Linus Torvalds (torvalds@linux-foundation.org) on 1/12/11 wrote:
---------------------------
>Max (max@a.com) on 1/12/11 wrote:
>>
>>I remember well Linus's "crazy" comment a few years that
>>if anything can challenge x86, it's ARM. (Not Itanium :-))
>
>So I've obviously not changed my mind about the whole
>"things grow up (rather than down)" model, and I will
>absolutely still pick ARM over Itanium as a future model.
>
>But I do have to say that people often seem to overlook
>the serious problems that ARM has. We have ARM nutcases
>(you know who you are) who think that ARM is a cleaner
>architecture than x86, despite all the evidence to the
>contrary.

You keep making these false claims but that doesn't make them any more true no matter how much you want to. ARM obviously has far less baggage than x86 (some of the original baggage such as FPA, 26-bit mode has even been removed).

>And they are easy to overlook, because almost all ARM uses
>are very much one-off. Users have very little view into the
>problems that arise from the fragmentation of the whole
>ARM ecosystem, because they almost never see them. So the
>lack of compatibility between different ARM cores isn't
>seen as the problem that it in reality would be if you
>really chased the x86 market.

There is full compatibility between the different ARM cores, you can run a binary compiled for one architecture on any other CPU of the same architecture or higher. And that is something that cannot ever be done on x86 - there will never be 2 x86 cores with the same instruction set (Intel and AMD support different non-overlapping ISAs). ARM on the other hand requires all ARM cores to be compatible, even if you design your own ARM core.

>And people don't understand how big the performance gap
>really is. ARM is absolutely not even close. ARM right now
>is struggling to compete with Atom, for chissake! Your
>high-end phone may sometimes feel "more responsive than
>your desktop", and people think ARM performs well, but
>that is entirely due to the problem size.

Actually the performance gap between Atom and ARM is widening quickly as Atom hasn't improved at all in the last 3 years...

Even the ancient Cortex-A8 can keep up with Atom clock for clock, while the out-of-order Cortex-A9 beats it by a big margin. There was an announcement of a 2GHz dual core A9 laptop, and several 1GHz dual core phones are about to be shipped.

Unless Intel does something radical to improve Atom's performance soon, Atom is simply history. And next year there's Cortex-A15 to compete against.

>So I absolutely believe that the only realistic threat to
>x86 is ARM, but at the same time I do think there's a very
>strong tendency for people to be more aware of the problems
>they have with traditional PC's, and unaware of the
>problems with the "new solution", because they simply have
>not had a lot of exposure to it.

I think you're mistaken to think people don't know ARM, there are hundreds of thousands of ARM developers. Just about every software engineer I meet knows all about ARM. Sure things are different in between architectures so you encounter different issues, but ARM is vastly simpler and cleaner compared to x86.

>And as mentioned, the people who have had a lot of
>exposure to ARM are embedded people who think that a
>product is a one-off event, rather than an ecosystem that
>spans several decades. And those kinds of people like
>being able to start from a clean slate, and think that
>long-term support is a bad thing.

Again, you're totally wrong. Just about all the mobile OSes that run on ARM run on many different ARM cores, have a well defined ABI/API and even allow users to write their own apps which run on multiple devices.

IIRC the only area where there was a lot of resistance and feet dragging from some people against using a standardized ARM ABI which would work for decades into the future was... Linux.

Wilco