LWN: Comments on "Intel CEO: Smaller gadgets will expand market (AP)"

why x86? community!

undefined — Wed, 04 Jun 2008 00:29:48 +0000

why x86? it's a "standard" and "open" platform on which to build a sustainable community.

i have a nokia 770, which is powered by an an arm processor & TI DSP, but the platform is
essentially dead. nokia and the "community" moved on to the n800 and n810. but the hardware
is so non-standard that no mainstream distribution supports it (unlike the ubiquitous x86
desktop or laptop). actually, i don't know of any "distribution" that supports it at all
besides nokia's abandoned IT OS 2006 (maemo 2.2 "gregale").

the x86 "platform" is popular enough that it's well known (from hardware to BIOS/EFI/firmware)
and well supported (from kernel, up through applications to distros).

point out another "platform" that has such on-going, widespread, linux community support,
except possibly for wireless access points (dd-wrt, openwrt, tomato, etc), but they are fairly
standardized across manufacturers and products (arm, flash, nvram, ethernet & wireless
chipsets).

maybe the 770 is a bad example because the hardware was too unique and this market segment is
too new (small but not a laptop form-factor). the EeePC seems to have done well
community-wise, but probably because it's more "mainstream" (though i don't think it would
have made much of a difference whether it was x86 or arm). but similarly no linux PDA has
done well (though i'm using FOSS apps on my 2007 palm device that were released in 2001, so
maybe software backwards compatibility has something to do with it).

okay, enough of my griping.

software is tested on x86, and may not work on other architectures

drag — Sun, 01 Jun 2008 09:20:52 +0000

Oh.. in terms of embedded development ARM surpassed x86 in popularity about 3 years ago.
(according to linuxdevices.com) 

According to Debian's popularitycontest Arm is now 3rd favorite behind x86 and x86-64. It beat
out the Apple-Linux crowd..

When it comes to fairly popular architectures like the ARM there really isn't going to be much
of a barrier for any open source application if the designers remotely cared about
portability. (If they didn't care much about portability then generally they didn't care much
about quality either, anyways...)

software is tested on x86, and may not work on other architectures

drag — Sun, 01 Jun 2008 09:06:08 +0000

> yes ARM is probably the next architecture to support, but that doesn't mean that the various
developers are going to take the time to do so.

Your right. It's the users that make it work. Always has, always will. The only reason why any
free software isn't supported on a platform is because not many users use that platform.

I've used PowerPC Debian in the past and I very rarely ran into any software that didn't work
that wasn't closed source. I even shared the same home directory between my PowerPC laptop and
my x86 desktop and that worked well (except for certain programs that used sqlite in a
non-endianess safe manner)

Hell during that period time my Ibook was better supported then any x86 laptop that I ever
used.. before Intel started helping out with getting Linux centrino support up to speed.

I figure as long as you have decent Linux kernel support for a platform, and the GNU stuff
supports a platform then most open source software should be mostly just a compile away. The
most important part is just having users to compile and test the software. Without those there
isn't any point anyways.

software is tested on x86, and may not work on other architectures

dlang — Sun, 01 Jun 2008 08:10:45 +0000

yes ARM is probably the next architecture to support, but that doesn't mean that the various
developers are going to take the time to do so.

the advantage of x86 is compatibility, it doesn't matter if it's closed source or open source,
it's almost certainly tested on x86 and it may or may not work with other architectures.

this isn't limited to compiled code either. I've run into perl code that was written on AIX
(powerpc) that didn't work on Linux without needing to go in and write compatibility routines
to handle the big/little endian difference.

why use x86 other than to support proprietary software?

drag — Sun, 01 Jun 2008 04:58:47 +0000

Probably ARM platform is going to be the most popular after x86-64/x86 platforms.

Cell phones, routers, NAS devices, etc etc. All those run Linux on ARM platforms. Most of them
are getting to the point were they can run Debian proper.. pretty much anything with the
ability to use flash media, 200mhz proccessor, and 64megs of ram (is what I figure).

Some examples...
KuroBox Pro
http://www.revogear.com/ProductDetails.asp?ProductCode=KU...
400mhz ARM9, 128MB DDR2 RAM, 256MB NAND Flash, USB 2.0 support, SATA support, 1x PCI Express,
1Gb/s LAN.

NeoRunner Phone
http://wiki.openmoko.org/wiki/Neo_FreeRunner
480x640 resolution, 802.11 Wifi, USB OTG, 400mhz ARM9, 128MB SDRAM, 256MB NAND flash.

Beagle Board
http://beagleboard.org/
600mhz proccessor, DVI-D video-out, 128MB ram, 256MB flash, SDHC support, USB OTG, Stereo Out,
Stereo In.

I would _love_ to have a mini-laptop with something like the BeagleBoard in it. The battery
life would be _fantastic_. I could probably last days for the same amount of power that is
required to run my Dell Laptop.

http://www.arm.com/products/CPUs/architecture.html
Those new ARMv7-based CPUs are fantastic.. those Cortex processors are pretty cool stuff. Even
some people are talking about multiprocessor support. They've added 64/128bit SIMD stuff in
the form of the NEON instruction set to vastly boost multimedia support.

Very cool. They seem to be approaching the same level of performance that you'd get with
low-end x86 systems.

