Apple to keep Intel at Arm's length: macOS shifts from x86 to homegrown common CPU arch, will run iOS apps
- Reference: 1592872938
- News link: https://www.theregister.co.uk/2020/06/23/apple_armed_for_x86_rebellion/
- Source link:
During its [1]2020 Worldwide Developer Conference keynote – an virtual event due to the coronavirus pandemic – the computerized bling biz said it expects to ship the first Mac sporting Apple silicon around the end of the year, with a full transition taking about two years. The unspecified device, presumably a MacBook Pro, will come with macOS 11 (Big Sur) which is slated to ship this fall.
Apple [2]shocked some by [3]switching to Intel processors in June 2006, ditching IBM chips. Now it's dumping Chipzilla for its own silicon, which it claims will lead to minimal disruption - although the status of dual-boot systems is still unclear.
Apple is introducing a new binary compilation target for apps in its Xcode developer tool called Universal 2 that bundles native code for Apple Silicon and Intel x86_64. It's also offering Rosetta 2, a virtualization layer to allow legacy x86_64 code to run on the upcoming Arm-based Macs. Rosetta 2 will translate x86_64 code on installation or on-the-fly in the case of browsers using JIT-compiled JavaScript or Java.
A new Virtualization layer is also in the works for running Linux VMs and Docker. Few details were provided during the WWDC video event, apart from Apple hardware SVP Johny Srouji demonstrating the launch of an Apache web server from a Linux VM via the command line. Apple is also patching [4]for various open source projects like Chromium, Node, and V8 so they will run on Apple silicon.
To help developers convert their macOS apps to Arm-based chips, Apple has introduced the Universal App Quick Start Program, which provides support in the form of documentation, forums, beta versions of macOS Big Sur, Xcode 12, and the "limited use" of Arm-based hardware for app testing – a Mac mini with an A12Z Bionic SoC, equipped with 16GB of memory and a 512GB SSD.
Developers in 31 countries are eligible to apply and, if accepted, must pay $500. "Limited use" means the hardware must be returned to Apple within a year of acceptance. The kits start shipping this week.
Heavy cost
It's been [5]estimated that Apple's slow-motion abandonment of Intel silicon – a transition that has been rumored for several years and will take at least two more – could cost Intel about $3bn annually or 4 per cent of its revenue. Intel's stock closed up on Monday, a sign investors have already incorporated the possibility of a transition into their share valuation.
Fujitsu, Japan strong-Arm their way to the top with world's fastest-known super: 415-PFLOPS Fugaku [6]READ MORE
Despite celebrating the shift to Apple silicon during his WWDC keynote presentation, Apple CEO Tim Cook made it that Intel isn't being kicked to the curb immediately.
"We plan to continue to support and release new versions of Mac OS for Intel based Macs for years to come," said Cook. "In fact, we have some new Intel-based Macs in the pipeline that we're really excited about."
Cook perhaps is referring to an update for Apple's Mac Pro workstation, where power utilization and battery life top out around [7]902 watts for the 2019 model.
To underscore the seriousness of the architecture transition, Apple SVP of software engineering Craig Federighi noted that Microsoft and Adobe have been testing Apple silicon-builds of Office and the Creative Cloud suite respectively.
In an email to The Register , Patrick Moorhead, president and principal analyst of Moor Insights and Strategy, said that the impact on Intel should be minimal but is still a hit to the company's reputation.
"I am very disappointed Apple didn't provide a lot of technical details to give confidence to users and developers in the Arm-based processor Mac experience. We know Apple will profit from a reduced BOM [bill-of-materials] cost, but before developers lift a finger, they should ask themselves what they get from making investments to changes in any of their code."
One benefit will be the ability to run iOS and iPadOS apps natively on macOS devices with Apple silicon. Going the opposite direction, running macOS apps to iOS/iPadOS, may eventually be allowed but apps designed for large screens like Logic Pro X can't just be moved to iOS without major changes. Also, Apple would need to harmonize its app review rules between the macOS and iOS App Stores.
