Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 09 Jul 1999 05:46:10 -0400
From:      "Christopher R. Bowman" <crb@ChrisBowman.com>
To:        Brett Glass <brett@lariat.org>
Cc:        freebsd-chat@FreeBSD.ORG
Subject:   Re: IA64
Message-ID:  <199907090946.EAA07069@quark.ChrisBowman.com>
In-Reply-To: <4.2.0.58.19990708092030.0466ca00@localhost>
References:  <Pine.GSO.3.95-heb-2.07.990708172513.21435A-100000@csd> <4.2.0.56.19990708071909.03f9cda0@localhost>

next in thread | previous in thread | raw e-mail | index | archive | help
At 09:27 AM 7/8/99 -0600, Brett Glass wrote:
>At 05:52 PM 7/8/99 +0300, Nadav Eiron wrote:
>
>
>>On Thu, 8 Jul 1999, Brett Glass wrote:
>>
>> > 
>> > Not so. Intel sold it as an embedded design, and *actually refused* to
sell
>> > it to any manufacturer who wished to use it as a general purpose CPU on
>> > any significant scale. Intel allowed MicroWay to build boards which used
>> > one or more i860s in combination with an x86 in the same machine, but they

>> > actively discouraged the development of any i860 products that would 
>> > cannibalize their precious x86 monopoly. The parts just wouldn't be
available.
>> > 
>> > I attended an Intel presentation at the time when the i860 was released
>> > which emphasized this.
>>
>>There's little info on the i860 on the web (it was discontinued too long
>>ago), and I don't have its architecture reference around (library closed 
>>early today for some reason) but, hey, see this quote from
>>this URL (remember, it's Intel, from the overview of MMX technology):
>>
>>http://developer.intel.com/technology/itj/q31997/articles/art_2b.htm
>>
>><quote>The definition of MMX technology evolved from earlier work in the
>>i860? architecture [3].  The i860 architecture was the industry's first
>>general purpose processor to provide support for graphics rendering. 
>></quote>
>
>Funny how Intel rewrites history to avoid antitrust problems. ;-) Fact is,
>they wanted the i860 to be used on graphics coprocessor boards but NOT
>as the system's main CPU. I may still have the handouts which said so.
>At the presentation, I asked an Intel rep whether the i860 could be used
>as a general purpose CPU for a workstation; he replied that Intel WOULD
>NOT SELL the CPU to a company that wanted to use it for that purpose.

Um, Brett, I can't debate with you what you may have on handouts or heard at
presentations.  However, in an odd coincidence I was just throwing out some old
Intel duplicates and marketing stuff I happen to have here, and what you
present above is diametrically opposed to the impression I get from that
material.  The 1992 Intel Product Overview book (section 4) clearly tags that
i960s as targeted for embedded systems, but makes no such mention for the i860s
and in fact instead mentions engineering workstations, scientific computing, 3d
workstations and multi-user systems.  Further the Overview of the i860XP
Supercomputing Microprocessor, and the  Overview of the i860 64-Bit
Microprocessor, Chapter 1 state:  the i860 CPU ... is designed to bring
supercomputing power to the desktop.  And later: This high performance can be
applied across a wide variety of compute intensive design implementations, from
supercomputers, through technical workstation, to graphics subsystems and
application accelerators for personal computer.  Also: The Intel i860 CPUs are
a family of supercomputing microprocessors designed to provide affordable
supercomputing class power to a range of systems from PC's to workstations to
massively parallel supercomputers.

Further Intel published a document called the i860 Platform Reference Design
Guide which was basically a reference workstation design containing "complete
hardware theory of operation descriptions, schematics, PAL equations, timing
and state diagrams and data sheets for the system's major components.

I get the distinct impression that the i860s were targeted at high end
engineering workstation type environment.  It is not clear if this was to
prevent cannibalization of the commodity x86 market or because these machines
would presumably have prices that would have kept them out of the lower end. 
Clearly, however, Intel was marketing the i860s at general
computing/workstations and the i960s at embedded design.

>This is illegal behavior, since it is discriminatory and an attempt to
>manipulate markets. But it was done nonetheless.
>
>>i860 - a general purpose processor. Let us check further, for instance:
>>http://infopad.eecs.berkeley.edu/CIC/
>>
>>The same... (anyway, graphics rendering on an embedded controller really
>>makes little sense).
>
>You're forgetting history. At this time, the TMS340 intelligent graphics
>controller was threatening to take over this market. (There were many very
>good one-card X servers based on that chip.) The i860 was part of Intel's
>response.

Ah the TMS340, I still have their data books here too.  As I remember, general
consensus at the time seemed to be that this chip was hot shit, and so I
eagerly ordered my data book.  I have to say upon reading it that I was not to
impressed.

>>Also, a search for i860 on Intel's site reveals some references to
>>massively parallel machines based on the i860, used for number crunching
>>(just like the one they have here in the department, which was not made by
>>Intel, or Microway). The reason most of these machines combine the i860
>>with other CPUs is that the i860 is notoriously bad at handling interrupts
>>and I/O, which meant that its main use was for number crunching - like a
>>sophisticated DSP perhaps, with something else handling I/O. However, it
>>was unsuccessful even in that niche. 
>
>The i860 was used in a number of demonstration machines that did
>floating point multiprocessing, because its floating point capabilities
>exceeded what could be done on any x86 processor available at the time.
>But this wasn't considered by Intel to be a "general-purpose" market, since
>these machines were produced in small volumes and were ludicrously expensive.
>
>--Brett
>
>
>
>To Unsubscribe: send mail to majordomo@FreeBSD.org
>with "unsubscribe freebsd-chat" in the body of the message
>  
--------
Christopher R. Bowman
crb@ChrisBowman.com
http://www.ChrisBowman.com/


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-chat" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199907090946.EAA07069>