So it looks like all three M1’s run at the same speed? (At least until the Air gets too hot?) So there is no speed bump from the Mini having 120v power supply?
Were they?In the test, though both M1 and Intel machines were building WebKit, but actually they were building for different targets (ARMv8 and x86_64). The compilation/optimization/code-generation/etc for different targets could vary in many ways.
Curious to know the numbers if we cross-build for x86_64 on M1 or cross-build for ARMv8 on Intel.
From the article: "Right up top I’m going to start off with the real ‘oh ****’ chart of this piece. I checked WebKit out from GitHub and ran a build on all of the machines with no parameters"Were they?
The test is not Apples to Apples then
For newbies who dont speak chinese, what do you say?From the article: "Right up top I’m going to start off with the real ‘oh ****’ chart of this piece. I checked WebKit out from GitHub and ran a build on all of the machines with no parameters"
with no parameters -- I guess this means the compiler will build for the same architecture as the host. It this is true, yap this is not Apples to Apples comparison. (This is indeed "Apple (M1)" to "Apple (Intel)" comparison though)
Ha ha, it's just that they used the default settings so that probably means the Intel-app compiler on Intel and the ARM-app compiler on M1.For newbies who dont speak chinese, what do you say?
I have a feeling Parallels and VMWare are either going to secure the ability to license Windows 10 on ARM from Microsoft to offer their customers (which has built in x86 emulation) or build out their own x86 emulation to continue offering Windows on ASi Macs. Given the performance we’re seeing from the M1 for native apps, a well crafted emulator could easily deliver performance on par with an Intel processor from a few years ago. CPU speeds and emulation capabilities have come a long way since the dark days of SoftWindows and VirtualPC in the late 90’s.I'd like to be shopping for AS, but TBH I'm not sure I'm ready to walk away from Windows virtualization and BootCamp quite yet. And I know the software support isn't there yet, either way.
Parallels has already announced it.I have a feeling Parallels and VMWare are either going to secure the ability to license Windows 10 on ARM from Microsoft to offer their customers (which has built in x86 emulation) or build out their own x86 emulation to continue offering Windows on ASi Macs. Given the performance we’re seeing from the M1 for native apps, a well crafted emulator could easily deliver performance on par with an Intel processor from a few years ago. CPU speeds and emulation capabilities have come a long way since the dark days of SoftWindows and VirtualPC in the late 90’s.
I did see that they announced M1 support, and the press release does mention Windows 10 on ARM, but has there been any word from Microsoft on whether they’ll allow non-OEM licenses? The press release you linked to contains yet another link to a Microsoft press release that still describes them as working with only hardware partners on this. That would seem to be the missing link (and the easiest problem to solve).Parallels has already announced it.
![]()
Parallels Desktop for Mac with Apple M1 chip
Best way to run Windows on Mac - Parallels Product Blog - Run Windows, Linux, and more on your Mac. Helpful tips and tricks, Apple macOS and Mirosoft Windows newswww.parallels.com
As I read it, that means no emulation on the Parallels side, but expecting Windows 10 on ARM to do x64 emulation within it.Since WWDC, our new version of Parallels Desktop which runs on Mac with Apple M1 chip has made tremendous progress. We switched Parallels Desktop to universal binary and optimized its virtualization code; and the version that we are eager to try on these new MacBook Air, Mac mini and MacBook Pro 13″ looks very promising. Parallels is also amazed by the news from Microsoft about adding support of x64 applications in Windows on ARM.
Well, I know only basic principles of programming but for Lr Classic to run natively doesn't the code need to be completely new? Eg. all calls to any X86 or X64 processes need to be removed or rewritten.What makes you think there will be a "complete rewrite"? 99.9% of the code won't be touched.
Very few modern programs use actual CPU-specific assembly code anymore. Most use high-level languages (Objective-C, Swift) that make calls to OS-level APIs which can be readily compiled to run on different CPU types. The big headache with the PPC->Intel transition is that most of the major apps were legacy Carbon apps that came from Classic Mac OS and were barely updated to run on Mac OS X. Most of these legacy apps have actually already been rewritten using modern Cocoa APIs since then - Adobe products were rewritten in Cocoa as part of the CS5 release in 2011 and Office was rewritten in Cocoa as part of Office 2016. Cocoa-based apps are easier to port to other architectures, so that means a lot of the legwork is already done and most of the tweaks will be in the form of optimization and bug fixes.Well, I know only basic principles of programming but for Lr Classic to run natively doesn't the code need to be completely new? Eg. all calls to any X86 or X64 processes need to be removed or rewritten.
So in layman terms, LrC will be same bloated mess as on Intel Macs?Very few modern programs use actual CPU-specific assembly code anymore. Most use high-level languages (Objective-C, Swift) that make calls to OS-level APIs which can be readily compiled to run on different CPU types. The big headache with the PPC->Intel transition is that most of the major apps were legacy Carbon apps that came from Classic Mac OS and were barely updated to run on Mac OS X. Most of these legacy apps have actually already been rewritten using modern Cocoa APIs since then - Adobe products were rewritten in Cocoa as part of the CS5 release in 2011 and Office was rewritten in Cocoa as part of Office 2016. Cocoa-based apps are easier to port to other architectures, so that means a lot of the legwork is already done and most of the tweaks will be in the form of optimization and bug fixes.
One possibility is that Apple has underclocked the M1 Mini, given its actual thermal capabilities.So it looks like all three M1’s run at the same speed? (At least until the Air gets too hot?) So there is no speed bump from the Mini having 120v power supply?
as a guy who knows nothing,I am very curious about the higher end processors. More cores? What configuration? 4 ice + 8 fire? GPU with more cores? More cache?
Higher clock frequency is unlikely, as it already is very high. The memory bus is already very fast.
I guess Apple can still do something more — but what? Keeping thermal limitations in mind.
Considering how much they're currently emphasizing unified memory... I doubt it'll be anything quite like that. Programming for an environment where different chunks of main memory are more or less accessible based on which core they're attached to is frustrating, which is part of why AMD's been moving away from that to locating all the memory channels on the uncore as I recall.as a guy who knows nothing,
I am guessing they include like 4 M1 CPU in Pros, each dedicating a full power to a specific task givibg a results of: the whole is more than the sum of its parts
It's called AWS GravitonI'd love to see this tech in server rooms. Being able to have far more computing power for the same power and cooling load would be amazing.
Yea I guess you are right. I just don't see how the pros will be faster since the M1 are already at 5nm. What can they do? Now I am guessing higher clock-speeds that will generate more heat but in PROs will have better cooling?Considering how much they're currently emphasizing unified memory... I doubt it'll be anything quite like that. Programming for an environment where different chunks of main memory are more or less accessible based on which core they're attached to is frustrating, which is part of why AMD's been moving away from that to locating all the memory channels on the uncore as I recall.
TIL: Amazon builds their own CPUs. Next Microsoft. RIP Intel.It's called AWS Graviton