WWDC 2024

[Jimmyjames]

Apparently this has been in development for a while. macOS 13 and later has been using this for external drives, but the library was limited to a private entitlement until now. Interesting.

Interesting. I didn’t know that. I knew it was used on iOS but was unaware of it’s use on macOS.

Huh. This is more surprising because we aren’t exactly seeing a proliferation of codecs these days. And hardware acceleration is becoming increasingly important for battery life.

That’s true. I’m more interested in format support like mkv rather than codecs, which definitely benefit from hardware acceleration, as you said.

There is an option to install them, but I haven’t used the beta long enough to see them in action.

Ok. No problem.

 

[Jimmyjames]

I wonder … supposedly there’s a private, unstable API for user space graphics drivers too. Maybe one day we’ll see that surface as well?

Wow. I do wonder with these user space APIs as replacement for kexts, whether performance can be good enough. It would be great for security if they can make as performant.

I wonder if these moves relate to my theory here:

Exclaves, XNU and the future.

Continuing my series of very niche interests, I was recently considering if anything significant has changed with regard to XNU. For anyone who doesn’t know, XNU is the name of the kernel in pretty much all of Apple’s products. It’s comprised of the Mach microkernel and quite a bit of...

techboards.net

 

[NotEntirelyConfused]

MediaExtension: QuickTime plugins are back baby!

MediaExtension | Apple Developer Documentation

This framework provides a means for developers to create format readers and video decoders for media that the system doesn’t natively support.

developer.apple.com

Huh. I wonder if Apple Music (nee iTunes, not the service) supports this, and if so will it finally be able to support some codecs Apple has refused so far?

Of course foreign codecs often come with foreign formats. So unless they have plugins for that too, FLAC still won't work. :-(

 

[Nycturne]

Wow. I do wonder with these user space APIs as replacement for kexts, whether performance can be good enough. It would be great for security if they can make as performant.

I wonder if these moves relate to my theory here:

Exclaves, XNU and the future.

Continuing my series of very niche interests, I was recently considering if anything significant has changed with regard to XNU. For anyone who doesn’t know, XNU is the name of the kernel in pretty much all of Apple’s products. It’s comprised of the Mach microkernel and quite a bit of...

techboards.net

I don’t think this provides evidence of anything specific. Precisely because this is also what securing XNU so it can continue to be useful in the future looks like. As I said in that other thread, userspace drivers are not new, and pretty much everyone (Apple, Microsoft, Linux) is pushing at least something into userspace to make the kernel more secure and reliable and has been for a while. In this space, Apple is playing a bit of catch up in some ways (FUSE dates back to the 00s), but they went all in to commit that everything will be userspace at some point first.

Performance depends a lot on how userspace drivers affect the quantity of context switches. In an ideal world, if I can minimize context switches (must I use the kernel to route requests to the driver?), things can be quite performant. There’s nothing inherent about the kernel that makes it faster (other than maybe scheduler priority), but rather it’s crossing boundaries between the driver and kernel to interact with the hardware and clients that causes problems.

Huh. I wonder if Apple Music (nee iTunes, not the service) supports this, and if so will it finally be able to support some codecs Apple has refused so far?

Of course foreign codecs often come with foreign formats. So unless they have plugins for that too, FLAC still won't work. :-(

In AVFoundation speak, these are assets (container) and tracks. It looks like you can create your own asset readers to handle MKV and FLAC containers, in addition to track readers for the media itself.

In terms of what apps pick up support, this looks like it all plugs into AVFoundation, so apps built on top should be able to work with it. What does iTunes use? No idea, it’s a legacy app so it’s hard to say just how it’s architected underneath and what has or hasn’t been updated over the years. Keep in mind it started out as a Classic MacOS app.

 

[Jimmyjames]

In AVFoundation speak, these are assets (container) and tracks. It looks like you can create your own asset readers to handle MKV and FLAC containers, in addition to track readers for the media itself.

In terms of what apps pick up support, this looks like it all plugs into AVFoundation, so apps built on top should be able to work with it. What does iTunes use? No idea, it’s a legacy app so it’s hard to say just how it’s architected underneath and what has or hasn’t been updated over the years. Keep in mind it started out as a Classic MacOS app.

macOS supports Flac In terms of QuickTime Player, quicklook etc.iTunes/Apple Music does not unfortunately. I doubt this will change that.

 

[NotEntirelyConfused]

macOS supports Flac In terms of QuickTime Player, quicklook etc.iTunes/Apple Music does not unfortunately. I doubt this will change that.

Hah. True, and I should have remembered that. :-( Oh well.

Also supporting this sad prediction, iTunes used to support media plugins, and they removed that capability. So it's likely that they do see this as strategic. Grr.

It's really too bad. Almost like Apple Mail (though not as bad as that cesspool of bugs), iTunes/Music has been going downhill for years. In the last few years they've chased away all professional or serious amateur DJs because they broke their player, and it now randomly skips half a second of music near the end of the track often enough to piss off all the dancers.

 

[Jimmyjames]

Hah. True, and I should have remembered that. :-( Oh well.

