I don't think the video output has any influence on amount of memory bandwidth other than the bandwidth required to generate the image itself. Bandwidth required for additional video output definitely isn't borrowed from main memory.
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.
Predicted display support on M2? More displays than M1? 8k? HDMI 2.1? DP Alt Mode2.0?
- Thread starter theorist9
- Start date
- Sort by reaction score
I don't think the video output has any influence on amount of memory bandwidth other than the bandwidth required to generate the image itself. Bandwidth required for additional video output definitely isn't borrowed from main memory.
Yes, it is. that is why they have a Unified Memory Architecture. Still, I don't think memory bandwidth is the limit here. more likely a marketing decision.
I am talking about the video output, not the rendering. Memory bandwidth doesn't have anything to do with bandwidth required for video output such as HDMI, DP etc.
Display Port Alt Mode 2.0 only requires 40 Gbps Thunderbolt 3 cables. Those cables can support 80 Gbps for sending display data. The only non-display data available would be USB 2 though. People expect more than USB 2 hubs on their Thunderbolt/USB-C displays.
are you sure that Macs don't already support dp2.0 alt mode?
wikipedia is saying that TB4 does, so...
There goes the entire raison d'être of the Macrumors website!
The usual quoted required bandwidth is 18Gbps for 4K@60hz and around 24Gbps for 5k. The later is roughly 5% of all available bandwidth on M1. Of course, there is likely some bandwidth-saving compression happening, but who knows really.
You still need to get the frame buffer data from the system RAM to the display controller. How are you going to do that if not using the system RAM bus? The alternative is that display controller owns a large pool of physically separate GPU-writeable memory from which drawables get allocated, but that’s a bit of a stretch. Not to mention expensive.
Amount of bandwidth required for the frame buffer data isn't significant enough to make much impact on M1 performance, let alone future M2 with potentially more bandwidth. None of the tasks performed M1 can't saturate 100% of theorical bandwidth anyways, it has a room for high-bandwidth displayports.
I have two 27" Thunderbolt Display which are not both supported (simultaneously) on any regular M1 machine (M1 Pro and up do work). I'm waiting on a M2 Mini to hopefully drive both of these at work. The M1 pro is way more than I need, and for 2k I'd rather have a 32gb 2TB M2 machine instead of a base M1 Pro.
Not at all. I was just assuming Apple didn't have it b/c it's a cool feature, and I've not seen either Apple or the various tech journalists that write about Macs mention it (Apple says "Display Port", but nothing more specific). In fact, when I Google "Display Port Alt Mode 2.0 Apple" with a restriction to find articles from the past year, the first listed result is this thread 😁!
I initially thought this is something Apple would want to have (and want to say they have), since it would enable a digitally lossless connection to their Pro Display XDR. However, the XDR uses TB3, and thus wouldn't support this standard (i.e., it doesn't have the ability to accept an uncompressed 60 Hz signal).
But you raise a good point: If it's an obligatory part of the TB4 certification standard, then anyone who claims TB4 (as Apple does) would have to offer it. The problem is those standards are hard to find. The only link I found: https://www.thunderbolttechnology.net/tech/certification
is broken.
But in the other reading I did, while DP Alt Mode2.0 can be run on TB4, it appears it's actually part of the USB4 specification, rather than TB4. And it wasn't available until after USB4 was released, so it's probably not written into the USB4 standard as a requirement. Net net, it looks like you can claim both USB4 and TB4 w/o having to implement DP Alt Mode 2.0. But this is just my best guess.
Last edited:
DP Alt Mode 2.0 actually doesn't use double the bandwidth of USB4. Each USB4 interface has two 38.88 Gb/s "channels", enabling it to transmit at that rate in both directions simultaneously. But when feeding a monitor, you only need to send data in one direction (not including status reporting the monitor might send back). DP Alt Mode 2.0 takes advantage of this, using both channels to send data to the monitor, yielding a unidrectional bandwidth of 77.37 Gb/s (not exactly 2x 38.88, perhaps because it reserves some bandwidth for the monitor to send status info. back to the computer). I.e., DP Alt Mode 2.0 simply turns a duplex system into a simplex system. The bandwidth is essentially the same; only the directionality has changed.
