1,474 comments

[ 1.4 ms ] story [ 522 ms ] thread
I was interested by the GPU that was mentioned in the slide, does this mean that Apple is thinking that they have the expertise to take on Nvidia and AMD in that space?
Whether or not they think that or they can, they have been developing GPUs for iOS devices. It'll be interesting to see what they do for their desktops.
I agree. I know they're currently using AMD GPUs, but how much longer will they keep using AMD if they believe that their own in house chips can do a better job? Also, I'm curious as to whether they would try to sell their GPUs (iGPU?) to gamers who traditionally don't stay within the Mac ecosystem at all. Having a 3rd major player in the GPU space would be wild.
> Having a 3rd major player in the GPU space would be wild.

Reportedly 4th as Intel is supposedly trying once again for that market.

Apple uses SOCs, the GPUs are on the same chip as the processor.
It helps explain why they haven’t added nvidia support. Apple is going the opposite direction and getting rid of both.
The advantage Apple has is that Nvidia and AMD and Intel have to actually make money from their chips. Apple can break even or make a loss on the chips themselves.

So maybe, they certainly have the budget to try.

I don't expect Apple to sell the hardware for cost. They are quite adept at selling products with a hefty profit margin.
I don't mean in pricing, but design. Other companies have to design chips that make sense in the market. Apple makes it for themselves, and they can make chips that would not make financial sense, but perform great.

And charge you a bunch for the whole package, of course.

I think it's their modified PowerVR cores they've been using.

Probably only focused on the iGPU segment.

Nearly the entire demo set was about GPU performance by demoing graphic responsiveness. Almost a bigger deal than the processor itself it seemed like.
Given the lackluster support Apple has for the AMD GPUs they currently ship I bet it will feel like an upgrade. Plus a lot of the product line uses the integrated Intel graphics so beating that isn't much of a challenge.
edit: why the downvotes? legit curious

I'm curious about the future, especially games, if they want to take on consoles.

PS4 and XboxOne are 7 years old and while the next gen looks really good Apple can refresh the productline much quicker (on the other hand it's a good question if people really would want to buy a new console say every year because Apple sure will be aggressive pushing out new models much more frequently)

macOS's gaming story has always been pretty abysmal. They nuked a ton of games when they dropped 32-bit support, and they refuse to implement the graphics API that the gaming industry is standardizing around. I don't foresee any improvements in Mac gaming from this announcement.
But now games like Fortnite, Minecraft, PUBG etc. that already exist on iOS can natively run on the Mac. So I think that's a pretty big thing
> But now games like Fortnite, Minecraft, PUBG etc. that already exist on iOS can natively run on the Mac. So I think that's a pretty big thing

With touch controls on non-touchscreen devices.

It isn't big.

I think the fact they chose to show off a fairly poorly running version of Tomb Raider as their demo goes to show that they still fundamentally do not understand the gaming market.

They chose Tomb Raider to demo Rosetta 2 precisely because it’s an old game built for x86
Note: the gaming industry is not actually standardizing around Vulkan. It is still DX on Windows, Metal on Mac/iOS, and then Vulkan on Android
am I stupid? how are they running a x86 vm on ARM...
Rosetta. Apple knows processors, ARM is just a licensed instruction set.
To be clear, it's not a virtual machine. Rosetta translates x86 instructions to ARM instructions on the fly.
I wonder if this is actually even related to the original Rosetta (which was actually an external vendor - QuickTransit by Transitive), acquired by IBM. Most of the Transitive team left IBM to go to Arm and Apple.
Are they, or is it an aarch64 build of Linux they showed?
The same way they ran m68k code on PowerPC CPUs and PowerPC code on x86 CPUs.
Emulation? I assume using the same kind of tech that let x86 run old PowerPC apps
It's interesting that their official virtualization support (in the demo) is Parallels.
Curious that they said nothing about virtualizing Windows. That can’t be a mistake or oversight. Odd.
It may be to avoid appearing to be misleading. They'll almost certainly be able to virtualize ARM-based Windows, but that's not what normal users are looking for when they want to virtualize Windows.
Why not? Windows 10 for ARM can run x86 apps.
...very poorly. And only 32-bit apps.
Yes. It appears that Rosetta 2 beats Microsoft's x86 emulation by a long shot. I wonder how they're avoiding Intel patents.
What patents? All Apple is doing is translating x86 instructions to ARM instructions, they’re not emulating the processor.
Keep in mind that current macOS runs ONLY 64-bit x86 apps and interestingly Windows on ARM emulates ONLY 32-bit apps. Since 64-bit extensions were designed by AMD, maybe they have some deal with them?
I wouldn't expect Windows for ARM to run on Apple's custom chips.
I did not know that.

Perhaps it's just to temper expectations then. If the performance or user experience isn't comparable to what today's users expect, it's probably better not to celebrate that use case.

> Why not? Windows 10 for ARM can run x86 apps.

Nobody who has ever attempted to use it would claim this.

Why?
Because to all intents and purposes it doesn't work.
How many people still use Parallels or other virtualization for their Macs?

I've stopped doing that years ago. No need now - if you want to play Windows games, buy a PC - it's a lot cheaper.

I don't.

I used Parallels for a while, and it was great, but then we kind of standardized on VirtualBox at work. It was okay.

But then our dev environments got too big to run on laptops, so now they're all cloud-hosted VMs. We still run Docker (which uses a VM under the covers for Mac) but that uses Apple's Hypervisor framework and isn't really "user-facing" virtualization.

Still alot, I'd reason.

Especially – and funnily enough – in business contexts, where you need to run software which are each only available on either Mac or Windows, but not both.

Do you have any examples? I was running VMWare Fusion until about 2015 for my office laptop for Visio and Project, at which point I felt no need on my next refresh to simply not request it due to cloud-based alternatives.
Yes, the NURBS (surface) modeler "Rhino" is one – or rather was; meanwhile it got a Mac version, but that one still has no feature parity, so you would still see this running in Parallels somewhere, I guess.
Indeed - only talked about Linux, of which there are of course ARM versions available.
Yeah, but what I care about is being able to run Docker with x86 containers so that I can use the same containers on my laptop as my servers.

Not all packages are available in ARM flavors.

Indeed - and I'm hoping they give some more details about that, whether it's ARM only virtualisation, or if their Rosetta 2 also supports x86 virtualisation.
Their dev documentation states that rosetta is only for user mode emulation.
That’s got to be coming, but maybe just wasn’t ready to demo today.
I noticed that too. Might have trademark or legal reasons though. Technically, if they're able to dynamically translate x86 to ARM as they've explicitly stated (with JIT of JS and the JVM as the example) they should be able to dynamically translate x86 VMs regardless of what OS they contain, which would allow x86 Windows to be virtualized with better-than-interpreted performance.

Or at least that's what I hope ;-) I'm relying on running x86 Windows and Linux within Parallels in addition to native MacOS Apps to do my daily work, which involves compiling and testing x86 binaries for these platforms, so whether this virtualization thing actually runs x86 OSes transparently is an absolute make-or-break feature for me to continue my usage of the MacOS platform for work.