Also supporting this sad prediction, iTunes used to support media plugins, and they removed that capability. So it's likely that they do see this as strategic. Grr.

It's really too bad. Almost like Apple Mail (though not as bad as that cesspool of bugs), iTunes/Music has been going downhill for years. In the last few years they've chased away all professional or serious amateur DJs because they broke their player, and it now randomly skips half a second of music near the end of the track often enough to piss off all the dancers.

What do they use now instead of iTunes?

 

[Altaic]

Huh. This is more surprising because we aren’t exactly seeing a proliferation of codecs these days. And hardware acceleration is becoming increasingly important for battery life.

We are actually seeing new codecs in machine learning. The so-called neural codecs are extremely space efficient with “good enough” quality. They’re mostly used for speech and have a lot of potential for multimodal models, though I wouldn’t be surprised if they soon extend into general audio, image, and video. Perhaps MediaExtension is part of Apple’s ML-platform push, but, of course, neural codecs also have uses outside of LLMs and such.

 

[NotEntirelyConfused]

What do they use now instead of iTunes?

I can't begin to speak for DJs - I only know my dance community, and there are dozens of different styles with large numbers of adherents. But within my community, most DJs have moved to using a player called "embrace" while continuing to organize within iTunes/Music.

It's a really pathetic situation. And even more appalling because for so long Apple was hugely better than windows for sound work because of OS issues - I no longer recall the details but I believe it had something to do with unpredictable high latency on DPCs in Windows. And now Apple can't even get music playback right! They're just sleepwalking out of another market. :-(

 

[Jimmyjames]

In the “Port advanced games to Apple platforms” session there a couple of interesting tidbits.

First. The addition of a reactive mask to MetalFX upscaling is supposed to increase fidelity is fast moving scenes. To quote directly:
"This optional feature may help you achieve higher upscaling fidelity in scenes that include fast-moving objects with inaccurate motion information, which may happen when using alpha blending."

Secondly, ray tracing. To quote again.
"Metal also adds other improvements this year to help bring ray tracing from other platforms. You can now specify the transformation matrices in your acceleration structures in row-major order. This may save you from paying the cost of transposing them at runtime. And direct access to on-chip intersection result storage improves performance by avoiding data copies and potential GPU memory spilling of passing arguments between ray tracing shader functions."

Nothing huge but nice nonetheless.

 

[dada_dave]

Courtesy of @testato at MacRumors an interesting difference in AI scores between iOS 17 and 18:

iPhone 15 Pro vs iPhone 15 Pro - Geekbench

Will be even more interesting if confirmed multiple times

 

[Altaic]

Courtesy of @testato at MacRumors an interesting difference in AI scores between iOS 17 and 18:

iPhone 15 Pro vs iPhone 15 Pro - Geekbench

Will be even more interesting if confirmed multiple times

Those text classification benchmarks are looking pretty excellent. However, I’m a bit put off by the INT8 results in general, since < 8 bit quantization is pretty common right now. Clearly that wasn’t a focus this time around, but I expect there’s low hanging fruit in the pipeline.

 

[quarkysg]

Those text classification benchmarks are looking pretty excellent. However, I’m a bit put off by the INT8 results in general, since < 8 bit quantization is pretty common right now. Clearly that wasn’t a focus this time around, but I expect there’s low hanging fruit in the pipeline.

Seeing that it is beta software, it could be due to lots of debug codes slowing down processing.

 

[Altaic]

Seeing that it is beta software, it could be due to lots of debug codes slowing down processing.

Even beyond that, basically everything in ML is a moving target for the foreseeable future. People use quantization for lower bit parameters to ease memory throughput bottlenecks right now, but ideally one should be able to force more into 16-bit parameters in smaller models via distillation and other methods; perhaps that’s what Apple is focusing on. Furthermore, evaluation and benchmarking are (and will continue to be for a long time) very active research areas, and it’s incredibly improbable that geekbench is actually useful

 

[theorist9]

The rumors continue to be that while the MBP's will get another update in late 2024*, we won't be seeing updated Studios until 2025. So two updates to the MBP's (last year to the M3, and this year to the M4) before we get a single update to the Studio. It's disappointing Apple won't give their Studio a bit more love.

I understand that updating the Ultra chip likely creates additional challenges, and it thus makes sense the Ultra isn't updated as often. But if that's the barrier, I'd like to see them put the MBP's, Mini, and Max Studio on the same release schedule, leaving the Ultra Studio and MP on a slower timeline.

Given that new Max chips are avaiable, any decision to not release a new Max Studio is obviously based on business considerations rather than technological constraints. But I'm sure Apple has reasons for what it does. Are the Max Studio's sales so low that the cost of putting a new chip in an existing case is too much to recoup if done annually?

*Ross Young now agrees with Gurman that new MBP's will be relased in Q4 2024:

M4 MacBook Pro on track for a late 2024 debut, analyst says - 9to5Mac

Apple began the transition to M4 chips last month with the new iPad Pro lineup. We’ve already heard a handful...

