Somehow impressive that they’re able to ship Universal Control (think Synergy only built-in) without M1 or U1. I wonder how do devices know where they are positioned relative to each other in physical space.
The device does not know where the other device is exactly although it knows it is clode because of Bluetooth. It will assume you want to move to the device you interacted with most recently when you move your mouse pointer past the edge of your Mac. From that moment on it will assume the device is on that side.
U1 doesn’t directly know direction between iPhone and airtags either. You have to move your phone around for it to deduce the direction. At least, that’s how it works right now.
Apple’s forced deprecation back at it again. They claim to care about the environment and for reasons other than technical decide not to support Intel devices (even though they still sell them TODAY!)
The day they stop selling Intel Macs, there will be an uproar from people who claimed that X months was not enough time to refresh their workplace's MacBooks--which need to be Intel because in spite of Rosetta 2 and the speed at which libraries and apps are being recompiled there's still some obscure thing that doesn't work yet on M1. They're kinda damned if they do, damned if they don't.
But enough of that, let's look at whether they decided this "for reasons other than technical". A lot of the features (the FaceTime image processing, the dictation and speech stuff, etc.) seem like the kind of thing that was built against their Neural Engine. Intel CPUs don't have custom circuitry for running neural nets, let alone an API that is congruent with Apple's. It makes sense that computers with certain advanced hardware are going to get features that other computers without equivalent hardware can't support.
Is it possible to use virtualization and x86 emulation at the same time now? I’d gotten the impression that people didn’t think that was ever going to happen, and it seems like a pretty non-obscure reason to want Intel support to continue.
They're only not supporting new features which have specific requirements of the device — the features it had when you bought it are still supported. The listed hardware support goes back to 2015 and historically they've continued shipping security and stability updates for the previous OS release, so whichever of those old devices are still running won't leave security support until the release of whatever comes _after_ Monterrey (presumably in the fall of 2022) when Big Sur falls out of support.
It’s not unusual for features of graphics APIs to be limited by the hardware rather than software; especially Metal which is a very low-level API that more-or-less directly exposes the capabilities of the GPU. Since Apple switched to an in-house GPU architecture for the M1, it’s not surprising that there are some features Apple’s GPU supports that AMD/Nvidia cards don’t. For instance, Apple’s tile-based-deferred-rendering pipeline allows fragment shaders to easily read back the previously color of a pixel in order to implement custom blending effects — an operation that would be prohibitively expensive on other GPUs.
My point is, everyone else does support older hardware, usually by disabling newest nicest effects (like ray tracing stuff) or switching to fallback version of said affects.
Not caring to do such thing is, probably, intentional push for customers to update their hardware
Oh yeah, Apple are relentless in their push for new platforms and abandoning old stuff. Maybe too much in the Metal sphere of things where there have been some big changes - Metal 1, Metal 2 their "modern" platform they we're pushing for starting at A12 and really now from A14-M1 and the future.
"Perform as intended" would imply an intention, which is what GP is talking about. Low fidelity is a reality of rendering on older devices. Apple should intend to render on them.
Certain features will obviously not work across devices but that is an artifact of the big change in device architecture.
The mentality that you're suggesting is akin to saying (for example) that desktops won't be supported on any display smaller than a given Retina resolution. That would be anti-consumer and totally unacceptable.
Hypothetically, the suggestion from Apple that they would stop supporting features on laptops that are currently for sale because they are "fallback" devices would be the height of arrogance and would piss off a lot of smart people who have each spent thousands and thousands on Apple devices over the past decade.
> everyone else does support older hardware, usually by disabling newest nicest effects (like ray tracing stuff)
Isn’t that effectively what they’re doing here? They’re not taking away anything that people are using now - just saying that, for example, you aren’t getting the new ability to copy text out of images which is a cool feature but not a game changer or something anyone has existing habits around.
It seems far less of a gripe than, say, when they required GPUs capable of running a desktop compositor or dropped 32-bit support.
Sure? It suggests they have a new TTS engine which they used for the new languages, but they kept the old engine around for Intel macs rather than dropping existing functionality.
Feels like the kind of thing where maybe there was an architectural initiative to move from a cumbersome speech data format to something more elegant that required the Neural Engine. Since they already had the old languages in the old format, they kept that data around, but for the new languages opted to only ingest the easier-to-work-with format.
But hey, I’m guessing and you’re Saurik: I have a feeling you know more about this than me and may have considered and dismissed this possibility for a good reason.
Nobody knows for sure but the rumours say it's been delayed due to a hold up in mini-LED displays. Probably will be something by September. But who knows?
> Text-to-speech in more languages, including Swedish, Danish, Norwegian, and Finnish
> On-device keyboard dictation that performs all processing completely offline
> Unlimited keyboard dictation (previously limited to 60 seconds per instance)
>> It’s because Intel Macs don’t have the neural engine these features require.
This is of course utter horseshit. Text-to-speech might be slightly easier with a "neural engine" but the fact that it can be done just fine in other languages without a "neural engine" belies this statement.
Dragon Systems (among others) was doing completely offline speech-to-text dictation on Macs that were 1000x less powerful than today's Macs 20 years ago, and doing it well. If Apple doesn't know how to do it today, that says a lot more about the competence of Apple's programmers than about their hardware.