(comment deleted)
I don’t think that would perform very well without some hardware support for it as well. Not an expert on this by any stretch but as I understand it modern virtualization is almost always hardware accelerated which I can’t imagine is a viable option if you’re translating the binaries with Rosetta.
I would guess that their Intel emulation doesn't support 32-bit x86, which many, many Windows apps depend on.
I really wonder if this will result in the Mac-ification of iPad. Would love to finally run a MacOS environment on a tablet.
It seems to be going the other way actually. MacOS Big Sur looks like iPadOS with a menu bar tacked on.
It’s going both ways - iPadOS now has drop down menus at the top, etc.
What they did to Finder is just...unacceptable.
I highly recommend Path Finder as a replacement anyway
Hoping there's a way to switch to a more compact view, but as shown in the keynote it looks terrible. What is with the insistence on wasting space in modern operating systems? Were people getting too productive with new, large displays?
yeah all the UI updates this year seem to be on the theme of stitching up the gaps between macOS and iPadOS.

I'm guessing that when the "two year" transition to all-ARM arch on the mac is complete, they'll merge them.

Also interesting that the dev kits use the same chip as the latest iPad Pros (A12Z). I’d love to see the two OSes merge to some extent because I love writing on the tablet but find iPadOS not the best for general dev work.
Oh, 1,000%. Even basic tasks like working on a file, then pasting said file to a cloud storage app and then to an email/slack channel take an order of magnitude longer than it would on a computer. Long way to go before the iPad becomes a true Pro machine, but for some fields, it's basically there. Still, such simple quality of life improves that are mainstay/muscle memory on computers essentially take time and animations and taps to execute.
Already Apple is making it a headache to run non-signed apps on the Mac. I don't think they're going to slow down the "convergence" of iOS and MacOS at this point.
This is a demo so take with huge grain of salt, but the x86 performance looks solid... better than I expected considering it's based on an iPad CPU.
IDK, "look how good our CPU support is by running a several year old GPU limited game" fell a bit flat for me.
Not even running it very well. All the graphics settings were turned down, the framerate looked choppy and it was only running in 1080p.
It’s not a demo of their new gaming-class GPU though.
No - but they weren't exactly showing us performance graphs to see where the game was getting caught on bottlenecks either, however.
So many people don't get this that it's astounding
It was very odd seeing Lara walk through an area with dappled bright light, and her body remain uniformly lit. It may be that the game has a very basic lighting engine though.
It doesn't, they just had the graphics turned down super low.
I was wondering that. I'm hopelessly out of the loop with exposure to modern game graphics, but to me, that looked worse than 360/PS3-era games.
It is like many triple A games in that it has a wide range of settings, all the way from full potato to RTX (it was ironically one of the first games to support that).
Again, people need to dial back their expectations here. You aren't going to see cutting edge games running well through emulation. There is a reason Apple made such a huge emphasis on native apps, native is always going to run much faster.

They didn't demo gaming to suggest this is a great machine for gaming, they demoed it to show that it was possible at all. The previous version of Rosetta during the PowerPC->Intel transition was not known for performance.

If gaming is important to you and you want a Mac then you want an Intel Mac or whatever games are released for Mac ARM. Emulated games are not going to compete with native.

But does it run better than on the current intel mac mini with integrated graphics? All it needs to do is beat intel in comparable circumstances.

It doesn't really matter what the graphics performance is, on high end macs they'll still ship a dedicated GPU from AMD. What matters is that the game is GPU-limited instead of CPU-limited.

The maxed out mac mini cpu is a 6 core 3.2Ghz i7 with turbo boost to 4.6Ghz. I wonder if they can beat that with a newly ARM optimized MacOS? The current i7 has tons of power still as an 8th gen Intel cpu.
> But does it run better than on the current intel mac mini with integrated graphics? All it needs to do is beat intel in comparable circumstances.

Having gone and checked, no. Not even close.

(Nor would it be plausible to expect to. But it's clear Apple have made a choice here, and that is that if you're a user who wants legacy software or desktop gaming, Apple do not care about you compared to their margins. It's that simple.)

It’s not about the game, it’s about running performance-dependent code written for x86 on an ARM chip. A lot harder to fake a stable frame rate in a game than in Photoshop.
I mean, it didn't look like a stable frame rate to me.
It’s mostly OpenGL/metal. Show me a cpu intensive task which isn’t already gpu native/jit
I think your expectations for what emulation is capable of are set a bit high. The fact that it is able to emulate a game that's a few years old at decent frame rate is more than acceptable. You didn't see Microsoft demoing games for their Surface on ARM systems at all and for good reason.
I mean, if I had things my way they wouldn't be switching to ARM at all and emulation wouldn't be necessary, so I don't think it's wrong to be skeptical.

> You didn't see Microsoft demoing games for their Surface on ARM systems at all and for good reason.

Those were also lower-end computers with poor GPUs.

> Those were also lower-end computers with poor GPUs.

You mean the Surface Pro X? It has similar price tiering to the iPad Pro whose SoC was used in these demos.

...ouch, okay, I just looked up the price of Surface Pro X, and it's $1,000. I forgot how much Microsoft was charging for that thing.

Still, the Pro X has poor graphics. I'm going to assume that a dedicated GPU was being used for Tomb Raider—they would have said something otherwise.

> I'm going to assume that a dedicated GPU was being used for Tomb Raider—they would have said something otherwise.

They said exactly what SoC they were using, and it's not known to have spare PCIe lanes lying unused in existing products. Apple pretty much just demoed an x86 game running on an overclocked iPad Pro.

Interesting—are you expecting the Thunderbolt ports in the Developer Transition Kit not to work then? Since it's using a Mac Mini chassis.
It's not like Apple can't change what connectors are available on the back of the mac Mini. The form factor may not have changed, but the ports available have in the past releases.

Don't be surprised if there is no Thunderbolt 3 at all, but just USB-C.

It may be that by the time these things are ready for an actual release they can be USB4, which is USB-C + Thunderbolt technology, but no longer Intel exclusive.
Tech specs on the DTK webpage don’t mention Thunderbolt at all, just USB-C (2 ports) and USB-A (2 ports), plus a HDMI 2.0 port.

Makes me wonder how the Pro Display used in the demos was being driven. DisplayPort with USB-C alternate mode?

It's entirely possible the display was only running at 4K instead of the full, native resolution.
They only said the demos were running off Apple silicon, not that they were running off a Mac Mini DTK machine. They probably have other systems more akin to Mac Pros that they use internally.
They said the demos were running on an A12Z.
> Those were also lower-end computers with poor GPUs.

Were you under the impression this $500 developers kit shipping with an iPad Pro CPU/ GPU is a high end computer? While it's a decent chip, it's essentially the same CPU as the prior generation iPad Pro with 1 additional core.

I can emulate a PS3 on my computer with better graphics at higher framerates, including GPU emulation.

It's not impressive at all.

Shadow of the Tomb Raider is a PS4/XB1 game, not a PS360 game.

---

Edit 2: Please disregard my first edit, below—I was right the first time, then I got the games mixed up.