9to5mac.com

 

[Cmaier]

The rumors continue to be that while the MBP's will get another update in late 2024*, we won't be seeing updated Studios until 2025. So two updates to the MBP's (last year to the M3, and this year to the M4) before we get a single update to the Studio. It's disappointing Apple won't give their Studio a bit more love.

I understand that updating the Ultra chip likely creates additional challenges, and it thus makes sense the Ultra isn't updated as often. But if that's the barrier, I'd like to see them put the MBP's, Mini, and Max Studio on the same release schedule, leaving the Ultra Studio and MP on a slower timeline.

Given that new Max chips are avaiable, any decision to not release a new Max Studio is obviously based on business considerations rather than technological constraints. But I'm sure Apple has reasons for what it does.

*Ross Young now agrees with Gurman that new MBP's will be relased in Q4 2024:

M4 MacBook Pro on track for a late 2024 debut, analyst says - 9to5Mac

Apple began the transition to M4 chips last month with the new iPad Pro lineup. We’ve already heard a handful...

9to5mac.com

it’s always going to be the case that apple focuses on products that make the most revenue. Macs already are low on that list, but studios are low even compared to other macs. The only reason we get new ones is probably because the engineers, themselves, insist on it.

I wonder if the new MBP’s will be OLED. I assume so? They would get some of the rumored thickness reduction from that. Still unlikely to replace my M1 MBP this go around. Maybe M5.

 

[Jimmyjames]

The rumors continue to be that while the MBP's will get another update in late 2024*, we won't be seeing updated Studios until 2025. So two updates to the MBP's (last year to the M3, and this year to the M4) before we get a single update to the Studio. It's disappointing Apple won't give their Studio a bit more love.

I understand that updating the Ultra chip likely creates additional challenges, and it thus makes sense the Ultra isn't updated as often. But if that's the barrier, I'd like to see them put the MBP's, Mini, and Max Studio on the same release schedule, leaving the Ultra Studio and MP on a slower timeline.

Given that new Max chips are avaiable, any decision to not release a new Max Studio is obviously based on business considerations rather than technological constraints. But I'm sure Apple has reasons for what it does. Are the Max Studio's sales so low that the cost of putting a new chip in an existing case is too much to recoup if done annually?

*Ross Young now agrees with Gurman that new MBP's will be relased in Q4 2024:

M4 MacBook Pro on track for a late 2024 debut, analyst says - 9to5Mac

Apple began the transition to M4 chips last month with the new iPad Pro lineup. We’ve already heard a handful...

9to5mac.com

If true, really disappointing that the Studio and desktops get so little attention. Doubly so if you are a desktop user as I am.

A while ago, Marco Arment said something I think has a lot of truth to it: he said “don’t care about one of Apple’s products more than they do.” He was referring to the HomePod, but I feel this is applicable to the desktops. Going forward I may have to consider a MacBook Pro as much as I dislike laptops.

 

[theorist9]

it’s always going to be the case that apple focuses on products that make the most revenue. Macs already are low on that list, but studios are low even compared to other macs. The only reason we get new ones is probably because the engineers, themselves, insist on it.

I hear you, but the Mac division's revenue only seems small when compared to Apple as a whole. As stand-alone company, it would be huge. Its average annual revenue over the past two years is $35B, which would put it ≈120 on the Fortune 500 list

So the idea that a division that big, and which sells only about dozen different products (not including accessories), doesn't have resources to update each of them annually (using existing chips and cases) doesn't quite make sense to me.

I wonder if the new MBP’s will be OLED. I assume so? They would get some of the rumored thickness reduction from that. Still unlikely to replace my M1 MBP this go around. Maybe M5.

Alas, any article reporting Young's predictions really should have mentioned whether he said the panels would be OLED or not, and the one I linked didn't. Unfortunately, Young's original tweet about this is behind a paywall. However, I did find this 6/18/2024 article from notebookcheck, which said Young doesn't expect to see OLED screens on the MBP until 2026:

"Ross Young, as well as other reliable tipsters, have repeatedly mentioned that Apple will inevitably switch to OLED displays for their MacBook Pro lineup, starting as soon as 2026. When asked by a fellow Twitter user why the switch hasn't been made already, Ross stated that Apple has been waiting for their tandem OLED tech to mature..."

OLED displays will allow for thinner MacBook Pros, claims reliable analyst

Recently, Mark Gurman claimed that Apple's recently launched iPad Pro marked the company's return to pursuing slimmer aesthetics, with iPhones and MacBooks next in line. Thanks to Apple Silicon's efficiency, the company can now achieve thinner designs without thwarting usability. Now, analyst...

www.notebookcheck.net

 

[B01L]

M4 Pro would be the "Brava" code-named SoC, M4 Max would be two of these "Brava" SoCs stitched together...

M4 Ultra & M4 Extreme would be, respectively, two and four of the "Hidra" code-named SoCs stitched together, so maybe the "M4 Max" variant for the Mac Studio desktop is actually a single "Hidra" SoC; and these "Hidra" SoCs are being stockpiled for not only the Mac Studio/Mac Pro/Mac Pro Cube desktops, but also for the Apple Intelligence servers...?