[comp.arch] i82786 R.I.P. ?

arthur@warwick.UUCP (John Vaudin) (06/22/89)

I read some rumours in a UK magazine that Intel have stopped production
of their i82786 graphics co-processor. This seems odd to me as it is quite
a new design. For those not familiar with the device, it supports all
the usual scan generation and drawing lines etc, but more interestingly
it suports hardware overlapping windows and does raterops. According to
the article I read Intel were expecting the chip to form the heart of
the VGA standard, and when IBM rejected it Intel decided to cease
production. This has, according to the article, produced much ill
feeling amongst small manufacturers here in the UK.

Is the rumour true? If so doesn't this seem a rather less than ethical
thing for Intel to do? It was claimed that the small manufacturers in
question vowed never to use Intel parts again, some even said they would
never use US parts again ( seems a bit drastic to me). But it must be
very difficult for small companies if even the large silicon companies
can't be trusted to continue to supply current parts. I realise that
some parts must eventually die out, but as I have said this is a newish
design. 

If this rumour is true then I am interested what implications this has to the
current RISC market. If manufacturers are going to stop production of
chips because sales are not as high as hoped for then what
happens if the 29000/88000/i860/MIPS/SPARC/.... etc. doesn't sell, will
the manufacturers keep production going?? There are so many RISCs on the
market at present that they can't all be successful, so 
does this mean that unless you back the winning horse you might be left 
with no supply at all ????

John Vaudin       Computer Science, Warwick University, UK.

e-mail arthur@uk.ac.warwick.cs

slackey@bbn.com (Stan Lackey) (06/23/89)

In article <1987@ubu.warwick.UUCP> arthur@flame.warwick.ac.uk (John Vaudin) writes:
>I read some rumours in a UK magazine that Intel have stopped production
>of their i82786 graphics co-processor. This seems odd to me ...

>If this rumour is true then I am interested what implications this has to the
>current RISC market. If manufacturers are going to stop production of
>chips because sales are not as high as hoped for then what ...

In cases where you are worried about this, like your company depends
on some part, there is a way to protect yourself.  Lore has it that it
is possible to get production rights if the supplier decides to stop
making it.  If you can, get the agreements signed before committing to
the part.  Now I imagine some suppliers might not want to sign such an
agreement, like they might not want to give away all their process
information if it is a custom process.  Then it's your decision if you
want to take a risk on the part, or perhaps use one from another
vendor that may not be as good.

Other approaches include buying up-front the entire volume you think
you will ever need.  (If the supplier won't sell them to you, it
should give you a real strong hint as to what is going on.)  This,
naturally, may be impractical for a small company.

If you do have production rights, and the process isn't too obtuse, 
you should be able to get the chip made.  It will probably end up
costing more, since the volumes aren't there.
-Stan    my opinions alone etc.

peter@ficc.uu.net (Peter da Silva) (06/23/89)

In article <1987@ubu.warwick.UUCP>, arthur@warwick.UUCP (John Vaudin) writes:
> I read some rumours in a UK magazine that Intel have stopped production
> of their i82786 graphics co-processor. This seems odd to me as it is quite
> a new design.

Not to mention that they're using it in their i520 Multibus-2 systems for
the console processor...
-- 
Peter da Silva, Xenix Support, Ferranti International Controls Corporation.

Business: uunet.uu.net!ficc!peter, peter@ficc.uu.net, +1 713 274 5180.
Personal: ...!texbell!sugar!peter, peter@sugar.hackercorp.com.

mash@mips.COM (John Mashey) (06/25/89)

In article <1987@ubu.warwick.UUCP> arthur@flame.warwick.ac.uk (John Vaudin) writes:

>If this rumour is true then I am interested what implications this has to the
>current RISC market. If manufacturers are going to stop production of
>chips because sales are not as high as hoped for then what
>happens if the 29000/88000/i860/MIPS/SPARC/.... etc. doesn't sell, will
>the manufacturers keep production going?? There are so many RISCs on the
>market at present that they can't all be successful, so 
>does this mean that unless you back the winning horse you might be left 
>with no supply at all ????

This is actually a different situation, because of the software implications,
i.e., devices intended to be reprogrammable devices visible to users:
	a) Not only need to be built for a (reasonably) long time, but
	b) Had better have followon products that are compatible
Hence, choice of a CPU is an even higher-stakes game than choice of
a peripheral chip.  When a company picks an architecture, it's probably
making a 5-10-year committment, and so they think hard about it.
These things come in cycles: in the early 80's, there was a fight
amongst the various CISC chips, and it took a few years for things
to settle out; as you note, the current fight amongst RISCs is similar.
Anybody has a choice, whose extrema are:
	a) Make a very early choice, hoping to get ahead of the competition,
		taking the risks that the parts may not be the long-term
		winners, or even, may be withdrawn from the market,
		which happens, even sometimes from large companies.
		Not every part described in the trade press is actually built.
	b) Make a very late choice, where the winners are obvious,
		but there is no technology advantage over the competition.

I've used the analogy before, but it bears repeating: doing a),  that
is,  picking your CPU early in the game, is like playing Russian Roulette
where you pick your CPU, pull the trigger, and then wait 1-2 years to
see if you've blown your brains out.
-- 
-john mashey	DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP: 	{ames,decwrl,prls,pyramid}!mips!mash  OR  mash@mips.com
DDD:  	408-991-0253 or 408-720-1700, x253
USPS: 	MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086

mmm@cup.portal.com (Mark Robert Thorson) (06/26/89)

mash@mips.com says:
> Anybody has a choice, whose extrema are:
> 	a) Make a very early choice, hoping to get ahead of the competition,
> 		taking the risks that the parts may not be the long-term
> 		winners, or even, may be withdrawn from the market,
> 		which happens, even sometimes from large companies.
> 		Not every part described in the trade press is actually built.
> 	b) Make a very late choice, where the winners are obvious,
> 		but there is no technology advantage over the competition.

There's also alternative c):
        c) Choose a processor which doesn't quite make it, but isn't a total
                loser.  The marketing department of your CPU vendor then
                re-targets it to the laser-printer market.  Let's see, which
                chips have been touted as the "ideal" laser printer controller:
                32000, 29000, 2167/9, ...

acockcroft@pitstop.West.Sun.COM (Adrian Cockcroft) (06/27/89)

In article <41891@bbn.COM>, slackey@bbn.com (Stan Lackey) writes:
> In article <1987@ubu.warwick.UUCP> arthur@flame.warwick.ac.uk (John Vaudin) writes:
> 
> >If this rumour is true then I am interested what implications this has to the
> >current RISC market. If manufacturers are going to stop production of
> >chips because sales are not as high as hoped for then what ...
> 
> In cases where you are worried about this, like your company depends
> on some part, there is a way to protect yourself.  Lore has it that it
> is possible to get production rights if the supplier decides to stop
> making it.  If you can, get the agreements signed before committing to
> the part.

If you buy a processor that is widely second sourced then you reduce your
risk. In the case of SPARC it is also possible to licence the right to design
and manufacture your own SPARC designs for your own use. This is what
Solbourne and Prisma (and others) have done. It is a different licence
from that held by LSI, Cypress, Fujitsu, TI and BIT since they have the right
to sell chips on the open market. Note that the current Solbourne design uses 
Fujitsu chips but Matsushita/Solbourne are working on their own design.