Edit 1: Oh wait, I forgot, SotTR actually did have an Xbox 360 port! It was one of the last big titles to have one. I think what they showed on screen looked better than the 360 version though, although it's admittedly hard to tell on a stream.

Not on low settings at fake 1080p with an inconsistent framerate, it isn't. It's PS4 level at 60fps on high.

Whereas I can emulate 4k60fps PS3 games with improved textures.

> SotTR actually did have an Xbox 360 port!

It did not, previous game in series (Rise of the Tomb Raider) did have but Shadow(...) was released only on current-gen consoles.

When its running on a iPad pro chip? It's pretty impressive., considering Microsoft x86 emulation on the Surface pro X is abysmal.
Is the target to match the performance of ten year old hardware? Then sure, that's matched. But it's not impressive. AMD FX CPUs have better performance than that, by a mile.
Yeah, but this is their existing CPU/GPU designed to fit into the constraints of the iPad form factor. They'll likely have something much more powerful for consumer hardware.
The iPad CPU/GPU is already thermal limited. An unlimited A12Z is right at the TDP of a laptop chip, at ~25-30W (5W per big core per anandtech, 4 cores, plus GPU and I/O, it's actually quite a generous estimate.

An unlocked A12Z is likely all you can get away with in a laptop, and inferior to SOTA x86 low power CPUs.

If the Tomb Raider game was actually running on a A12Z system (without any external GPU - note that this is the same CPU/GPU on the iPad Pro!), then that demo is actually really impressive, even if the game settings are set to low-quality and the framerate is a bit choppy.
I would have been more impressed with them accessing an game service like Steam and launching random games from there.

Plus they insinuated they were running on the A12 but later demos including the game did not state exactly the hardware being used.

If they are limited to games support Metal then that is jettisoning a large number of games. Granted Catalina already did a of that work for them.

Are they going to support non Apple video cards? The skipped right over that but I suspect they don't think they need to

*edit on that last note it would be a good reason they never brought nvidia chips back as they would know they would not need them

(comment deleted)
On the other hand, it probably doesn't use much CPU just because it's single threaded. No game ever uses more than 10% of my 16-thread CPU. But that also means that emulation could seriously tank single-thread performance and ruin the game.
It might be the other way around. The might have gotten around the big issue with running arbitrary x86 code on ARM (the way weaker memory model) by pinning all x86 threads in a process to a single core. Which would be unfortunate.
Curious how this will effect "the developer experience" considering that cloud instances are also transforming to ARM based processors.
Besides native apps from Apple, Microsoft, and Adobe (Universal 2), the demo included Maya and Shadow of the Tomb Raider (Rosetta 2), and Parallels Desktop (unspecified virtualization improvements). I couldn't tell if the guest O/S was Debian for x84-64 or Aarch64.
I was curious about the hardware they used for the demo.
One of the About This Mac dialogs showed an A12Z with 16 GB, which is consistent with the developer transition kit, basically an iPad Pro in a Mac mini case: A12Z, 16 GB, 512 GB SSD.
It said A12Z. Mac Minis with A12 available next week to devs.
It was stated the demo hardware is the DTK (available this week), which includes an A12Z processor from the latest iPad Pro coupled with 16 GB RAM.
Wow, that's awesome.

I was having intermittent xfiniti issues during the stream.

Tim Cook just said they have new intel macs in the pipeline. Those will go over like a lead balloon.
And I reckon those of us who just bought a shiny new Intel Mac in the past few months might be questioning our decision. Should have held on to that 2010 MBP one more year?
I was thinking the same, but I always avoid gen1 Apple gear anyway, so it’d basically be another 2 years before it’d have been time to upgrade.
Nah, I'd have to wait until early next year to buy an ARM Mac, and mine will be fine for a long time to come.

I am pleased I didn't splurge for the 32GB of RAM, however. I'll do that with my next one.

I was waiting to upgrade my 2013 16gb/512. I really see no point
And without upgradable parts, it really doesn’t make sense
Typing this on a 2010 MBP, bought a new MBP this weekend (was later canceled by the seller b/c of inventory issues) and thinking about waiting another year
Buying a Mac right now, need to replace my EOL Mac Pro. I definitely won't be getting a Pro now. Probably just an iMac with more RAM and SSD instead of Fusion. I really wish I knew when they were updating their iMac and Mac Pro lines.
The last Macs that will support Boot Camp will be the best selling Macs for years.

Much like the 2015 Macbook Pro I expect you'll be able sell it for more than you paid for it years later.

Remember the MacBook Pro with a DVD drive lived on for years after all its siblings were discontinued. We can expect some x86 support to remain available for years after the last new ARM machine is launched.
I think that's the mid-2012 MBP? They were still selling them in 2015, I bought one of the last ones before it was pulled from sale. It remained in the lineup because it continued to be a best-seller.

Apple was already grumpy about servicing it by 2018 or so though, even with AppleCare. (I've switched to Windows now.)

I think it could be tied to large contracts and ensured availability. I suspect there will be an x86-based Mac available well past the two year transition that was announced.
What is really sad is that I don't think the majority of Apple consumers really understand what this means for the Intel macs in the long run. I fear a number of people will buy them not realizing they have a very limited lifespan.
If those devices will have 4-5 years worth of use, then I think most people would be fine with the purchase. 4-5 years already does require upgrading for either performance or quality of life features.
Yes, I think they will support them for at least 5 years.
Your comment made me look how old my current laptop is. "Mid-2014".

It's still as powerful as any of the current Apple laptops - progress has been slow these last 6 years.

Part of the motivation for the transition is the hope that the progress will be much faster over the next 6. If that happens, those intel macs may age out sooner.
(comment deleted)
They won't, necessarily. Fat binaries worked ok the last time around. Sure, Apple could stop making updates available for Intel Macs, but they can stop making updates available for any older model of Mac if they want to.
Their lifespans will be no different than any other Apple hardware, which typically has very long usable lifespans.
(comment deleted)
> […] not realizing they have a very limited lifespan.

It may depend on one's definition of "limited": the Rosetta was first released with Mac OS 10.4 in 2005, [1][2] and was last available in Mac OS 10.6, which first released in 2010, [3] but whose last update was in 2011.

Six years of transitional support is not unreasonable.

[1] https://en.wikipedia.org/wiki/Rosetta_(software)

[2] https://en.wikipedia.org/wiki/Mac_OS_X_Tiger

[3] https://en.wikipedia.org/wiki/Mac_OS_X_Snow_Leopard

Rosetta is for the other case where you want to run old software on a new machine. This is about running new software on an old machine.
Nowadays we can use a same computer for around 8 years easily if we bought with decent spec.
I wouldn't be so sure, there are a fair number of people who want a machine that will run x86 for various reasons. Windows support/ Linux support. Even considering how impressive x86 VMs looked in the demo, lots of people will prefer using intel silicon for guaranteed compatibility.
Linux and Windows have both supported ARM for a while now.
The end goal isn't just using Windows for Windows sake, the reason people use Windows on mac hardware is to get access to apps that run on Windows. And most of those apps still run like garbage (if at all) on the ARM versions of Windows.
They showed Maya (x86 binary) running on their Chip. It has some ability to run intel binaries on Apple silicon through at least two options-emulation and something that sounds like “jit interpretations”-lack of a better word.
I think that is a slightly different use case, though. That demo was an x86 binary running on ARM MacOS via a translation layer. So if there is a MacOS x86 version of the app you want to run, that might be an option.

