Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

KensaiMage

macrumors regular
Original poster
May 25, 2017
235
65
Can we expect a fanless macbook with low power intel processor + A11x chip?

It would be my dream, but I bet it will be just T2 or T3.
 
I like the "Amiga" idea which I saw in another thread, where A11x would take certain tasks off the Intel CPU, while keeping general Intel architecture to avoid rewriting code – I just don't know which ones – sound? graphics? something else?
 
I would like to see a hybrid device from Apple similar to the Dell XPS 13 2in1.

In tablet mode it would run iOS otherwise macOS with some clever data exchange.

Maybe when IPad sales will further drop. But probably not the next couple of years
 
I like the "Amiga" idea which I saw in another thread, where A11x would take certain tasks off the Intel CPU, while keeping general Intel architecture to avoid rewriting code – I just don't know which ones – sound? graphics? something else?


Apps would still need to be rewritten so they'd be able to know what instructions to send off to the chip though - This is the best concept if we're going to have an ARM chip in a Mac, but it's not just plug-and-play.
[doublepost=1517594195][/doublepost]
In tablet mode it would run iOS otherwise macOS.


So depending on what mode it is in, half of your data on the device is inaccessible?....
 
So depending on what mode it is in, half of your data on the device is inaccessible?....

Well better not:) they would have to develop some exchange. But for pics, music and office files it would work already know.
 
There are reports out there that Apple plans on letting iOS apps run on the Mac. If that's true, and forgive my ignorance, but wouldn't MacOS just hand off the processing for these apps to the ARM processor, essentially running the apps natively? For Macs without ARM, emulation would be done. Meaning, no need to rewrite apps. Apple did this for PowerPC apps to run on the first Intel Macs as a temporary bridge between platforms.
 
There are reports out there that Apple plans on letting iOS apps run on the Mac. If that's true, and forgive my ignorance, but wouldn't MacOS just hand off the processing for these apps to the ARM processor, essentially running the apps natively? For Macs without ARM, emulation would be done. Meaning, no need to rewrite apps. Apple did this for PowerPC apps to run on the first Intel Macs as a temporary bridge between platforms.


Yes and yes. That would work. But that limits the use of the ARM chip to apps written for iOS. (The emulation already exists - check out an app called Simulator bundled with Xcode).

Also, the rumour mostly point to a super-API that works on both macOS and iOS, not so much towards Macs running unmodified iOS apps. That would also lead to quite a bad UX really.
 
Apps would still need to be rewritten so they'd be able to know what instructions to send off to the chip though - This is the best concept if we're going to have an ARM chip in a Mac, but it's not just plug-and-play.
Aye – I was thinking of "partial Rosetta" where the system intercepts a call, then either sends it to Intel CPU, or "translates" it to ARM call. (You can tell I don't exactly have proper technical vocabulary for this.)

I'm mostly just curious what sort of improvements could be achieved by doing that.
 
Do we know if the T2 on the iMac Pro performs any sort of neural offloading in the fashion that the A11's coprocessor does? Hypothetically, the use of a coprocessor in this way on something like the MacBook could potentially benefit energy efficiency substantially, especially as AI/ML tasks become more commonplace, wouldn't it? (forgive me if my terminology is off...I'm still learning about this area.)
 
Aye – I was thinking of "partial Rosetta" where the system intercepts a call, then either sends it to Intel CPU, or "translates" it to ARM call. (You can tell I don't exactly have proper technical vocabulary for this.)

A system like that would kill performance and efficiency, eliminating the whole point. Plus there are licensing issues with respect to the x86 and x86_64 instruction set. And Apple bought Rosetta back in the day. Rosetta is also much simpler than a software scheduler that would be able to distinguish between two entirely different architectures, and that would be able to receive instructions from both architectures. What architecture would this middle layer run on, and how would it then be able to intercept instructions from the other architecture? As far as I can think of, it would have to run simultaneously on both, exchanging data for each operation, to see where it would be best suited to run, which would mean you could never power-gate either processor, making it a total power hog. – At least that's the only way of implementing it I can think of. Aside from having developers manually specific operations for an ARM chip and recompiling the code with these new APIs.

Do we know if the T2 on the iMac Pro performs any sort of neural offloading in the fashion that the A11's coprocessor does? Hypothetically, the use of a coprocessor in this way on something like the MacBook could potentially benefit energy efficiency substantially, especially as AI/ML tasks become more commonplace, wouldn't it? (forgive me if my terminology is off...I'm still learning about this area.)

As far as I know, the T2 only control the following

All SMC tasks including fan control
SSD controller duties as well as RAIDing the two SSDs in the iMac Pro
ISP/DSP duties for the iSight camera (DSP/ISP = Digital/Image signal processing - this theoretically is part of what should make the iSight camera in the Pro better than the regular iMac)
Security operations with the secure enclave. (AES, RSA, etc)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.