
"...
Max Displays 4
..."
https://www.amd.com/en/product/11596
As posted above that 5 is pragmatically an over-count because these "second DisplayPort connections" are not demonstratively actually independent connections. If 100% coupled to another connection then not really a separate one. To get it onto the Thunderbolt protocols for transports it may be labeled that way. But that is more so adapting to the constraints of transport, than demonstration that is actually independent hardware 'channel' .
Apple has made different design trade offs. It is targeted more at narrowing range of monitors connected in a more efficient fashion (e.g., the XDR ). Intel is focused on TB4 port provisioning which (drives two per port). And AMD is keeping up with Intel (Thunderbolt not embedded into the SoC yet, but same wide display target market as Intel is targeting) with slightly different (but closer aligned) trade-offs. (AMD main GPU complex is bigger than Intel's )
When Apple gets to TSMC N3 for the entry, 'plain' , M3 (or whatever) then probably won't see much of a core count increase and that some of that increased transistor budget goes to putting two 'big' display controllers to remove the constraint. ( or perhaps not if DisplayPort 2.0 demand a substantively larger budget. ) N3 could also allow for an even bigger one display controller that did more. This probably isn't a very long term constraint. At least to pick up Thunderbolt v4 certification on the plain Mx entry SoC systems and the Pro/Max/etc SoC has some other features to do better market segmentation on.
higher external resolution which Apple has a prime candidate for sale. Apple's display controller also is not trying to support the widest range of monitors out on the general market either. Not just the concurrent number of displays.
It is a difference between looking at the overall use case model that the general monitor market tries to support versus looking at the Apple docking display user case model. Apple optimizes for their product ecosystem and whatever happens to work that is outside of that is "OK".
The touchbar probably has a minor roll in the constraint limitations also. Probably something in here to optimally efficiently drive it also (whereas the general monitor/PC market isn't really working in that dimension).
Intel ( and AMD) also had different display controller priorities and that is another driver of the regression. Stuff that Apple had bought anyway from those two but really wasn't interested in.
I don't think we can say that until after Asahi has figured out how to do DisplayPort. Maybe they'll find a way to use the extra DisplayPort connections. It would be useful even if they have to do a dual tile setup using two equal sized displays.
I think this is the core point. Apple is not interested in supporting the widest possible array of hardware. Apple targets a certain subset of devices which are close enough to what they offer and what they design for. That's also why folks have reported issues with running monitors with exotic aspect ratios etc.
Yes I use my M1 Pro with dual displays and sidecar.
However bear in mind there's some memory leak attached with using Sidecar in Monterey.
I'm sorry to contradict but I think a little context is needed here.
First of all the Intel Airs that could support two external screens were famously so pathetically underpowered you couldn't use that functionality for anything worthwhile without the machine gasping for its last breath. The M1 and M2 Airs are so immensely more powerful in every measurable metric, they're almost incomparable to the wheezing asthmatic spaceheaters that came before them, and the trade-off, be it for technical reasons or just pure marketing, is they only support two displays: the internal display and one external.
I haven't stood at the Exit door of an Apple store to ask how many people were put off from buying an MBA solely because it does not support two external monitors, but I would expect the quantity to be minuscule. It's just not high importance for most people.
Secondly, the minority of users who want two external monitors 'on the cheap'* can satisfy their needs by purchasing a Mac Mini, rather than moaning that a powerful laptop which wipes the floor with most machines that came before it can only support one external display.
Priorities: on this occasion, Apple have absolutely got them right for most people.
*By Apple standards.
Last edited:
I think you are extending the context a bit too far. We don't need to sing more praises on how M1/M2 Airs are leaps and bounds ahead of Intel Airs in terms of performance and efficiency. Everyone acknowledges that. Let's just stick to the topic of this thread.
1. It's a key regression in terms of I/O functionality. The improvements in any other performance metrics are revolutionary but irrelevant here. If anything, it shows just how close the M2 Air is to the perfect portable Mac.
2. No need to bash intel Airs that hard. "so pathetically underpowered"? The Intel Airs were on top of list of laptop recommendations for almost a decade for a reason. They are still capable machines, not like the 12" Macbook.
3. Suggesting a desktop when others searching for a premium portable Mac to have a one-computer setup both at home and on the go is just not helpful.
The topic of this thread is the limitations are probably likely to be hardware-based. The point I was making, which IMO is very topic-relevant, is most people will take those hardware-imposed limitations, in exchange for the scorching performance/battery life/ProRes, and run to the hills with it in case Apple comes to their senses and turns the Air back into a gutless POS (like the 2012 Air you cited) that can fire an image to a wall of monitors but can't play two streams of 1080p video in iMovie without tripping over itself.
Come on: we all know it was a conscious Apple decision to impose these limitations. Even if if it's a hardware limitation, Apple designed the hardware, so it's still their fault. But, if it's been a case of Apple only wanting to throw a certain number of features at the Air, then IMO, and in the opinion of the hoards of people who are buying it faster than Apple can build it, they got the combination right.
Come on: we all know it was a conscious Apple decision to impose these limitations. Even if if it's a hardware limitation, Apple designed the hardware, so it's still their fault. But, if it's been a case of Apple only wanting to throw a certain number of features at the Air, then IMO, and in the opinion of the hoards of people who are buying it faster than Apple can build it, they got the combination right.
I think the important topic to discuss is about the merit of Apple's approach. To refresh, it has been observed that the display controller on M-series chips occupies a significant amount of space — the two display controllers on M1 die are similar in size to the P-CPU cluster. This of course explains the lack of additional display support — die space cones at high cost after all. And these display controllers are truly high-end devices capable of very efficient operation due to their own large RAM buffers as well as 6K resolution support.