But I know a lot of people still run Windows because they want the Windows version of an app, either because it isn't available at all on Mac or just because the Windows version runs better (Excel was a classic example of this for a long time, might still be). In that case, I don't know if that same translation layer will have the same performance (if it can run at all outside of MacOS) when running an entirely different OS.

FWIW I remember running PowerPC binaries on Intel macs via Rosetta was pretty painless. They mentioned explicit support for linux/windows emulation so they know it’s an important use case.
They did not mention windows emulation, only Linux/docker.
I don't believe Rosetta will work for linux/windows VMs.
> “jit interpretations”-lack of a better word.

I think this is called "transpiling" -- a version of compiling that's mainly translating from another architecture. And it didn't sound from their description like it was JIT -- it sounded like it would do the transpile when you first installed it (or maybe first ran it?) and keep the results.

Transpiling (as much as I hate that word, because the more you know about compilers, the more meaningless it is), is about source to source translations, not binary translation.

And they have a first pass AOT, with a JIT backup from the sounds of it to support JITs like browsers, node, and java.

Like, a joke version of Windows, yes. As a developer, if my machine can't run Visual Studio then it's not interesting to me at all. I can see it being acceptable to people who work predominantly with tools that have an ARM binary though.
Windows on ARM can run vsCode fine now. Traditional VS is on the way, but you can run x64 Windows apps on ARM.
Visual studio code has an arm build now [0]. I don't know what regular VS is written on but assuming it's .NET framework I assume a build will show up as the framework itself improves on ARM.

[0]: https://github.com/VSCodium/vscodium/releases/tag/1.46.1 (basically just MS source minus branding)

VS is not built on top of dotNET.

VS Code was expected to run on anything, being nothing more than Javascript & HTML wrapped via Electron.

VS Code is not what anyone means by Visual Studio in that context. No one uses Windows for VS Code.

Of course VS Code works on ARM, it's an electron app.

Visual Studio proper is a mix of C/C++, .NET, and COM.
I use Visual Studio, but I could just as easily use JetBrain’s Rider and be just as content.
But you still need ARM specific binaries. If your software doesn’t support your specific flavor of ARM, then you’re out of luck.

(I recently tried to install Zoom on arm64 Debian and it didn’t go so well... Rstudio wasn’t even close to an option)

...and we need new, more hardware for CI/CD that's just dedicated for Apple OSes. Pretty sure we won't be able to use the 5 unused raspberry pi we have.
I found the same problem when I tried running off Pi OS 64-bit for a day—almost every app that _did_ have a Linux binary available was only able to run under x86_64, not on arm64... More here: https://www.jeffgeerling.com/blog/2020/i-replaced-my-macbook...

Zoom, Bluejeans, Dropbox, pretty much all the popular apps I used where I could find a Linux version for my Dell laptop, I couldn't find a way (though got close at least, with Dropbox) to run them on an ARM64 CPU.

Dropbox could run relatively well with qemu user mode.
For Dropbox, you could quite trivially get by getting an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem.
Rclone is what I ended up using and it worked decently well.
Zoom does not release binaries for ARM linux.
No-one runs Linux on Apple laptops these days; it's completely impractical. On OS X, Zoom will obviously recompile their app for the new architecture.
I am – right now even – but I agree it's become impractical.

This announcement is the final nail in the coffin I suppose. I hope some company will come out with great quality, well-designed, minimal laptops. That run Linux perfectly.

If it ran Linux perfectly, it could only do so if the manufacturer maintained its own Linux distribution. Laptop hardware and power management is just too complicated these days for anything else to work well.
All the Linux users I know prefer to use Zoom in the browser.
Unless the new ARM chips can handle VirtualBox well, I’ll pick up the last intel model. I have to work with old ASP apps in IE VMs side-by-side with stuff on my Mac every day.
If it uses the virtualization kit, it shouldn’t be a problem. Unfortunately only veertu, xhive, and part of parallels do so
Unless it only allows ARM VMs.
They demoed Parallels running an x86 Debian VM
IE is dead and will be gone before Intel Macs are.
I know, that’s why I run it in compatibility mode in a VM. Just because it’s dead, doesn’t mean there aren’t organizations that have built mission critical web apps that only work in IE and those of us who have the burden of dealing with them.
You’ll be able to do that for so long that by then you can emulate the whole OS it runs on in a web browser. It’s obsolete after all.
What's the state of using cloud Windows images vs. local for this sort of testing?
Dunno. The free Windows VMs I have saved can be reloaded in VirtualBox and used pretty much forever. They can’t run any serious software very well (on my old MacBook anyway), but I’m just using them to support some legacy web apps.
Nobody cares about windows per se. People want to run legacy win32 x86 and x64 binaries.
Not necessarily. I'm working on some hard-core Excel spreadsheets, so the simplest solution was to run Office on Windows 10 in Parallels.
Sounds like you wanted to run office, not windows, which was exactly my point.
Ok, the lack of commas made it seem to me you were only talking about legacy software.
> Linux and Windows have both supported ARM for a while now.

Many people who want dual boot Windows/ Macs want them for games and gamers want the sort of performance you don't get with emulation.

Windows have both supported ARM for a while now

Windows has always been cross-platform, x86, MIPS, Alpha and PPC back in the day. There was even a SPARC build at one point.

Did they say the Linux VMs in the demo were x86? They said "virtualization", which to me, implies ARM Linux.
Yeah, the fact that they didn't show off Windows, but instead Debian of all things, was very telling.
And strange - Windows already runs on ARM doesn’t it?
I don't think there's a way to have a licensed Windows ARM copy right now on arbitrary hardware. I thought they only provided OEM versions on certified hardware.
> Yeah, the fact that they didn't show off Windows, but instead Debian of all things, was very telling.

I suspect developers running VMs with Linux on them is far more common than developers running Windows VMs. Likely by an order of magnitude. Web developers want Linux VMs, Windows developer have Windows laptops.

You'd be surprised. It was the only option for those writing native apps to have a platform that could legally run all of your tooling if you shipped Mac/Windows/Linux.
> It was the only option for those writing native apps to have a platform that could legally run all of your tooling if you shipped Mac/Windows/Linux.

Sure, but the number of people writing cross platform apps is a small fraction of the number of web developers out there.

But given the amount of time they talked about how large a part native apps play into the transitions, an extremely strategically important segment for them.

Additionally, there's the Android/iOS crowd in the same boat, where emulation of non x86 in Android dev is pretty limited (but I can see that being rectified with the newer virtualization extensions).

MacOS Subsystem for Linux
This is true, it might just be virtualizing ARM, I think I just assumed it was x86 emulation based on the context.
Since arm has a more relaxed memory model they will likely never be able to emulate x86 with multithreading well.
>how impressive x86 VMs looked

You mean that they opened and could run a command line app? I'm not actually sure they implied it was x86.