Last edited:
I think it’s important to note that RAM bandwidth is a bit of a weak point of M1 (unlike Pro/Max/Ultra with their very wide LPDDR5 interfaces). M1 only has 67GB/s of bandwidth available, but wants to compete with entry-level dGPUs which often have much faster VRAM. Two external 4K displays plus the internal one add up to around 6GB/s, reducing the available RAM bandwidth by significant 10%, and thats just the needs of the display controller. One also needs to render to those frame buffers (here the required bandwidth will depend on what one does). So yes, additional external displays can significantly cut into the RAM bandwidth economy, potentially to the level that limits performance elsewhere. So I don’t think we can discount the possibility of Apple limiting the external displays on the first-gen M1 in order to guarantee a certain minimal level of performance.
Not entirely accurate. iOS/iPadOS can both drive external displays separately from the internal display. The real difference is that iOS/iPadOS doesn’t manage the external displays itself (i.e. the windowing system is still primitive compared to the Mac). However, an app can absolutely watch for an external display and do whatever it wants with it by providing a UIWindow to be displayed. An app can also hijack AirPlay mirroring in the same way by providing a specific UIWindow to display over AirPlay.
A vanishingly small number of apps have put in the work to do it. It’s not hard to do. It’s more that there’s not enough demand for the functionality if only one app can control the screen at a time. But I know it’s possible because I have written a small app that uses the functionality.
Correct. The USB-C alt mode isn’t a requirement (either in the USB spec or the DP 2.0 spec). And I suspect the controller blocks are still in development that will go into GPUs/monitors/etc to support DP 2.0. On the PC side, maybe we’ll see with it the next GPU gen?
But my understanding is that the alt modes are all optional, but some like the DisplayPort alt mode are common. Thunderbolt 3/4 is a USB-C alt mode, just one that‘s been rolled into the USB-IF, and remains optional for USB4 as well.
I wonder if there’s a noticeable difference benchmarking an M1 Mini headless vs w/ a monitor.
As @theorist9 says, it turns a duplex system into a simplex system. It just doesn't have the bandwidth to carry both DisplayPort *AND* USB3+ speeds at the same time. People may *want* 4K+ screens and more than USB2 speeds, but AFAIK, it's not possible with current connections. If/when Thunderbolt moves to PCIe 5+, maybe then we'd get the bandwidth. Unfortunately by then, we might have 8K screens and people complaining how they "expect more than USB 3 hubs on their Thunderbolt/USB-C displays." That's what sucks about technology today, the amount of data increases, but the connection speed doesn't increase at the same rate.
Of course, if Apple releases an M2 Pro Mini at WWDC, we will soon have an M2-generation chip with support for at least 3 external displays and 32 GB RAM.
Since, on the MBP, the low-binned M1 Pro (8-core CPU, 14-core GPU) is $500 less than a comparably-equipped (RAM & SSD) low-binned M1 Max (10-core CPU, 24-core GPU), I'm guessing a low-binned M2 Pro Mini will be $700 less than the comparably-equipped low-binned M1 Max Studio ($500 less for the chip, $200 less for the less expensive box), and would thus start at $1200. [That's assuming they have a doubly-binned (CPU & GPU) M2 Pro the way they have a doubly-binned M1 Pro; otherwise it would be higher.]
Since, on the MBP, the low-binned M1 Pro (8-core CPU, 14-core GPU) is $500 less than a comparably-equipped (RAM & SSD) low-binned M1 Max (10-core CPU, 24-core GPU), I'm guessing a low-binned M2 Pro Mini will be $700 less than the comparably-equipped low-binned M1 Max Studio ($500 less for the chip, $200 less for the less expensive box), and would thus start at $1200. [That's assuming they have a doubly-binned (CPU & GPU) M2 Pro the way they have a doubly-binned M1 Pro; otherwise it would be higher.]
