4 ms·

I keep coming back to the implications for the rest of the industry: 1. What does this mean for x86? 2. What does the answer to question #1 mean for Windows a

I keep coming back to the implications for the rest of the industry: 1. What does this mean for x86? 2. What does the answer to question #1 mean for Windows and Microsoft? Are they going to sit by and let Apple lap them? 3. What does the answer to question #2 mean for cloud computing? AWS is already making ARM chips... does the M1 change anything fundamentally about where the cloud goes with ARM?

pishpash 6y ago
#1: The writing's on the wall for x86, probably. Once its former monopoly rent is only justified by high-end "scientific computing" uses -- well you know what happened to IBM/DEC/SGI/HP and their workstation solutions...
gtirloni 6y ago
At least for #2, see Windows ARM https://docs.microsoft.com/en-us/windows/arm/ https://docs.microsoft.com/en-us/windows/arm/
__d 6y ago
1. x86 (and then AMD64) has had vast amounts of R&D money and talent thrown at it for decades, with the result that it has out-performed every competitor: SPARC, MIPS, Alpha, PA-RISC, Itanium, etc. As a consequence of this domination, it has a massive software, tooling, and expertise advantage. But ... ARM, which through a series of coincidences, ended up being the right architecture at the right time for mobile devices, has had its own injection of talent and money, and has advanced rapidly in the last 15 years. The M1 is a long way from the ARM710 or StrongARM of the late 90's! Similarly, computer circuitry has evolved. Early minicomputers used small circuit boards with just a few gates to assemble their CPUs. In the early 80's, bigger CPUs were built from multiple "bit slice" chips ganged together. The modern PC motherboard has been a thing since the era of the Apple ][, growing through 8, 16, 32 and now 64 CPUs with ever-increasing speed and power, but maintaining a model of a motherboard and expansion cards. SoCs have been a thing for embedded devices for some time now: the M1 is perhaps the first really high-profile "computer" to use this technique to integrate its major architectural components in one "chip" (package). I think it's inevitable that others will follow, given the obvious performance and power gains to be had. AMD and Intel do this to some extend already. The AMD chips powering the new Platstation and Xbox consoles are SoCs, and Intel has sold embedded variants of its CPUs that are basically SoCs for some years now too. x86 has a long future ahead of it, but for the first time since Alpha (?) it's seeing a major competitor. Unlike DEC though, ARM has a broad and deep ecosystem behind it, and (as the article points out) there are some advantages to the instruction set that x86-64 will perhaps struggle to overcome. 2. Microsoft already has the Surface. They've known for some time that Apple's integration of hardware and software is hard to beat, and that the iPhone represented a technology that could be applied to laptops and desktops any time Apple wanted to do so. So ... they've already made steps in the direction of competing, but they have a lot of business-model challenges to overcome in order to do so. Honestly, I'd be much more interested in the futures of Dell, HP(E), Lenovo, etc, than Microsoft. This is a massive challenge for them. 3. No. The cloud vendors are already moving towards ARM for power reasons. The AWS Nitro board is begging to be integrated with a Gravitron -- I would be surprised if Amazon doesn't come up with a compute platform that's powered by a similarly integrated SoC in the very near future. The M1 helps accelerate this process because all Open Source software will now have to treat AARCH64 as a first-class citizen (which, to be fair, the Raspberry Pi has already helped to do). Similarly, commercial software vendors now have another market that will need a native ARM version in 2 years (after Rosetta2 is phased out, like Rosetta was for PPC). So .. I don't think it's changed the cloud direction at all, but it could make the transition faster.
kristianp 6y ago
Intel is going to be strongly motivated to pull their finger out and match the performance of the M1. Increase the performance of the chip, and release SOCs with RAM on the package for laptops, not just for phones.

pishpash 6y ago

#1: The writing's on the wall for x86, probably. Once its former monopoly rent is only justified by high-end "scientific computing" uses -- well you know what happened to IBM/DEC/SGI/HP and their workstation solutions...

gtirloni 6y ago

At least for #2, see Windows ARM https://docs.microsoft.com/en-us/windows/arm/ https://docs.microsoft.com/en-us/windows/arm/

__d 6y ago

1. x86 (and then AMD64) has had vast amounts of R&D money and talent thrown at it for decades, with the result that it has out-performed every competitor: SPARC, MIPS, Alpha, PA-RISC, Itanium, etc. As a consequence of this domination, it has a massive software, tooling, and expertise advantage. But ... ARM, which through a series of coincidences, ended up being the right architecture at the right time for mobile devices, has had its own injection of talent and money, and has advanced rapidly in the last 15 years. The M1 is a long way from the ARM710 or StrongARM of the late 90's! Similarly, computer circuitry has evolved. Early minicomputers used small circuit boards with just a few gates to assemble their CPUs. In the early 80's, bigger CPUs were built from multiple "bit slice" chips ganged together. The modern PC motherboard has been a thing since the era of the Apple ][, growing through 8, 16, 32 and now 64 CPUs with ever-increasing speed and power, but maintaining a model of a motherboard and expansion cards. SoCs have been a thing for embedded devices for some time now: the M1 is perhaps the first really high-profile "computer" to use this technique to integrate its major architectural components in one "chip" (package). I think it's inevitable that others will follow, given the obvious performance and power gains to be had. AMD and Intel do this to some extend already. The AMD chips powering the new Platstation and Xbox consoles are SoCs, and Intel has sold embedded variants of its CPUs that are basically SoCs for some years now too. x86 has a long future ahead of it, but for the first time since Alpha (?) it's seeing a major competitor. Unlike DEC though, ARM has a broad and deep ecosystem behind it, and (as the article points out) there are some advantages to the instruction set that x86-64 will perhaps struggle to overcome. 2. Microsoft already has the Surface. They've known for some time that Apple's integration of hardware and software is hard to beat, and that the iPhone represented a technology that could be applied to laptops and desktops any time Apple wanted to do so. So ... they've already made steps in the direction of competing, but they have a lot of business-model challenges to overcome in order to do so. Honestly, I'd be much more interested in the futures of Dell, HP(E), Lenovo, etc, than Microsoft. This is a massive challenge for them. 3. No. The cloud vendors are already moving towards ARM for power reasons. The AWS Nitro board is begging to be integrated with a Gravitron -- I would be surprised if Amazon doesn't come up with a compute platform that's powered by a similarly integrated SoC in the very near future. The M1 helps accelerate this process because all Open Source software will now have to treat AARCH64 as a first-class citizen (which, to be fair, the Raspberry Pi has already helped to do). Similarly, commercial software vendors now have another market that will need a native ARM version in 2 years (after Rosetta2 is phased out, like Rosetta was for PPC). So .. I don't think it's changed the cloud direction at all, but it could make the transition faster.

kristianp 6y ago

Intel is going to be strongly motivated to pull their finger out and match the performance of the M1. Increase the performance of the chip, and release SOCs with RAM on the package for laptops, not just for phones.