I haven't looked at GPU code, or higher-level libraries for GPUs, since before metal or cuda existed, so I have no idea about this. Can someone who is familiar with both cuda and metal comment on the likely feasibility of a translation layer? I sort of assumed it wasn't easy (even possible?) since it would be such an obvious step otherwise, but I don't actually know.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple's M3 Pro, M3 Max, and M3 Ultra Chips Could Offer Even More CPU and GPU Cores
- Thread starter MacRumors
- Start date
- Sort by reaction score
Last time I've checked metal (or the gpu) doesn't support double precision (64bit) numbers, and that makes the idea DOA.
Building every chip with the capability it's already has then removing it costs something. The chips were designed to take advantage of multple displays. Then they cut the feature out. Someone some analyst said " sure it will cost X$ to remove this feature , but we will make X$ by forcing those users to upgrade to a higher tier machine.
Totally an apple thing to do.
Apparently having multiple displays is a PRO feature. Definitely not something average people use....
Thanks Apple. My new i9 Asus laptop for $960.00 and a 4k 144hz hdr display works amazing. I made the right choice over the MacBook air. And saved money.
That is my biggest complaint about the 13" Pros. It wasn't the end of the world (it's not something most people will do often), but the situations where I found myself annoyed at my workdesk because I couldn't plug in multiple external monitors made the laptop feel much less premium to me. On the times when you do want to do it, it leaves a lingering feeling of having invested in something lower end.
I get the market segmentation thing, but I'd prefer that they didn't do market segmentation over smaller things like this. There are plenty of ways they can make the 14"/16" systems worth the upgrade (performance, screen size, etc) without making tradeoffs over comparatively simple things.
While I agree Apple as a business is shady at times, but your above assessment as to how Apple Silicon chips are designed is, well, wrong.
They DID design the M1 and M2 chips to be limited on display capabilities. The Display Buffers are hardware portions on the SoC die, which can even be identified by x-ray die shot and can be counted. And for people trying to render a very high resolution screen, such as 5k HiDPI on 8K, or 4k HiDPI on 6K, they have discovered the Apple Silicon framebuffer resolution limit to be quite small, something like less than 10,000 pixel wide. In contrast a low class dGPU or even an Intel iGPU are in the ballpark of 65,536 pixel wide.
And as I noted above, strategically Apple chose to use the same M1 M2 chips across desktop Macs, laptop Macs, and tablets, while considering all use-cases over all these product segment there is an overwhelming evidence that number of external display is not a priority. You can look at it this way: The MacBook Airs just happen to be using a tablet chip that's powerful enough, but not necessarily having all the features of a "traditional" laptop chip including external display capabilities.
I just feel the need to point this out, it is not like Apple just artificially castrates its lower tier products in hopes of forcing users up the product ladder. In fact their balls are deeper than that, they designed the chips with hardware limitation much lower than previously with Intel / AMD, because they feel the products these Apple Silicon chips go in do not need those higher capabilities.
The lengths some of you go to defend the undefendable...
When did it become the norm for Apple products to have designed limitations instead of a wow-factor?
M1/M2 doesn't have entry level performance. It's powerful enough for a lot of professional stuff. What Apple does is limits it's hardware to casual consumers with stuff like these, and the higher-end machines to graphics professionals with the rigid specs and excessive pricing. And they wonder that the sales drop. They are making a historic mistake by focusing on short-term profits instead of grabbing market share from pc while they have an edge over intel and amd. All while they are supposedly try to shift to more service revenue.
You do realize we are agreeing with each other right, if we strictly speak of accusing Apple under-delivering for its consumer range products.
I was only trying to clear up your bit of "The chips were designed to take advantage of multple displays". Which is not true, it is not just a software switch like some others would suggest. In fact they went even further, by designing inferior hardware in the first place.
Mostly, but not supporting two external displays in clamshell mode - which would be the case for most of us complaining about it, or not implementing displayport daisy chaining is a software limitation. The latter worked on intel macs with windows, but not with macos.
That was indeed the case with Intel Macs, under bootcamp I believe MST worked flawlessly in Windows 10. Though I am unsure if Apple's reason is like you said an intentional up-selling, or they couldn't be bothered to write the relevant parts in drivers and macOS for MST to be supported. Some ventured to guess they wanted to push for display daisy-chain being exclusive to Thunderbolt Displays, since that's what they sold, but ever since TB3 the displays no longer have an output (or have enough bandwidth for a 2nd 5k anyway).
Now with Apple Silicon it seems the bottleneck is all the way back to the (absence of) hardware.
B
The M2 chips in the MacBook Pro is the same M2 chip in the MacBook Air M2 model. Same die, same size.
The Pro machines HAVE the capability of multiple displays. Apple specifically cut down the functionality of those lesser model chips to stick in lower tier models. And is charging premiums for these devices. They specifically designed the chips with multiple display functionality and REMOVED IT, by choice on lower tier machines.
And no one is asking for 4 extra 8 K displays on a MacBook Air… that’s obviously a pro functionality. But the fact a 13 or 15” M2 MacBook Air can’t run 2 , 1080p external displays is a joke.
Apples loosing a lot of business because not every professional has $3000 for a laptop. The sweet point for professional machines is $1500
And the PC side of things hands down slaps apple silicons capabilities out of the sky. Yes they are efficient and run quiet and at low temps. But the PC segment is quickly catching up. Apples illusions of grandeur in its walled off garden have people seeking to peek what the world is like outside that garden. More and more people are looking to break that wall down and experience life outside of it.
I don’t think you understood my above comment.
The M2 chips in the MacBook Pro is the same M2 chip in the MacBook Air M2 model. Same die, same size.