Last edited:
If Apple was to release a 6k-7k resolution monitor with 120Hz refresh rate, let's say a successor of XDR, it's pretty much settled it wouldn't run at 120 Hz with current M1 lineup, right? Assuming M2 will most likely be more of a single core spec bump and possibly more cores without a huge architectural jump in overall chipset design, such a monitor would still have no way to be driven by a M2 machine - unless via some weird design like double Display port.
I guess the conclusion is there won't be a 120 Hz 6k-7k res display anytime soon
I guess the conclusion is there won't be a 120 Hz 6k-7k res display anytime soon
...there's a difference between what you can get away with and what is reliable enough for Apple to officially support. Last I looked, although the retina MBPs could drive 3 external displays, there were some reports of overheating.
...two external 4ks, plus the internal retina display, potentially in fractional scaling mode in a fanless machine with potentially only 8GB of unified RAM from which to allocate video memory? I suspect that your mileage may vary.
...yes, probably because it uses LPDDR5 which supports higher-capacity chips. The regular M1 uses LPDDR4. As I said - the M2 could support 32GB since it is likely to use LPDDR5 (like the M1 Pro and Max, although they also have more RAM channels and support more RAM chips). Still, Apple would have to decide to offer the option.
Apple are presumably confident that enough of those people will go for a 14" MBP instead.
A 6-7k display at 120ghz would need over 60-70Gbps video signal bandwidth, maybe even more for wide gamut stuff. That's out of reach for Thunderbolt 4.
I doubt that Apple will "debut" the M2 Pro in a Mini when their flagship 14/16" MacBook Pros would be first in line for that honour - and they're not even a year old yet (plus they have a large backlog of orders). It would also cannibalise the recently released Studio Max.
OTOH all of the "regular" M1 machines are ready for an update, so the more plausible rumours are for a M2 Mini alongside the M2 Air - and I think that's still likely to be of interest for people waiting for a M1 Pro Mini. But then the rumor mill got themselves in such a thorough knot with the Mac Studio that I don't give them much credulity at the moment.
(nb: it's also slightly odd that they'd announce a new MacBook Air unless they could have it shipping in quantity in time for the US "back to school" season - so while I wouldn't rule out an announcement at WWDC, my money is on a "preview" of the Mac Pro as the only hardware launch).
An M2-Pro Mini does not necessarily mean that the regular M2 will support more displays or RAM than the current M1. It just means that the M2-Pro will.
Last edited:
You know, given the current delivery delays I don't think Apple is worried about cannibalisation. The more folks can get a Mac into their hands sooner, the better for them. There is still that one awkward Intel Mac Mini Apple sells, would be nice to have it replaced by something decent.
...that assumes any new Macs won't be subject to the same supply problems as the current ones. They've got a lot of orders "in the bag" which could be switched if something new comes out that's more readily available - and a "Pro" Mini would be a significant "downsell" from a Studio.
Sure - there's a "M1 Pro Mini"-sized hole in the lineup - but Apple could have filled that any time since last October, and they haven't. However, as I've said before it's a bit late for a new M1 Pro machine now because even the regular M2 could give it a run for its money - and possibly beat the low-end binned M1 Pro - and I'm sure that the debut of the M2 Pro/Max (assuming it follows the same path) will be reserved for the 14/16" MacBook Pro update, which isn't quite due yet.
I wonder if Apple considered and then ditched the prospect of M1 Pro-powered desktops - it would explain the bizarre case of two 24" M1 iMacs with different cooling and port arrangements, and the demise of the lower-end 27" iMacs.
Right? It's odd...My completely uneducated and arbitrary guess is a shortage of LPDDR5X modules.
Yeah, I was thinking the same