I guess it's not a VM, but they were running Maya with Rosetta2 weren't they?
I could see the "final Intel Macs" having a value to folks; somewhat similar to the 2015 Macbook Pros which many considered the "last good Macbooks" before Apple fumbled things with the 2016-onward Macs and their gimpy keyboards.

In this case, I don't think the first ARM Macs will have undesirable hardware ala the first few years of Touchbar Macs, but there will be some straggler software whose ARM ports will be delayed or will never happen. For those who depend upon that software, the final Intel Macs will be invaluable.

I still have my dual G5 PowerMac. Right now it runs Lubuntu.
I'm in the market for a new laptop, and if I knew how long the Intel macs would be supported, I could see myself picking one up. Unfortunately, nothing was said about that so I guess I'm holding off.
They said the transition should take 2 years. I wouldn't expect much more than that.
Disagree. He said Intel Macs would be supported for "years to come" and that they had more in the pipeline. They'll be supported for significantly longer than 2 years.
Standard Apple support is 7 years from last date of sale. 2 years for transition, so seems reasonable to expect support most of the way to 2030.
Apple dropped macOS support for their powerpc architecture after only 3 years, so any claims of hardware support being 7 years is not very meaningful
Those are the rumored "iPad Pro" aesthetic Macs which were initially rumored to be released today.
I was waiting for Craig to say that every Mac with a T-series chip could be used as a dev platform, since they're pretty strong on their own. Oh well. Maybe to smooth things over they'll ship an A-series co-processor with those Macs.

Apple released a bunch of PPC Macs after they announced the switch to Intel, so this isn't unprecedented.

The T-series CPUs are a bit weak, to be honest…
Personally I'd rather grab one of those as my daily driver, and wait a few years for all the issues to get worked out during the transition.
It was a backup plan if the Apple Silicon was delayed. Might as well still execute on it.
I wonder if these intel based ones are going to be on the high end of the Macbook Pro lineup or iMac.
Its total transition to Arm in 2 years. Likely they will build different SoCs performance/power for desktop and notebook.
Do you think most people know who makes the CPU inside their Apple machine? Go into an Apple store and take a survey. I almost guarantee no one will know.
That is so people don’t stop buying Macs into end of year
Apple announced the Mac in 1984 and sold Apple // based computers until 1993.

Apple announced the PPC in 1994 and sold new 68K Macs for 3 years at least.

If the PowerPC to Intel transition was any indication, they'll do just fine. Most consumers won't care, or will want to wait until Apple works out the kinks.

To put it a different way: the initial sales of the ARM-based chips won't be as strong as you think they might be, or as wide spread.

Having both Intel and ARM based systems in the market will even out the sales.

History is repeating itself, it's worth reviewing the previous transitions before making assumptions.

- https://everymac.com/mac-answers/macintel-faq/why-did-apple-...

- https://en.wikipedia.org/wiki/Apple%27s_transition_to_Intel_...

- https://tedium.co/2020/06/16/apple-powerpc-intel-transition-...

- https://mashable.com/2016/06/29/intel-macs-at-10/

With the powerpc to x86 transition Apple was moving towards where everyone else already was. It was one less thing developers had to worry about when supporting Mac. With this migration it's the opposite, and MacOS is still a very small slice of the desktop/laptop market.
I agree, it is a very small part of the market, but I'm not sure I grok your point. I think Apple's has leapfrogged Intel in some ways?
I’m pretty sure their point is that last time, the incentives/burden was more aligned between Apple and soptware producers, because it meant devs had one LESS architecture they had to care about. This time they will have one MORE, and it could very well be an architecture they have no familiarity with. And Apple isn’t exactly known for their technical documentation of new platforms.
I don't know that it's one more platform. I mean, sure, it's not exactly an iPhone, a Pine 64, a Raspberry Pi, or a tablet. It is, however, a very broadly supported processor family. More and more support is coming out for ARM all the time. If you're targeting desktop and anything smaller (phones, tablets, watches, SBCs, smart home devices) then the something smaller if it's bigger than an Arduino is probably ARM.

I think quite a few developers would be happy to use the same basic software stack on on a phone app and a laptop app.

If I'm making a professional desktop app like Blender or whatever, then right now I only need to worry about x86 SIMD optimizations & behaviors. I don't care about iPhone/iPad/Raspberry Pi performance. I may care that it works, because why not? But I don't care about optimizing for it.

Now, though, suddenly I need to figure out how NEON works if I want to continue to support professional users on MacOS. I need to increase my hardware costs to ensure adequate test coverage. I need to use something other than Intel's glorious VTune for profiling & optimization. When I'm step-debugging my C++ or assembly tight-loops, I need to know double the amount of assembly than I do today. I need to learn ARM's memory model & cache behaviors.

If I'm just writing some silly native app that could have been a website as is trendy then yeah this is all no big deal, who cares? But if you're really pushing professional app boundaries, where time is money? That's a different story.

Many things in the world of software are neither webapps nor contain a lot of highly optimized hand-written assembly. That's the worst false dichotomy I've read so far today, and I read political Twitter.

Just because your use case is different from those who have C or C++ code and a good optimizing compiler doesn't mean you need to disparage every other type of developer on the planet.

There are a lot of people who write native ARM for phones and tablets for performance purposes, too. Some of them get down into ARM assembly. If they port to Intel and want to do assembly-level things there, they need to learn twice as much assembly coming the other direction.

It depends on where your perception of "where everyone else already was" in this context.

Is the future more like Intel desktops or more like ARM-based mobile devices? If you think the latter then Apple is definitely moving towards where people will be (insert Gretzky quote: "skate to where the puck is going to be, not to where it has been.)

I remember Apple's m68k to PowerPC transition. Platform changes were a much bigger deal back then because it determined what kinds of apps you could run, but even so it went without a hitch. We lived with fat binaries for a while but otherwise it wasn't a big deal.

Apple's focus has always been the consumer market, and for most consumers, they're not likely to notice the platform change.

One might say, Apple gained a huge developer demographic when it moved to x64 from PowerPC, won't this matter to them? But most of those developers were creative types that don't work at the binary level anyway. If you're doing front-end web dev on VS Code and almost never fire up gcc/clang, your life isn't going to be impacted significantly.

Those whose lives were impacted (e.g. scientific computing folks who struggled to get certain Linux-specific software to compile on what was a BSD-ish platform) have probably moved off the Apple platform and on to Intel Linux years ago or at least they're likely to be running a Linux VM on their Macbook Pros. (homebrew doesn't really cut it)

Also unlike back in the day when most people were still running desktop apps, so much of what we do on a daily basis is on the browser nowadays. Calendar? Cloud-based. Email? Cloud-based. Word processor/spreadsheet/presentation? Cloud-based -- and Microsoft Office is being ported to ARM. We're living in much more platform-agnostic world than we were even 10 years ago.

I expect the x64 to ARM transition to be seamless to most people who are currently on the macOS platform.

A bunch of Python wheels might need to be recompiled though.

> I expect the x64 to ARM transition to be seamless to most people who are currently on the macOS platform.