Beyond its altered underpinnings, the major changes in macOS Big Sur, available to developers today, have to do with aesthetic improvements and enhancements to apps like Maps, Messages, and Safari.
Federighi touted Safari's JavaScript speed and page loading. "When loading frequently visited websites, Safari is now an average of more than 50 per cent faster than Chrome and Safari delivers this amazing performance while continuing to deliver industry-leading battery life," he said.
Safari will make ad tech more visible to users by showing which websites implement tracking code, the sort of information previously revealed by various privacy-focused browser extensions. In the forthcoming version of Safari, he said, "users can click on the privacy report button in the toolbar when they visit a site to better understand how that site is treating their privacy."
Perhaps this will encourage more people to use the desktop version of Safari, which only has about 10 per cent global market share, according to [8]StatCounter .
Extending support, and potential problems
Apple also plans to distribute Safari extensions through the Mac App Store. "We're adding support for the Web Extensions API so developers can easily bring over extensions that they built for other browsers," he said. "And we're building an all new category in the Mac App Store to showcase Safari extensions, so users can easily find them."
Acknowledging the persistent privacy and security problems Google has wrestled with in its Chrome Web Store, Federighi said Safari will ask users to grant access permissions to extensions as follows: "Allow for One Day," "Always Allow on This Website," or "Always Allow for Every Website."
Apple's iOS 14 delivers a Home Screen cleanup with a feature called App Library that allows users to organize serial screens of apps into categorized collections collected on a single screen. It's like iOS folders but outside the context of the screen-swiping interface convention that falls down at scale.
Widgets have been rethought and in iOS 14 will be able to be affixed to Home Screen pages. And Picture in Picture, enjoyed by Android users since Android 8.0 Oreo in 2017, is coming to iOS, so users will be able to watch videos over other applications.
Siri is getting a minimized mode, to be less intrusive, and should be more competent at answering questions. Siri's dictation capability will work on-device and offline.
Apple is also taking on Google Translate with its own Translate app for handling realtime conversational translation. The major selling point is privacy: The app's speech recognition model can operate on-device, so no data about conversations gets sent to the cloud.
The iMessages app, like its macOS cousin Messages, is getting touched up with the ability to pin chat conversations, with group chat improvements like inline replies and mentions, and new Memojis for representing different hairstyles and COVID-19-era masks.
The recently revised Apple Maps is headed to Canada, Ireland, and the UK later this year with an additional mode of transportation option, Cycling.
Clips, but not Clippy
Possibly the most substantive change in iOS has to do with the way apps get found and distributed. Apple is launching a developer-oriented capability called App Clips that allows iDevice users to download minimal versions of apps on-demand.
Think of them as Apple's take on [9]Google Play Instant apps or [10]Progressive Web Apps , code optimized for easy discovery and no-fuss installation. App Clips can be no more than 10MB, to ensure that they download quickly.
To facilitate distribution, Apple has developed App Clip Codes, which encode a URL and incorporate an NFC chip reference. These codes can be referenced by NFC tags, QR codes, Smart App Banners, links in Messages/iMessages, and place cards in Maps. The idea is that an iPhone user can wave a phone over an App Clip Code and get a vendor-specific app more or less immediately.
App Clips can also contain multiple applets. "We made it possible for apps like Yelp, which support multiple businesses, to create applet experiences for each of the places they work with," said Federighi.
The full app is not installed on the device (and thus doesn't add more clutter), though full installation can be accomplished via the Recents category of the new App Library, where the App Clip reference is temporarily stored.
iPadOS, watchOS, and tvOS were also mentioned. We're looking forward to watchOS and its enforcement of public health norms, in the form of Automatic Handwashing Detection.
"Ou approach here is using machine learning models to determine motion which appears to be hand washing, and then use audio to confirm the sound of running water or squishing soap on your hands," explained Kevin Lynch, VP of technology.
