5 ms·

> > what if Apple outperforms both Intel and AMD with their Apple Silicon? > It wouldn't matter at all to either AMD or Intel until Apple starts supplying thes

> > what if Apple outperforms both Intel and AMD with their Apple Silicon? > It wouldn't matter at all to either AMD or Intel until Apple starts supplying these chips to other OEM's. Till then, it will just be used in the walled garden of Apple products, which is the very thing that makes it very unattractive for many users. Well, there is another possible path: if Apple Silicon is seen as a PoC, someone else could decide to pursue the server market with an Apple Silicon inspired ARM-based offering. Perhaps aimed at FaaS deployments or something similar.

wmf 6y ago
someone else could decide to pursue the server market with an Apple Silicon inspired ARM-based offering In other words Nuvia.
_ph_ 6y ago
There are already ARM based servers available in AWS. The big problem is: almost no one has desktop machines, where the software development happens, with ARM. Linus once commented that ARM doesn't sell so well on the server, because developers are lacking desktop machines with that architectore. Apple Silicon could close that gap. You can develop with a Linux VM on AS, and deploy on ARM in the cloud.
- pjmlp 6y ago
  Around the first .com wave, I used to develop on Windows (C and C++) and deploy on SPARC/PowerPC/PA-RISC servers. The only x86 servers were running a mix of NT 4.0 and Windows 2000. Apparently it worked.
  - _ph_ 6y ago
    Of course it works. Otherwise there would be no ARM on the servers. But it is second class to a setup, where the development and deployment happens on the same platform. Having good ARM machines available on the desktop will give ARM on the server a boost. As I wrote, don't take my word for it, listen to what Linus hat to say on that topic.
    - pjmlp 6y ago
      Ironically UNIX is exactly one kind of platform where remote development is first class. One just needs to properly configure mount points and remote sessions. There was hardly any difference between my X sessions and local development.
      - _ph_ 6y ago
        Sorry, in general that is not true. You need a very good network connection, both with bandwidth and latency, to make remote devlopment workable. Still, it never equals local development. If you follow the discussions here on hacker news, which terminal software has the smalles latency, remote development never can compete with that.
        pjmlp 6y ago
        Apparently having UNIX servers on premises is a forgotten art.
        _ph_ 6y ago
        It is. Even if companies have their own hardware, it is often in separate compute centers. And of course anything on AWS or Azure is not on premise.
        pjmlp 6y ago
        Not everyone is FAANG, or pretends to be one.
Iwan-Zotow 6y ago
Nobody cares about server market, literally if it runs on abacus fast enough it's ok. Developers! DEvelopers! Developers! ARM should be in consumer space, millions of laptops for developers first running Windows and Linux
- pjmlp 6y ago
  With the cloud and managed runtimes (or stuff like POSIX), I don't care one second what CPUs are running there (beyond endiness and using the right widths), I never wrote hand Assembly for such kind of deployments and the optimizer was good enough for our workloads. The only issue would be libraries available only in binary form for AOT compiled languages.

wmf 6y ago

someone else could decide to pursue the server market with an Apple Silicon inspired ARM-based offering In other words Nuvia.

_ph_ 6y ago

There are already ARM based servers available in AWS. The big problem is: almost no one has desktop machines, where the software development happens, with ARM. Linus once commented that ARM doesn't sell so well on the server, because developers are lacking desktop machines with that architectore. Apple Silicon could close that gap. You can develop with a Linux VM on AS, and deploy on ARM in the cloud.

pjmlp 6y ago
Around the first .com wave, I used to develop on Windows (C and C++) and deploy on SPARC/PowerPC/PA-RISC servers. The only x86 servers were running a mix of NT 4.0 and Windows 2000. Apparently it worked.
- _ph_ 6y ago
  Of course it works. Otherwise there would be no ARM on the servers. But it is second class to a setup, where the development and deployment happens on the same platform. Having good ARM machines available on the desktop will give ARM on the server a boost. As I wrote, don't take my word for it, listen to what Linus hat to say on that topic.
  - pjmlp 6y ago
    Ironically UNIX is exactly one kind of platform where remote development is first class. One just needs to properly configure mount points and remote sessions. There was hardly any difference between my X sessions and local development.
    - _ph_ 6y ago
      Sorry, in general that is not true. You need a very good network connection, both with bandwidth and latency, to make remote devlopment workable. Still, it never equals local development. If you follow the discussions here on hacker news, which terminal software has the smalles latency, remote development never can compete with that.
      - pjmlp 6y ago
        Apparently having UNIX servers on premises is a forgotten art.
        _ph_ 6y ago
        It is. Even if companies have their own hardware, it is often in separate compute centers. And of course anything on AWS or Azure is not on premise.
        pjmlp 6y ago
        Not everyone is FAANG, or pretends to be one.

Iwan-Zotow 6y ago

Nobody cares about server market, literally if it runs on abacus fast enough it's ok. Developers! DEvelopers! Developers! ARM should be in consumer space, millions of laptops for developers first running Windows and Linux

pjmlp 6y ago
With the cloud and managed runtimes (or stuff like POSIX), I don't care one second what CPUs are running there (beyond endiness and using the right widths), I never wrote hand Assembly for such kind of deployments and the optimizer was good enough for our workloads. The only issue would be libraries available only in binary form for AOT compiled languages.