Texas Instrument is finally coming around also, releasing massive amounts of documentation and
even are trying to get 2D/3D acceleration chips for emdedded systems to be openned up for
Linux folk. (I think)

This article explains pretty well the sort of costs that are associated with the legacy CISC
system for embedded systems. With large systems it doesn't matter so much as the logic in the
cpu to translate and deal with the x86 CISC instruction set is minor compared to the rest of
the system.. but when you shrink the CPU to embedded sizes those costs start to add up.
http://arstechnica.com/articles/paedia/risc-vs-cisc-mobil...

why use x86 other than to support proprietary software?

dlang — Sun, 01 Jun 2008 03:47:29 +0000

recompiling code for the other CPUs assumes that the code will compile cleanly on those CPUs.
that's not always the case. 

for many projects the maintainers don't have the time, interest, expertise, or hardware to
make their code work cleanly on different systems. they make sure that it works cleanly on
their desktop systems, which are x86 varients (and nowdays, frequently make sure it works on
64 bit versions of x86), but that doesn't say that it will work on other systems.

while it's true that with Open Source code you can go in and fix it, doing so for every
package in the world takes a lot of effort. The Debian folks spend a _lot_ of effort on this,
but they are only up to 20,000 packages, and to maintain them they have now dropped support
for some CPUs down to second-class status.

why use x86 other than to support proprietary software?

zooko — Sat, 31 May 2008 19:13:37 +0000

It seems to me that the only advantage that x86 architecture has over ARM, MIPS, or PowerPC
for small devices is that it can run binaries that you can't recompile, which is probably
because they are proprietary and you can't get the source or are legally in danger if you
recompile it.

This web magainze is a good source of news about small device components which are linux
friendly:

https://linuxdevices.com

I'm hoping that the fact that Free Software can more easily be ported to different CPU
architectures, and that ARM, MIPS, and PowerPC perform better in small devices than do x86 or
amd64, will complement one another and give Free Software a boost in the market for small
devices.

Too bad my sons' OLPCs have x86 architecture, and pay a small cost in wasted power and reduced
performance for it.

It's all about the packaged platforms

drag — Fri, 30 May 2008 15:45:57 +0000

AMD screwed up because they depended on Via and Nvidia.

I haven't seen a competitive motherboard come out of Via for years, in terms of performance.

Nvidia, I think, suffers from strategic short-sightedness and never put a lot of effort into
courting the small-n-cheap PC market.. I mean in terms of laptops you may see their stuff in
the big expensive 'gaming' or 'desktop replacement' style laptops, but not anything that
normal people will want or can afford.

For desktop/server systems AMD's cpus are, frankly, more interesting then Intel. Things like
better memory bandwidth, NUMA support, stuff like Nested Tables for virtualization. But with
Intel I know I can get a good, inexpensive package, that includes open source video drivers,
stability, good SATA controllers, good ethernet, and good power management. (as long as I
stick with a high-quality board maker like Asus) Those things are much more hit-or-miss when
your dealing with AMD for lower-end systems. (think 600-900 dollar PC desktop market)

Hopefully AMD can finish turning ATI over pretty quick so they can start being very
competitive again.

I still have high hopes for ARM systems, too, of course.

Think like this: http://beagleboard.org/ or http://wiki.openmoko.org/wiki/Neo_FreeRunner are
intensely interesting and I am pretty sure is the forerunners for the next generation of
personal computing devices. (PCs in the most pure sense)(of course much more powerful machines
will be desirable for a very large number of good reasons, but I don't see why people
shouldn't be able to carry a terminal and multipurpose computing device in their pockets and
find very good uses for them)

Intel CEO: Smaller gadgets will expand market (AP)

NigelK — Fri, 30 May 2008 09:39:36 +0000

I agree with the quoted comments (yeah, I didn't read the article - not yet, anyway). Those of
us who grew up with computers decades (plural) ago have been spoiled rotten with today's
technology and freely available software. This has made us take things for granted. "Oh look -
higher resolutions, more polygons, better sound, faster machines, higher capacity, yawn..."

The whole concept of small, cheap, and yet entirely capable machines is a new one. No longer
are we looking at more powerful, but at more available, more flexible, more disposable. In
five years time I wouldn't be surprised if the number of small computers per household matches
the number of television sets or radios purely because they become so cheap and capable you
could buy them on impulse without noticably denting the wallet.

On the market side of things, it's also going to be an exciting battle where monopolies have
little chance of succeeding because the cost of entering the market is very low indeed. We've
already seen CPU models and manufacturers be mentioned with the response of "er, who??" Intel
and AMD will have to compete properly again, and the price of paid-for operating systems
running on these things will be knocked down to next to nothing.

Rob Landley (Busybox developer) said it best when he reported that the margins on these
devices are really sucky, but the profits are made on sheer volume.

After years of not being interested in the hardware market beyond keeping track of prices and
capabilities, I'm certainly excited to see where the UMPC market goes from here. With costs
becoming lower, we could be seeing a lot more imagination in this field.

It's all about the packaged platforms

darwish07 — Fri, 30 May 2008 00:01:16 +0000

Last year I've been an intern at Intel for 3 months. I remember very much my mentors saying
(who was also been told by upper management) that the most strategic stuff for Intel was its
'platforms' and each day passes I feel they were damn right. 

The number of platforms done there was great, not only the UMPC and the classmates (where the
CEO of the region has been honored for doing great classmate-pc deals in Libya and Nigeria and
knocking the competition/OLPC - using the honor certificate own words). There were also a lot
of failed platforms that wasn't lucky enough (power efficient platforms for some of the
low-power cities of India, specific machines for Chinese schools, Models done for
ultra-low-power embedded devices, ...).

That have always made me wonder what the hell is AMD doing, watching all of this happens
beginning from the first packaged platform, aka the 2003-2004 Intel Centrino platform.