I completely agree with you IF Apple contains this to just the Macbook market. If they push this to iMac Pro / Mac Pro, too, though? Or even Macbook Pro? That's a different story. For developers it probably won't matter much, but for users currently relying on things like Protools to make a living? There's a good chance this won't be a friendly transition to them in the short term.

The pro tools (like Pro Tools) are usually the tools that are being ported first. It was the same with all the other transitions.

If not, there will be Universal 2 binaries with both x64/Intel binaries for a while and an emulation mode for legacy binaries. Underoptimized at first? Maybe, but they work well enough and performance catches up eventually -- the transition is over 2 years. Same situation as before. Software that work will continue to work on existing and new hardware for a while. Nothing is going to stop working the moment the new processors come out.

We've seen all this happen before (m68k - PowerPC - Intel). And it's really not a problem.

Did the keynote give any indication that some machines would stay on Intel?
Yes, at the 1:46:10 mark. Tim Cook says they have some Intel based Macs in the pipeline.
ARM is a much bigger platform than x86 nowadays.
Apple has a strong story for x64, going forward, which I think will ease fears.

They aren’t rushing x64 out the door.

Maybe it depends on the pricing for the x86 vs. ARM macs. If the ARM machines are significantly cheaper the x86 sales will probably suffer quite a bit. As someone else in this thread said, most people have no idea what processor is in their computer. They just care that it's cheaper and runs all of their stuff (browsers, photos, messages, etc.)
Apple released new mac minis and powermacs when they announced the intel transition.
The Mac Pro (and possibly the iMac Pro) will still be Intel for a while. I expect the portable lineups to get the ARM treatment first (as they are the most similar with the iPhone/iPad in terms of power requirements), and then scale it up for actual high-power desktop CPUs.
Apple could continue to ship arm/x86 machines in parallel for years if the fat binaries+emulation work well. If say the desktop mac pro maintains an edge somehow, or Intel comes back from their current slump in 5 years this could go on indefinitely.

(that probably totally depends on whether Intel actually designs a desktop chip for the first time in nearly two decades, rather than continuing with their current sell server chips as poorly designed desktops path)

On the contrary, I'm tempted to buy the latest Intel Macbook Pro so I can stick with Intel for a few more years. Whether or not I'll end up doing so depends a lot on how good the new ARM-based laptops are in terms of performance (esp. wrt virtualizing x86/x64 VMs). But the current lineup, however imperfect, is mature and a known quantity, in contrast to the unknowns surrounding the first generation of machines with the new architecture.
They said the same with the Intel transition, and also projected a two year timeline. The transition was over much sooner than that and there weren't any new PPC products to speak of, as far as I remember. And of course, PPC Macs only got one more version of OS X before support for them was dropped.

They're just trying not to Osbourne themselves too badly. They want you to keep buying Intel Macs, but who's to say how long they'll keep support.

PPC was complicated by coming after years of failing to have decent chips actually ship. In this case, Intel and AMD are both still heavily invested in pushing x86 forward and Apple doesn’t have to do all of the work maintaining things like compilers or the JVM so it’d be a lot easier to keep that option alive if there was a good enough reason.
Now will Qualcomm feel a little more competition or not? There chips are slow in mobiles and they are slow in PCs, compared to Apple's processors.
They did not seem to care for the past years either so what does this change
Different segmentation - Qualcomm powered laptops are the weakest ultrabooks, compared to Macbook Pros. I really wish Qualcomm chips were better though.
Nobody gives a crap about the Qualcomm powered notebooks. They're a joke market segment. As long as Apple doesn't start selling A* chips to Samsung or other Android OEMs Qualcomm isn't going to worry.
> As long as Apple doesn't start selling A* chips to Samsung or other Android OEMs Qualcomm isn't going to worry

Even if Apple did sell their chips to other parties, Qualcomm will be still collecting its sweet patent royalty from both parties, so it couldn't care less.

The patent royalties are fairly small compared to selling a full chip. They would still be feeling the pinch.

Of course even if Apple did start selling those chips (which they won't) they would charge an arm and a leg for them, because they could. They would only be used in the top of the line flagship phones. Apple has shown zero interest in competing with the bargain basement hardware producers.

“We've been down this road before”

Yes, and then you ‘upgraded’ the OS and gratuitously stopped running older software. My final Mac Pro — yes, I was a buyer of Mac Pro grade hardware — still runs 10.6.8 for that reason.

https://youtu.be/UDfAdHBtK_Q

MacMini with an A12 chip. Available now as a part of a developer kit. Will be fun to see what that can do.
I wonder what the cost of the transition mini is.
The last time Apple transitioned to a new architecture, the dev kits were loaners you had to give back.
And you most likely will have to be an established developer, meaning you’ll have to have an app in the App Store. Apple’s not gonna give these to people that just want to play around with a new machine.
Yep, confirmed.

https://developer.apple.com/programs/universal/

"Submit a brief application for an opportunity to join the program. Selected developers will receive a link to order the Universal App Quick Start Program from the online Apple Store. Priority will be given to applicants with an existing macOS application, as availability is limited."

Those dev kits ran on P4 processors. IIRC, they later sent all the devs a Core 2 mac pro.
If it's like the switch to Intel then it will be $999 and you have to return it when the real versions come out but you'll get a replacement with one of those real versions. Except with the Intel switch you were shipped a slightly repackaged beige box... and that was 15 years ago so my guess is $1499.
Didn’t that also include a discount on the final replacement equivalent?
Yep it did. I think you could get an iMac for 1000-1500 dollars once they were released to the public.
It is $500 and you have to return it at some point.
And a two year transition period.

That sounds ambitious as the bigger Apple desktop machines are quite powerful. But suppose you could just fit them with multiple chips.

If they keep the desktop thermal budget I have no doubt they could scale up the A series to be just as powerful as any intel chip.
I doubt whether making big chip just for Mac Pro/iMac Pro profitable. Chiplet approach looks like benefitical.
I wonder if they could release something to turn iPad Pros into development kits. Would be readily available for a lot of devs.
I was seriously considering upgrading my 2012 MacBook to the new 16" model soon, but now I wonder what kind of longevity or support I can expect from that hardware...
I just got a 16" MBP... Still, the waiting game sucks too. Apple said their first Arm mac will come out at the end of the year, but that could well be a very low end Mac mini, or Macbook Air. It's unknown when the pro line will come out.
There was an article posted here on HN yesterday speculating that the first ARM Mac will be a new 13” MacBook Pro.
That might explain why all but the top line version of the barely-refreshed 13+ MacBook Pro are missing Intel's 10th-gen processors.

Sandbag the upgrade for the 13" so it's not that desirable and you won't cannibalize your own demand for the release of the first Apple Silicon machine

If that's the case, Apple is deliberately hindering their intel version so that the future ARM version destroys it in benchmarks.

They'll win anyway. But comparing a 2021-ish Apple ARM laptop against intel's early 2019 chip isn't particularly fair.

People will see the benchmark as a blow to intel (and it is), but the fair benchmarks would be a 2021 ARM mac against the best intel has to offer

Given Apple's track record of supporting older Mac hardware, I would expect Intel Mac hardware to be supported for many years and many OS versions after the last Intel Mac has been produced.