Still, one can argue whether it's a good use of die space. Is an ultra-efficient display controller really necessary with external displays for example. Maybe it would make sense to have asymmetric display controllers — a "fancy" large one for the internal display (for excellent battery life) and multiple smaller ones for the external displays. These smaller controllers could for example use smaller internal memory buffers and operate less efficiently, but one could fit more of them on the same die.
P.S. Just to be clear before the interpretations start rolling in — I am not advocating anything in particular, I am merely raising an academic question. Personally, I don't care about running multiple monitors and the current solution works perfectly fine for me.
Two? The die floorplan labeling in the first post is of the M1 Pro; not the M1. The Pro/Max have two, but are also substantively bigger dies. The M1 also has fewer Thunderbolt (TB) controllers. Can't really increase the TB controllers without increasing the Display controllers (and still try to hit TB4/TB3 certification requirements. ). Four TB soak up substantive die space also. But having them integrated saves a substantive amount of power.
Second, it two Display controller complexes are approximately as bing as one (of the two) P core clusters. Not bigger than the two P core clusters. if you round up just the P cores without L2, the DisExts are close to just the P cores. Throw in the P2 core L2 cache and it isn't that close if compare two to two. That was a bit of Apples to Oranges comparison. The Display cores with substantive framebuffer working space RAM and the P cores with registers and L1 cache. Yeah substantive RAM capacities takes up more space on the die.
The cache replacement policy for the Display Controllers is going to be largely different from the GPUs and CPUs (and NPUs). So there are several upsides to given it is own working space if that would partially trash the shared L3 case of the CPU and GPU (both line replacements policy mismatches and bus congestion to/from the L3 ) ... those could have to take a performance hit. ( similar coupled resourced issues when the Display controller runs hotter and there is thermal bleed over into the other function units. More P cores that run into a thermal limit is helpful how on general workloads? )
One of the issues Apple is juggling here is there they are more than a few relatively 'hungry' bandwidth consumers sharing the same stuff. GPU want lots. CPU when fired up on high multiple threads want lots. Video decoders are shipping lots of output around on the internal bus. NPU are bursting on computations. The display controllers have higher latency tolerances to meet service level requirements, etc. They all have to share and also not share as appropriate.
It is also likely the DispEXT allocation is also coupled to the TB controller allocation. So where the TB port count is low then the display count is going to go lower in lock step.
Folks get caught up in the P core counts. If Apple is trying to build a System on a Chip then need to have more than maximum P core count. If there is primarily just P cores on the die, then it isn't much of a 'System' ( well rounded , completed package).
I/O off the chip is where they are going to take lumps on power. That is one reason why Apple is using PCI-e v4 more so to reduce the number of pins out from the SoC than the generally increase bandwidth off the chip. ( x4 PCI-e v4 == x8 PCI-e v3 as opposed to trying to tag larger overall bandwidth off the SoC than the Intel packages replacing. They aren't. ]
I suspect the single Internal display ( that gets no where near 6K ) is the problematical power and bandwidth consumer problem here.
And that if want to have "world's fastest" iGPU then some of the other internal function units have to get some compensation for the "bandwidth hog" that GPU is going to present. ( GPUs get the bulk of the transistor budget but going to get some die grow because other units need room also just to share with that "problem child" .)
Sorry, my mistake.
As to the rest: excellent and insightful write up from you as usual.
No problems. Lots of content flys by rapidly on Macrumors.
This DispExt makes some sense on the laptops but on the desktops there is likely a tipping of the cost/benefit trade offs. Once start doubling up laptop targeted dies this is lots of space with diminishing returns.
M1 Pro 2 Units
M1 Max 4 Units
M1 Ultra 8 Units.
If go to the tech specs on the MBP 14/16 and Studio.
M1 Pro 14" can drive two 6K displays.
M1 Max 14" can drive three 6K displays (and on 4K HDMI) .
M1 Max Studio can drive four 6K displays ( and one 4K HDMI).
then curiously the Ultra has same limitation. ( probably were Apple's tech spec page drifts itno being the marketing spiel page. To get TBv4 certification six TB ports should drive six 4K displays. if can also do the HDMI at 4K then seven. ). For the Ultra pretty good chance has one whole DispExt sitting around doing nothing ( perhaps a 'spare' so can squeeze out very incrementally better yields. )
A quad , laptop optimize die would be drifting into the wasteful space zone. 16 units and probably still only 7 video out ports. That would be a more dubious die space expenditure. Apple really does need a more desktop die if only to use two pair with once go past 1 . Something like a Pro's 2 Units allocation and spent the rest on something else that don't use in laptop space. Or even just staying at the Ultra's 8 units allocation; which is already 'too many'.
The "Ultra" being an exact twin made sense for generation 1. But to be effective in the desktop it really shouldn't be an exact twin. It may overall be cheaper for Apple to skip doing desktop specific , but it would not be surprising if how the Ultra package changed a bit in gen 2 ( or perhaps gen 3 ) as Apple builds bigger sales of the top end of the desktop line up.
In terms of DispExt perhaps something like
new Ulra 4 + 2 = 6 units (and 7 outs on box )
new "Quaddro" 4 + 2 + 2 + 2 = 10 units ( and 9 outs on box. But if 7 outs better to be 3 units over needed than 8. )
The extra large DispExt work effectively on several criteria, but they don't unit scale well.