[threeseed]

> This is a market failure more than an Apple failure

Arguably it's a Microsoft failure more than anything.

If they had delivered a version of Windows for ARM capable of seamlessly emulating x86 code with no compromises and released it a decade ago the world would likely be different.

[aleph_minus_one]

> If they had delivered a version of Windows for ARM capable of seamlessly emulating x86 code with no compromises

This is hardly possible on most ARM chips because x86 has a much stronger memory model than ARM. Also concerning "with no compromises": common x86 implementations have a very fast implementation of the SIMD instruction sets (SSE..., AVX/AVX2, perhaps AVX-512) that is much slower to emulate on ARM because their SIMD instruction set is different. The only reason why people don't realize this is that a lot of common software makes no intense use of these SIMD instructions. Then there are the subtle parts that (as far as I am aware) the ARM FPU handles multiplication of denormalized floating-point numbers slightly different than x86 (both implementations are allowed by the standard) etc.

[rob74]

Well, Apple seems to have achieved this (for all intents or purposes) on MacOS (https://en.wikipedia.org/wiki/Rosetta_(software)#Rosetta_2 - "In some benchmarks, x86-64-only programs performed better under Rosetta 2 on a Mac with an Apple M1 SOC than natively on a Mac with an Intel x86-64 processor"), so there's no reason why Microsoft shouldn't be able to do the same for Windows? Of course, Apple designed the M series ARM CPUs specifically for their devices, but I'm not sure how much care went into optimizing x86 emulation?

[happymellon]

The M1 was significantly faster than other ARM chips.

Why? My impression is that due to Apples permissive licence, they are able to make more changes. If we just look at the mess that ARM is making of Qualcomm situation, where we finally have a performing alternative chip.

Arm hasn't granted anyone else the ability to make significantly modified chips for multiple platforms, so Nuvia were fine while it was all just research.

> Arm's claim against Qualcomm and Nuvia is about protecting the Arm ecosystem and partners who rely on our IP and innovative designs, and therefore enforcing Qualcomm's contractual obligation to destroy and stop using the Nuvia designs that were derived from Arm technology

Its rather telling that Qualcomm are allowed to make mobile chips, but not these ones because they want to "protect the ARM ecosystem".

RISC V can't come soon enough.

[rickdeckard]

  > Arm hasn't granted anyone else the ability to make significantly modified chips for multiple platforms,

Of course they have. Many of the licenses are TLA (Technology License Agreements) but ARM has also made ALA (Architecture License Agreements) with several companies, apart from Apple also with Qualcomm, nVidia, Broadcom, Samsung and many others.

  > so Nuvia were fine while it was all just research.

Nuvia wasn't just "research", they got a very permissive license from ARM to create ARM-based server-architecture and build a business with that, with the licensing contract explicitly limiting their IP to use for servers and only to Nuvia.

Qualcomm acquired Nuvia with the clear plan of using their IP for other use-cases ("powering flagship smartphones, next-generation laptops, and digital cockpits, as well as Advanced Driver Assistance Systems, extended reality and infrastructure networking solutions"). Apart from the fact that Nuvia had no license for these markets, the acquisition voided the license they had and the contract states that without license all the IP developed under the license has to be destroyed.

There is no dispute on this, Qualcomm confirmed this in court, but they argue that they have ANOTHER license they want to transfer the IP to, and that the terms of the ARM-Nuvia Agreement are offending and "a threat to the industry"

  > Its rather telling that Qualcomm are allowed to make mobile chips, but not these ones because they want to "protect the ARM ecosystem".

Because Qualcomm is in progress to fragment the ARM-ecosystem, by using its dominant position in Smartphone chipsets to establish its custom architecture as the new standard for other industries. For decades, ARM is carefully avoiding this to happen, by allowing selected partners to "explore" evolutions of the IP in a industry but with methods to make sure they can't diverge too much from ARM's instruction set.

ARM has designed new architectures (Blackhawk, Cortex-X) which achieve comparable performance to Nuvia's IP, but Qualcomm's assumption is that they can apply Nuvia's IP on top of their existing architecture without the need of licensing any new ARM design.

[voidfunc]

Why would Microsoft build something almost no customers were asking for?

Even now ARM is niche.

[h0l0cube]

I'll assume you mean niche in the desktop space.

So there's a catch 22 here. ARM was niche on desktop because no-one made the first move. That is, until Apple launched the M1, which is exactly the failure of imagination of Microsoft that GP is talking about. Though a quick google shows they did indeed investigate this avenue in 2016 for Windows 10, though they obviously didn't commit to it:

> On December 7, 2016, Microsoft announced that, as part of a partnership with Qualcomm, it planned to introduce support for running Win32 software on ARM architecture with a 32-bit x86 processor emulator, in 2017. Terry Myerson stated that this move would enable the production of Qualcomm Snapdragon-based Windows devices with cellular connectivity and improved power efficiency over Intel-compatible devices, and still capable of running the majority of existing Windows software (unlike the previous Windows RT, which was restricted to Windows Store apps). Microsoft is initially targeting this project towards laptops.

https://en.wikipedia.org/wiki/Windows_10#Features_in_develop...

[teknologist]

To show them that their battery life could go from 2.5 hours to 8 or more, perhaps. That was the Mac experience that customers weren't asking for yet "wowed" them when it came.

[antifa]

> customers weren't asking for

they definitely were asking for longer battery life and cooling systems less noisy than a fighter jet, but the supply side of the market was a desert filled with minefields and duds until the M1.

[aleph_minus_one]

> To show them that their battery life could go from 2.5 hours to 8 or more, perhaps.

x86 already was on its way towards such a goal.

[threeseed]

Years later.

And it took the threat of the PC industry moving to ARM to light a fire under Intel.

[teknologist]

x86 is an overcomplicated behemoth never intended for low power computing. It’s lipstick on a pig for mobile x86 at this point

[franga2000]

ARM tends to have better performance per watt, leading to improved battery life and thinner/lighter devices at the same performance and price point. Are you sure nobody is asking for it?

[jocaal]

Correction, ARM chips tend to be designed for better performance per watt. The instruction set has nothing to do with it. For a long time intel and amd neglected the mobile markets. That is about to change.

[everettp]

What? Per Wikipedia: "With over 230 billion ARM chips produced, as of 2022, ARM is the most widely used family of instruction set architectures."[1] [1] https://en.wikipedia.org/wiki/ARM_architecture_family

[PetitPrince]

Niche in the laptop and desktop caategory I assume?

[roshankhan28]

i dont really agree with the microsoft failure part, could explain why do you have this point of view? would love to hear it :)