Also given Apple's recent MacBook track record, the first version of ARM Macs may not be as much of a slam dunk as people hope.

Which track record are you talking about? Apple discontinued OS support for PPC only 3 years after intel was announced. That means no more support for xcode and new commandline tools after 3.5 years.
I think we’ve got some time. I’m on the last leg of a 2012 too. Just like the sibling comment said, since we don’t know if the pro line will get a refresh there could possibly be a longer runway.
My personal philosophy is to buy the third gen of every new type of product. The second gen tends to fix al big issues in the first. The third irons out all the small issues left. After that there probably won't be many changes except for some performance bumps.
Docker and Linux virtualization. Fat binaries to make it easy to update Mac apps. iPhone and iPad apps. Rosetta to run Intel apps that aren’t updated.

Apple is gonna knock this out if the park.

It already has. I'll Def try to get my hands on the A12Z Mac Mini. The GPU performance should be vastly better than existing Mac Mini options.
I wonder what this will do to Electron. If the iOS apps are really 1:1 on macOS, then the need to maintain an electron app will probably diminish. As long as they both support the same OS APIs I can see devs that can learn a new language (Swift) ditch Electron.
Isn't this already the case?

iOS apps can be built, for Intel, using Catalyst, with very trivial changes in code, but we still see Electron hanging around today like a bad smell.

I can't imagine this will change anything.

But have you actually used a Catalyst app on mac? They're terrible ports as it is right now.
Eh, Catalyst apps are as good or bad as the developer wants them to be. Voice Memos and "Find My" on the Mac are two fantastic Catalyst apps, and certainly better than Electron.

Anyways Catalyst won't even be relevant in the long term, once iPhone apps are written in SwiftUI

Interestingly there's a few mentions of Catalyst improvements in the Xcode 12 release notes:

> When bringing iPad apps to macOS, you can now use the Optimize Interface for Mac target setting to use native macOS controls and Mac resolution

Hopefully this means Catalyst apps will feel a little more like actual Mac apps.

I often close the Twitter app and Apple's news and stock apps on my Mac mini because the performance is terrible. Hoping they've spent time tuning this more.
Catalyst and SwiftUI are (were?) very immature technologies, not yet ideal for production software. I'd imagine that we're at least a year out from seeing real SwiftUI software in the wild.
Mac and iOS aren't the only two platforms.
But if you're developing with Electron, the purpose is typically cross platform desktop support. This won't change that?
I'd wager for a significant amount of shops it's less about cross-platform support and more for being able to throw existing generalist or web frontend developers into native development. If a business wants to ship a desktop app quickly, it's hard to argue against electron because your existing teams can become productive without too much training.
Hrmn. Wouldn't this be leaving out android and windows? It might be practical for some apps, but that's an awful lot of users.
Electron already has support for ARM64, but no official releases yet. But it needs to be build from x86 machine. No native compilation on ARM64 yet. I think with apple moving to ARM, google will add native ARM64 compilation for chromium. This in turn will be picked by electron. Chromium has been running on ARM for a long time with Android and Chrome OS so it has all the optimizations.
Is there anyone building on Electron and only targeting the Mac desktop environment? Windows is still the king there.
Is MS Teams for Windows written with Electron? It is on Mac.
Apple had a list at the State of the Union of open source technology projects they had built pull requests for to add ARM support. Electron was up there, as was Python 3, OpenJDK and Go, notably.
Would love a link to that if you still have it on hand.
The processor architecture is not that relevant if you are working at the abstraction level that electron or swift-ui or similar provides.
I am soo much looking forward to using native Slack and Teams instead of their horrendous electron apps, that even don't use GPU acceleration on iGPU Macbooks!
There's gonna be a weird moment soon where developing docker images for Armbian hardware (Pi and friends) is going to be more straightforward on a Mac than developing docker images for Intel servers.
And that may be the moment that ARM servers start picking up steam.
I wonder what sort of involvement ARM server vendors could be offering during this period. I don't expect Apple 'needs' anything from them, but there might be some Docker work they could be jumping on.
First mac with Apple chip by the end of the year.

Also Apple is committing to a long term pipeline of Intel chips. This makes sense since so many apps will take years to transition. At the same time, they're willing to put their own chips side by side with Intel and they believe people will voluntarily switch. I'm looking forward to the benchmarks.

If the performance in the demo is accurate, and considering that the actual Mac chips will have a lot bigger power headroom, silicon die size and more thermal headrooms than the A12Z, Intel will be swatted out.
Here is a though how about making Rosetta run those 32 bit applications you dropped on the floor with Catalina.
An app the vender thought not worth updating to 64 bit isn’t worth saving.
So, if I install numpy via conda on a Mac now, it's backed by Intel MKL and is thus amazingly fast. What will it be replaced with? Has anyone at Apple thought about use cases like this?..
I bet that they already have some low-level math library that uses ARM NEON intrinsics; you would definitely need them to port performance-demanding apps like Final Cut Pro / Maya / Photoshop / etc.
PCs are for gaming. PCs are for work. Macs are for... students and middle managers apparently.
And about 30% of software devs, according to Stack Overflow.
Right, apparently only a quarter of all developers uses them (https://insights.stackoverflow.com/survey/2019#technology-_-...).
Including me. Will those same developers buy a machine that is completely locked down? I will not.
Sorry, I don't understand what you're referring to. Locked down in what way?
I'm surprised that Windows has ~50%.

I'm not a developer myself, but When I engaging with them on various OSS projects I always feel Windows is being treated as a second class citizen in almost every aspect (smaller projects often has no test, tool chain setup, tutorial, or all of above for Windows, as a starter).

I think for this reason they are not going to discontinue intel based macs as x86 is the default for many high performance software.
They claimed they would "complete this transition" in 2 years.
They also stated they'll be supporting Intel Macs for a long time, and have new, unreleased, Intel-based machines in the pipeline.
Isn't that what VecLib is for? Also there are other math libraries to use on ARM. I think we use EigenBLAS in our products.
numpy is backed by BLAS and LAPACK. It just happens that on your system on macOS those libraries are provided my MKL. There are other implementations of BLAS and LAPACK out there.
numpy/scipy/etc will have to support Accelerate instead of MKL.
SciPy did support Accelerate, but they dropped it in 2018 because Apple’s implementation of LAPACK was so out of date. Apple have been crap at updating this stuff which is why people don’t use it.
Rosetta 2 will take care of translation and JITing
Well, that new MBP I bought 4 weeks ago and planned to keep for 5 years just got a muuuuch shorter life span if I can run iOS apps on an Arm Mac, my wife is going to be livid.

I guess this also means Catalyst’s life span is pretty short and SwiftUI will becomes the focus?