"During this, you'll get a little coaching to do a good job. You'll see a countdown, along with haptics and sounds, to make sure you wash as long as you're supposed to. If you pause early there's a plate note to keep washing. And when you're done, you'll see, hear and feel it." ®
Sponsored: [11]Running Your Modern .NET Application on Kubernetes
[1] https://developer.apple.com/wwdc20/
[2] https://www.theregister.com/2006/05/19/pasemi_apple/
[3] https://www.theregister.com/2006/08/07/apple_xeon/
[4] https://twitter.com/wongmjane/status/1275177255681982464?s=20
[5] https://www.wsj.com/articles/apple-expected-to-break-up-with-intel-transition-to-in-house-chips-for-macs-11592844482
[6] https://www.theregister.com/2020/06/22/japan_arm_supercomputer_fugaku/
[7] https://support.apple.com/en-us/HT201796
[8] https://gs.statcounter.com/browser-market-share/desktop/worldwide
[9] https://developer.android.com/topic/google-play-instant/overview
[10] https://web.dev/progressive-web-apps/
[11] https://go.theregister.com/tl/1956/-8477/running-your-modern-net-application-on-kubernetes?td=wptl1956
"Intel never thrilled me"
Everything before the 80386 was horrible. x86-64 isn't at all bad. Who cares where the CPU comes from? How it performs is what's important. These days, security at the hardware level is a hot topic. Intel's doing things the old monolithic way. I suspect the ARM approach allows CPU designers to come up with more secure CPUs quicker than Intel's way of doing things - especially in the case of Apple, which can now design a complete package of CPU, full computer circuitry, OS, and software developer tools in a fully controlled ecosystem. Authoritarianism is a rotten way to run a country, but often works well in engineering.
- my main point there being: since Apple's got control of everything from CPU design upwards, it can implement systems to mitigate security flaws at any given level by tweaking security systems at any other level.
"Rosetta 2 will translate x86_64 code on installation or on-the-fly in the case of browsers using JIT-compiled JavaScript or Java."
I lol'd
Rosetta
The original Rosetta from 2006 worked pretty well: PPC applications mostly did run on new Intel Macs without trouble.
Prior to that transition, Apple implemented the "Classic" environment so that you could boot up the old-style MacOS 9 on a MacOS X PPC Mac, and run 68k code while you were about it. That worked well too.
The software to run 68k code on a PPC Mac came from when Apple moved Macs from 68k CPUs to PPC CPUs. It mostly worked well - although back in the days of the 68k to PPC transition, I did notice a performance hit.
As it happens, there's a thing called SheepShaver that lets you install MacOS 9 on an Intel Mac running the current Mac OS and run old PPC software on Intel Macs even now. It's not from Apple, but it works.
Apple has form in this area: it's done well in the past. I'd bet more than a pint that Rosetta 2 will outperform its predecessors.
Re: Rosetta
MacBook Pro?
Surely a low power MacBook would be the best choice given the need to Rosetta most apps to start with?
I’m not sure why you wouldn’t do an arm coprocessor, for a transition.
Getting away from Chipzilla is all well and good, but to ditch the entire architecture instead of chatting up AMD seems like a bit of a stretch. Plenty have tried this whole mainstream ARM thing before, and all of them have failed, or at least not succeeded in the way they thought.
ARM?
Hasn't been successful before?
Never underestimate the power of the Sheeple. Apple's attempt may prove to be the mouldbreaker.
Re: ARM?
I guess we'll see. IMO it will either work wonderfully, or flop spectacularly. I don't foresee much middle ground.
I'll give Apple the benefit of the doubt on this one - They've managed the last two processor transitions reasonably well, and for a project that's supposedly been swirling around for nearly a decade, I get the feeling they wouldn't have announced it now if they weren't confident it's ready and they can pull it off.
Really?
ARM is mainstream if you're looking at smartphones. Current smartphones have more CPU power and more memory and faster communication than the most powerful Microsoft or Apple OS desktop PCs of the 1990s. Quite a lot of them have higher resolution screens, too.
