Subject: Re: [IP] more on IBM VC calls for 'open' hardware
From: DV Henkel-Wallace <gumby@henkel-wallace.org>
Date: Mon, 11 Apr 2005 17:19:44 -0700

 Mon, 11 Apr 2005 17:19:44 -0700
> From: Simon Higgs <simon@higgs.com>
> Date: Mon, 11 Apr 2005 22:17:51 +0100
>
> Isn't this what Gil Amelio tried to do with the Apple platform before 
> Steve
> Jobs returned?

This is not at all what the licensed Mac model was.  Gil Amelio was 
attempting to recreate the success of the PC model that was sparked by 
IBM's publishing the PC documentation.  However the trigger for that 
market explosion was Phoenix's reverse-engineering IBM's proprietary 
BIOS.  Amelio's effort was closer to IBM's Microchannel effort (a 
failed attempt to push the horse back into the barn).

What Carballo's talking about now is at a completely different level of 
abstraction.  SoC (System on a Chip) is the much-vaunted 
application-driven integration of various functional elements onto a 
single die.  "Integration" is generally taken to mean CPU, pager, IO 
controller, dedicated logic (e.g. MPEG decoder or crypto accelerator), 
ethernet controller etc and even analog circuits (e.g. "mobile phone on 
a chip").  The general idea is that the OEM can source "parts" at the 
VHDL level as they now source ICs, and that new businesses could arise 
licensing logic for these systems.  The drivers are 1> engineering sex 
appeal, 2> lower COGS, and 3> increased reliability, in that order.

People have been trying to make this work without a huge amount of 
success for 10-15 years.  Search for "hardware-software codesign" on 
the search engine of your choice or look at http://www.vsi.org/ -- in 
fact I just went there and look at the first seminar they announce: 
"Survival of the Fittest. Will Small IP Makers Make it?".  You can also 
see companies like iReady whose original business plan was to be one of 
these suppliers.  In fact I founded a company which at one time thought 
it would make money as this market developed (it made money other ways 
instead).

There are two fatal problems for this model today.  The first is that 
the logic-level abstractions are just not as clear as the ones 
available in straight software.  The second is that by trying to 
integrate your software flow with your hardware flow you increase the 
complexity, not decrease it.  To make that worse the software people 
and hardware people speak different languages -- not just VHDL vs C++, 
but timing driven vs algorithm-driven.  This latter problem IMHO also 
makes life extremely hard for companies that try to bridge just that 
software gap (e.g. CoWare).

The ARM guys have had some success.  The has generally only been 
possible when there are huge enough volumes at stake (basically the 
mobile phone business) to justify design, but even then the resulting 
system is not as specialised as the SoC people dream of.  They don't 
support an ecosystem of 3rd-party IP like you see in the back of Dr 
Dobbs, and I doubt they ever will.


Carballo is making a different point, and I think it's an interesting 
one.  Software has gone through various stages of abstraction 
consolidation.  Into the 1980s\s, we all wrote our own programming 
languages, but that consolidated down to a few.  Still lots of people 
wrote their own O/S, especially embedded guys.  What Linux did was 
consolidate the OS abstraction layer as well.

But  how  it did it was even more interesting.  Since Linux is open, 
people fix it.  It wasn't really that good, but it was close enough to 
good enough that it was worth improving.  And because of how it was 
licensed, many of those improvements became "Linux" too (or "Linux" 
became those improvements).  Now it's quite good.  By contrast C never 
really was quite "good enough" in the same way, and improving that at 
all took quite a long time.

And "despite" being free software (actually  because  it is) , tons of 
businesses both small and not-so-small have been built around it.

So the interesting points in Carballo's talk are that 1> if someone can 
crack the component abstraction model SoC might actually take off and 
2> any solution to problem 1 will most likely follow Linux's example -- 
just as Linux and the Linux market built upon many predecessors as 
well.

Sorry for such a long answer to your brief question.

David Henkel-Wallace


>> Date: Sat, 09 Apr 2005 21:16:51 -0700
>> From: Dewayne Hendricks <dewayne@warpspeed.com>
>>
>> IBM VC calls for 'open' hardware
>>
>> Richard Goering
>>   (04/08/2005 12:53 AM EDT)
>>   URL:  <http://www.eetimes.com/showArticle.jhtml?articleID=160502705>
>>
>>   MONTEREY, Calif.  A revolution is coming to the electronics 
>> industry,
>> says Juan-Antonio Carballo, partner in IBM's venture capital group 
>> and it's based on "open" hardware that will foster the kind of
>> collaborative effort that Linux has brought to software.
>>
>>   Carballo presented his argument before a somewhat skeptical audience
>> at the Electronic Design Processes (EDP) workshop Thursday (April 7).
>> But Carballo insisted that various kinds of "openness," including
>> collaborative efforts and open-source models, will change the world of
>> system-on-chip (SoC) venture financing.
>>
>>   "There's a pretty important wave coming in our industry that most of
>> us are not noticing, and it could change it completely," said 
>> Carballo,
>> who happens to be an avid surfer. "The open-source model is quickly
>> extending from software to hardware, and it will provide a similar
>> swell of collaborative innovation." [...]