This if anything makes Catalyst more important.
I'd be surprised if the iOS apps did not work on Intel based macOS. Project Catalyst already works on Intel macs and my guess is many iOS apps have been migrated already.
I'm sure there will be emulators. But I'm also fairly confident that they'll be relatively slow. Emulating across architectures is rarely performant, and if Apple had solved the problem they would be talking a lot about it right now. In the past they've gotten away with this because the architecture they're moving to was so much faster than the previous architecture that even with a 50% or 75% performance penalty the apps would run faster than they did on the old hardware. With this new hardware it is likely only going to be marginally faster than the old Intel chips since the focus is more on power efficiency, so emulated apps are probably going to feel sluggish.
The difference is that the entire (64bit?) iOS App Store back catalogue would likely be available to run as unmodified binaries without developers having to lift a finger.
I think we’ll know more after the Platform State of the Union
SwiftUI and Catalyst are not mutually exclusive.

I started a new iOS app recently, and there was essentially no cost to building for catalyst.

I am using SwiftUI and I considered targetting MacOS directly, but the extra work wasn’t worth it at this stage.

I think catalyst is the new Cocoa - it will grow to consume all of MacOS.

Is it just me or did apple just bungle this entire announcement by not announcing a consumer facing ARM MacOS device, only a hot-rodded Mac Mini with an iPad Pro chip inside.

How many devs actually have the setup in place to use a non mobile device?

I also wonder if the current+last gen iPad Pro that has the new keyboard + trackpad case will gain the ability to run Xcode and native macOS apps in the near future.

What? It's exactly the same thing they did last time, except the dev kit was in a Power Mac G5 case.
They do their hardware releases in the other conferences at predictable times of the year
(comment deleted)
Native virtualization and the ability to run iPhone and iPad apps directly on the Mac. This could unlock a whole new level of focus and optimizations for Apple.
Will BootCamp continue to be supported for the next ten years on my June 2019 MacBook Pro so I can install an operating system that continues to develop new features with the assumption I'm running on x86-64?
(comment deleted)
What do you mean by "continue to be supported"? Your 2019 MBP will only need new drivers for Windows if Microsoft decides to change things. Otherwise, Apple doesn't have to do anything at all for your current MBP to continue to be able to run Windows. While Windows may be something of a moving target, your particular machine isn't.
I mean, there's nothing too magical about boot camp—even if apple removed all support for it from macOS, you can probably install windows on a partition the normal way without too much trouble. I think the only thing that might break is if Windows stopped being compatible with the Mac hardware drivers, and Apple didn't update them.
My comment from a week ago:

"What would keep Apple from shipping machines with both ARM and Intel CPUs in them? The ARM CPU would run the OS and decide when to ship jobs over to the Intel CPU. I can’t imagine that the home-grown ARM CPU would add much to the total price of the computer."

https://news.ycombinator.com/item?id=23530084

It is a lot of work and now you have to buy two CPUs. No cost savings, possibly more power usage in worst case, and maybe even thicker computers.

It would restrict it so you can never remove the Intel CPU. It is best to just emulate the Intel CPU for a while until everyone ports their code across.

Dual CPUs never made any sense. That’s why they are doing emulation and binary conversion.

Maybe Apple adds the i86 instruction set as a module in their SOCs, but they aren’t paying intel CPU prices on their ARM Macs.

Did I misinterpret today? Didn't Apple just release a machine with ARM and Intel?
It's an ARM only Developer Kit.
Ah. I saw references to "dual architecture" online but looks like that means Apple will be releasing new Intel machines even as they are encouraging everyone to port their apps to ARM. The "dual" part just means two different lines of machines that will be sold contemporaneously for a little while, not two architectures inside one machine.

Oh well, I got excited for a minute there.

It’s an ARM chip that can run x86 software using translation tools (Rosetta).
This level of asymmetric processing is very difficult to achieve, at nearly every possible level, from part sourcing, through hardware design, firmware, software to user experience.
Which was why I thought Apple might blow everyone away by doing it. But I was obviously wrong.
Well, in theory both AMD and ARM use MOESI so it's not as crazy as it would be with Intel's MESIF. But that's a really scary rabbit hole to be going down and I'd be terrified of memory permissions attacks striking at the boundary there. So probably not worth the risk unless each processor gets its own pool of RAM.

https://en.wikipedia.org/wiki/MOESI_protocol

Rosetta 2 - the interesting bit was that it was going to pre-translate binaries instead of at runtime. The implications for actual VM emulation is that Rosetta won't work for run time environments like OS emulation. They touched on it briefly with the emulation technologies bit, but it looks like it will be separate from, and likely much less performant than Rosetta.
The emulation thing seemed to me to be Hypervisor.framework for ARM, as they showed Linux and Docker running (which both run on ARM), but not Windows (which an average user may be more interested in).
It's doing both AOT and also hooks JIT compilers. They mentioned JS engines as an example for the latter.
The same slide also mentioned supporting JIT translation (for x86 web browsers and Java), so Rosetta doesn't run only at installation time.
They explicitly said that it can perform both static and dynamic translation for JITs. I wouldn't be surprised if there is substantial hardware support too.
My brain skipped over that part apparently. It is interesting that they are calling out virtualization improvements separately though.
Apple's chips haven't supported that yet, so it's certainly something new.
I was wondering why did they mention the virtualization. Let's see what technology do they use. Whether it is going to be proprietary or something like Xen.
It's going to be an ARM clone of Hypervisor.Framework.
In the state of the union they mentioned an Apple hypervisor
> They explicitly said that it can perform both static and dynamic translation for JITs

Can a JIT be static? Isn't that not possible by definition?

I meant that it supports dynamic translation in order to support JITs such as turbofan.

But to answer your question, yes a JIT can be static. JIT just means that the compilation happens at runtime, and "static" in this context means that the compilation is happening at the very start of runtime. You could imagine a JIT that compiles all bytecode to native code immediately on launch. The reason this technique is not used often is that it tends to lead to long startup times. But if the result is cached somewhere then it might be acceptable.

> Rosetta won't work for run time environments like OS emulation

They literally showed a demo of Parallels with Ubuntu for Intel running inside it.

Did they show the architecture at any point? I didn't see the architecture shown in the VM at all, but could have missed it.
How do you know it was for Intel?
Yes, but probably a native ARM version of Ubuntu rather than an emulated x86 version.
It was debian, and in the state of the union talk, they called it out as an ARM build of debian.
Good time for anyone who was using macOS for gaming or anything else for that matter to switch to Linux. OpenGL bit rot, refusal to support Vulkan, dropping of 32-bit, dropping of x86_64 architecture - all that should have been a hint. Backwards compatibility is not even an option there (besides for emulation).
I started using MacBooks when they switched to Intel around 2008. Maybe this is a sign to move on. Currently using a mid 2014 model, I like the slim form factor, but I don't like the absence of an ethernet port. I don't care much for the retina display, my 2008 matte model was better. Been looking at the Lenovo T470. I need something that would be easy to replace quickly anywhere in the world. What should I get?
Lenovo Thinkpad laptops are pretty good, and in my experience run Linux very well. Get something with AMD APU (there should be new models coming out soon with Zen 2 + Vega).

The only annoyance is their refusal to refund Windows tax. But now Lenovo partnered with RedHat/IBM and started selling some laptops with Linux pre-installed (Fedora), so you can avoid Windows tax there, even if you don't plan to use Fedora. I hope that will eventually extend to all their models.