Tablets are another field where ARM has a strong presence. Desktop PCs are becoming less important; laptops more so. Laptops absolutely must have good electrical power efficiency: ARM scores highly there.
And don't forget: Apple designed Macs with Motorola's 68k CPUs, switched to the Motorola/IBM PPC CPU architecture, then dropped that for Intel. This is simply the latest change of CPU architecture for the Mac line: hardly "a bit of a stretch", rather more "business as usual".
Re: Really?
Current smartphones have more CPU power and more memory and faster communication than the most powerful Microsoft or Apple OS desktop PCs of the 1990s. I'd even go so far as to say the mid-late 2000s, and that's possibly being conservative.
Apple baked chips
Whatever.
But, a great reason to increase pricing due to all that secret holy Apple-tastic circuitry.
more Apple marketing
If it works as well as they claim it does, it won't make a bit of a difference to the end-user. It's just a different compiler target. And different opcodes. Does the end-user care about opcodes?
If it doesn't work that well, what's the point?
The only thing worse than being talked about is not being talked about.
Compatibility is gonna be a problem.
If Windows, that can run the kitchen sink from three generations ago, needs an emulator to run 32 bit software and some of it even needs a virtual machine, imagine how it will work with Apple who is way more happy to break compatibility with old stuff.
It'll work.
Apple has done a Mac CPU switch twice before. Macs started out on 68k Motorola CPUs. 68k failed to keep up, so Apple went for PowerPC (Motorola/IBM). PPC could have kept up, only IBM wanted paying up front to develop updated CPUs, so Apple went for Intel because Intel could deliver adequate CPUs for less money.
Each time, the transition worked well enough.
Each time, Apple provided adequate backwards compatibility by enabling most old CPU code to run on the new machines well enough for long enough for most users, and made arrangements for updated applications to install with code appropriate for both old and new CPUs (back in the 68k/PPC era, one could select one or the other so as to save on disc space, even if that did sometimes require a third party utility to strip out the unwanted code). I used Macs through both transitions. Rosetta - the original version from 2006 which let you run PPC applications on Intel Macs - worked very well. I expect Rosetta 2 will work better
And now, Apple has figured out how to roll its own CPU well enough that it beats Intel's offering for PC/laptop jobs. So for the third time, Apple's going to switch the Mac CPU to a new architecture. It's a well-trodden path for Apple. The firm knows how to do it, the user base knows what to expect, and the developers know what to expect.
Really, it's not that big a deal. Mac users will in future have more power-efficient hardware which runs the code they want to use at good speed with good reliability and good security. Some Mac users will end up having to abandon ancient software that's not being maintained (that'll be me, then). The change in Intel's sales will be tiny.
But perhaps this is a sign of what is to come.
Intel's x86 architecture has been around for a long time. Then again, x86 is more of a marketing term than anything else these days, since 21st century x86-64 architecture has pretty much nothing to do with the original 8086 CPU from 1978. Given how long Intel's been in the CPU game, and how well it's done, I suspect the firm will work out how to keep its head above water - and maybe the term x86 architecture is on its way to the history books.
Wot?
Cook perhaps is referring to an update for Apple's Mac Pro workstation, where power utilization and battery life top out around 902 watts for the 2019 model.
What battery other than a very tiny watch sized one does the Mac Pro have? The Mac Pro, like the iMac has to be fed with 230vAC (of 115VAC). I agree for the macBook pro and that it needs more battery life.
My 2015 MBP (bought secondhand) is getting rather long in the tooth. As I mostly write fiction on it these days I may well get one of the new devices in a year or so. I'll let all the Fanboi's rush out and get them first. Then they can wail long and hard about this function or that function not working like it did in 2009. ROFL. Things change people. I may even pick up one secondhand at a good Apple Tax free price.
Two years to work out the kinks and I am buying one
Intel never thrilled me (and I know this may be an irrational decision).