Yes they had text to speech, but not as good as today, not nearly as robust, you had to train it. Google do also have on device speech transcription, for their own phones with neural engines. Their blog post points out that it's quite a resource intensive method of doing things.
> Dragon Systems (among others) was doing completely offline speech-to-text dictation on Macs that were 1000x less powerful than today's Macs 20 years ago, and doing it well.
Generally speaking these features maybe are faster on M1 but I don't fully buy into the claim that this is a technical requirement. Seems more like a technical decision to me.
As a user it would be vastly preferable for Apple to deploy these features to both cpu architectures, and sure M1 would offer acceleration and better performance.
Considering that Apple hasn't finished fully migrating their hardware lineup to M1 (and are still selling intel macs), this deprecation feels a bit rushed.
33 comments
[ 2.3 ms ] story [ 94.4 ms ] threadSource: https://www.theverge.com/2021/6/8/22523613/macos-monterey-ww...
But enough of that, let's look at whether they decided this "for reasons other than technical". A lot of the features (the FaceTime image processing, the dictation and speech stuff, etc.) seem like the kind of thing that was built against their Neural Engine. Intel CPUs don't have custom circuitry for running neural nets, let alone an API that is congruent with Apple's. It makes sense that computers with certain advanced hardware are going to get features that other computers without equivalent hardware can't support.
They're only not supporting new features which have specific requirements of the device — the features it had when you bought it are still supported. The listed hardware support goes back to 2015 and historically they've continued shipping security and stability updates for the previous OS release, so whichever of those old devices are still running won't leave security support until the release of whatever comes _after_ Monterrey (presumably in the fall of 2022) when Big Sur falls out of support.
This is not true. The reasons are technical.
>Live Text for copying and pasting, looking up, or translating text within photos
>An interactive 3D globe of Earth in the Maps app
>More detailed maps in cities like San Francisco, Los Angeles, New York, and London in the Maps app
>Text-to-speech in more languages, including Swedish, Danish, Norwegian, and Finnish
>On-device keyboard dictation that performs all processing completely offline
>Unlimited keyboard dictation (previously limited to 60 seconds per instance)
Most of them make sense; the M1 has the Neural Engine and ISP, and Intel Macs don't.
Not sure what's stopping them from enabling additional text-to-speech languages or rendering a globe, though.
Text to speech uses Neural Engine as well. Rendering Globe likely uses newer version of Metal only supported on M1.
Apple publishes a full compatibility matrix showing which features are supported on each GPU hardware family: https://developer.apple.com/metal/Metal-Feature-Set-Tables.p...
Not caring to do such thing is, probably, intentional push for customers to update their hardware
Not really. Apple has always had a policy of only supporting features that perform as intended.
Low fidelity “fallback” versions are simply not part of Apple’s approach to design.
Certain features will obviously not work across devices but that is an artifact of the big change in device architecture.
The mentality that you're suggesting is akin to saying (for example) that desktops won't be supported on any display smaller than a given Retina resolution. That would be anti-consumer and totally unacceptable.
Hypothetically, the suggestion from Apple that they would stop supporting features on laptops that are currently for sale because they are "fallback" devices would be the height of arrogance and would piss off a lot of smart people who have each spent thousands and thousands on Apple devices over the past decade.
There is no reason Apple should make things that don’t work as designed.
> saying (for example) that desktops won't be supported on any display smaller than a given Retina resolution.
Yes, this is an absurd exaggeration, because in fact we’re talking about a few new features that are not central to the user experience.
> That would be anti-consumer and totally unacceptable.
It would, which is why they don’t do it, and why it’s not representative of anything.
It’s unclear what point you are trying to make other than that they could make bad design decisions if they wanted to, but in fact don’t.
Isn’t that effectively what they’re doing here? They’re not taking away anything that people are using now - just saying that, for example, you aren’t getting the new ability to copy text out of images which is a cool feature but not a game changer or something anyone has existing habits around.
It seems far less of a gripe than, say, when they required GPUs capable of running a desktop compositor or dropped 32-bit support.
But hey, I’m guessing and you’re Saurik: I have a feeling you know more about this than me and may have considered and dismissed this possibility for a good reason.
> On-device keyboard dictation that performs all processing completely offline
> Unlimited keyboard dictation (previously limited to 60 seconds per instance)
>> It’s because Intel Macs don’t have the neural engine these features require.
This is of course utter horseshit. Text-to-speech might be slightly easier with a "neural engine" but the fact that it can be done just fine in other languages without a "neural engine" belies this statement.
Dragon Systems (among others) was doing completely offline speech-to-text dictation on Macs that were 1000x less powerful than today's Macs 20 years ago, and doing it well. If Apple doesn't know how to do it today, that says a lot more about the competence of Apple's programmers than about their hardware.
https://ai.googleblog.com/2019/03/an-all-neural-on-device-sp...
Not even close to modern standards.
As a user it would be vastly preferable for Apple to deploy these features to both cpu architectures, and sure M1 would offer acceleration and better performance.
Considering that Apple hasn't finished fully migrating their hardware lineup to M1 (and are still selling intel macs), this deprecation feels a bit rushed.
It’s a technical requirement if the designers specified a required performance level.
Jailbreakers found a way to enable Siri on it, and it honestly felt exactly the same as the 4S.
F*ck me, right? :(