The Pro machines HAVE the capability of multiple displays. Apple specifically cut down the functionality of those lesser model chips to stick in lower tier models. And is charging premiums for these devices. They specifically designed the chips with multiple display functionality and REMOVED IT, by choice on lower tier machines.
And no one is asking for 4 extra 8 K displays on a MacBook Air… that’s obviously a pro functionality. But the fact a 13 or 15” M2 MacBook Air can’t run 2 , 1080p external displays is a joke.
Apples loosing a lot of business because not every professional has $3000 for a laptop. The sweet point for professional machines is $1500
And the PC side of things hands down slaps apple silicons capabilities out of the sky. Yes they are efficient and run quiet and at low temps. But the PC segment is quickly catching up. Apples illusions of grandeur in its walled off garden have people seeking to peek what the world is like outside that garden. More and more people are looking to break that wall down and experience life outside of it.
This. Making machine far more prone to shortened life spans and less capable isn’t going to work out well when these machines enter their 4 and 5th years of life.
Apples biggest competitor is their own products. The used market. We have 2015-2016 MacBook pros still churning like they were new. And I don’t think apple like that. They want you in a new machines the moment your new device leaves the safety of apple care. They don’t want to keep the parts setting around. And they 1000000% don’t want third party repair services extending the life of their machines. Which is ever apparent by friggen flex cables being serialized. That’s the great thing about the intel Mac’s. Parts were interchangeable. And now when apple makes devices obsolete “ vintage “ there will be piles of these devices everywhere creating E waste with zero ability to be fixed. And all of this is by design.
Apple has changed a lot as a company and I’m glad I’ve refused to change with it. A company I once venerated is a shell of its former self. Full of buzz words and flashy marketing. I have owned one apple silicon machine. A first generation Studio Mac. That’s been used for basic productivity. That’s been back to apple three times for not wanting to boot.
After the third round I sold it.
Never again.
I do not enjoy owning the John Deere of computers.
Judging by your response I don’t think we are on the same page, but for this reason:
The M2 MBP (13” touch bar that is) which shares the same M2 chip as the MBA, absolutely does not output to more external displays than the MBA.
It takes the 14” 16” with M1 Pro or M2 Pro to exceed that limit. Judging by similar limits on mini, which is headless thus a total of two, and the iMac which also has an internal screen where it can do 1 just like the MacBooks, we can conclude the total display buffer count is exactly 2 for these M1 M2 chips, with no exceptions (on Macs).
At no point did anyone mention the 13” MacBook Pro. It was specifically omitted.
Apples charging more for less.
That’s the argument.
Dude you said this:
”The M2 chips in the MacBook Pro is the same M2 chip in the MacBook Air M2 model. Same die, same size.
The Pro machines HAVE the capability of multiple displays.”
If not the 13” touch bar MBP that you were referring to, then what was it? Seriously I am at a lost.
Regardless I do agree that *all* MacBooks should really support better external display capabilities. If that’s what matters in the context of this discussion
I don’t think they care. Especially considering that half of the Macs sold in any year (for awhile now) have gone to folks who’ve never owned any Mac. For each model, they focus on “who’s going to actually buy” and included in that group is “people who don’t care enough about upgradability for it to alter a purchase decision”, which is a very large number of the potential market.
Fair, but my comment was about the rumored M3 Max beginning to prioritize performance over power efficiency. The M2/M1 Max and Pro have the same CPU configuration so unless you are utilizing the GPU the peformance & power draw is very similar.
I think the M3 Max (assuming the rumors are correct) will still be a great portable system but there may be a bigger difference between the Max and Pro for battery life.
This is highly unlikely in the general case.
What I mean by that is, for a given workload, the M3 Max will not be less efficient than the M3 Pro (or M2 Max, for that matter). If the M3 Max can run at higher clocks than the Pro, it will still be able to run at lower clocks as well. And if it does indeed have more cores than the Pro, it will be able to deep sleep or totally shut down cores that aren't in use.
If you have a workload that crushes an M3 Pro, then the M3 Max may draw more power to handle it. But it will still be at least as efficient (possibly more efficient if it benefits from race-to-idle). The only exception to this will be whatever "high power mode" it may support, but that will be at the user's option (and probably only when plugged in).
Thus for normal workloads, you'll see the M3 Max battery life be superior to the M2 Max, wh for wh, and likely no more worse than the M3 Pro's than the M2 Max is worse than the M2 Pro (which, IIRC, is not too much, and likely attributable mostly to other hardware like the screen). Of course, if you run a power virus, you'll be able to drain the battery faster on the Max than the Pro, because you'll have more cores available to suck the power down.
...or, wait, no, not even then. If you have only N watts to spread among your cores (which is true - you're limited by cooling), for any large workload that isn't bound by single-threaded performance, you will probably do better with more cores than with fewer. That is, if you can run either 8 P cores or 12, you're always going to do better with 12 (assuming you can fill all 12 with work). That's because if you divide your N watts among 12 cores instead of 8, you'll have to give each of them 33% less power, but (since power consumption correlates with the *square* of the clockspeed) the clock on each core will go down by a lot less than 33%. That will get you a net performance gain for the same amount of power used.
If you think about it, this is just another iteration of the same argument that Apple followed in building their cores all along: wide > high clocks.
Last edited:
Apple Silicon requires one display controller per external display, and the M1 and M2 only have two display controllers. There's no "it was designed for it then taken away" story here.
I don't know if it's possible, but even if in clamshell mode you can get two displays from the regular M processors, that would be great.
I believe that isn't possible for mechanical reasons, i.e. one of the controllers is hardwired to the internal display.