Diversity for its own sake doesn't seem that compelling: there was previously much more diversity in CPU architectures, and presumably there were valid "evolutionary" reasons for that changing (e.g. network effects WRT software support, economies of scale WRT building foundries).
The network effect around software support to binary architectures is irrelevant (next to, actually) when applied free and open-source software.
An ARM netbook has been proved viable regarding software availability because Linux-based netbooks have already proved themselves viable. The fact they employ x86 processors is a historical accident.
Now we have most of the core DSP functions optimized for NEON SIMD, we can achieve realtime CIF video encoding on a ~400mhz Cortex A8. Now start doing the math...
A Cortex A9 is faster than an A8, per clock.
The chips are planned to scale up to 2Ghz, 5 times faster than ours.
Quadcores are already hitting silicon.
At this rate, there might even be a market for ARM servers in a few years--low-power, high-performance devices designed to minimize cooling and electricity costs in the datacenter.
My startup is about virtualizing OS kernels on these processors.
I am a single founder and a kernel hacker. Single because not many people have enough skills + want to start a startup. I always wonder if there are any real kernel hackers here that might want to get involved.
http://www.l4dev.org
Currently spread around the world as a few engineers. Works well so far. Planning to have U.S. presence - either boston or the valley, and been to the valley for exhibitions.
Making money is via dual-licensing the project. GPLv3 is too restrictive to deploy, so people buy a commercial one.
Skill set needed: Linux Kernel, C, Assembler, Linkers & loaders, Python.
I think this might spawn a whole new category of very-small, energy-efficient laptops primarily used to access electronic mail, world-wide web sites, and those new-fangled instant-messaging services.
It's too tiny. My MIPS-based IBM Workpad z-50 did last 10+ hours on a single battery charge. My current computer is orders of magnitude faster, but it still can't go much longer than 4 hours.
17 comments
[ 4.6 ms ] story [ 50.5 ms ] threadBesides that, this x86 world is utterly boring.
An ARM netbook has been proved viable regarding software availability because Linux-based netbooks have already proved themselves viable. The fact they employ x86 processors is a historical accident.
Whilst this was a definitely trend over time, and there are earlier examples, AMD pretty much nailed it when they bolted x86 onto a RISC core.
http://x264dev.multimedia.cx/?p=142
Now we have most of the core DSP functions optimized for NEON SIMD, we can achieve realtime CIF video encoding on a ~400mhz Cortex A8. Now start doing the math...
A Cortex A9 is faster than an A8, per clock.
The chips are planned to scale up to 2Ghz, 5 times faster than ours.
Quadcores are already hitting silicon.
At this rate, there might even be a market for ARM servers in a few years--low-power, high-performance devices designed to minimize cooling and electricity costs in the datacenter.
I am a single founder and a kernel hacker. Single because not many people have enough skills + want to start a startup. I always wonder if there are any real kernel hackers here that might want to get involved. http://www.l4dev.org
Making money is via dual-licensing the project. GPLv3 is too restrictive to deploy, so people buy a commercial one.
Skill set needed: Linux Kernel, C, Assembler, Linkers & loaders, Python.
<a href="http://the-future.com>The Future</a>--The Future