Index Home About Blog
Subject: Re: Ticked off about "Analyst" view of Pixar (PIXR)
From: mash@mash.engr.sgi.com (John R. Mashey) 
Date: Dec 04 1995
Newsgroups: misc.invest.stocks

In article <jpanicoDJ0z9F.LAI@netcom.com>, jpanico@netcom.com (Joe
Panico) writes:

|> : 1. The entire TS movie was rendered on Sun workstations Not SGI (who
|> : will be dead within 14 months, but that's another story).
|> 
|> According to the aritcle in Wired, yes the movie was entirely *rendered*
|> on Sun workstation, but much of the 3-D *modelling* was done on SGIs.

Yes.


|> Pixar also makes Renderman, commerical rendering software, and some of
|> their revenue stream is derived from it. Also, the value that Pixar adds
|> to digital movie making is in their proprietary modelling and rendering
|> software, not simply in their use of off-the-shelf software. Now that
|> SGI owns Alias and Wavefront, SGI's modeling and rednering software and 
|> techniques are indeed in direct competition with those from Pixar. If
|> Alias and Wavefront can produce better software than Pixar, then in short
|> time they, or anyone buying their software, could produce movies that
|> are as graphically appealing as Pixar's.

1) Alias/Wavefront are primarlily in business to sell tools to people
who use those tools to do things.

2) Pixar, these days, seems to be more of a creative animation house
that sometimes sells tools, and that like some computer companies, does
not just use widely-available tools, but does some of its own (i.e.,
Renderman) ...  While not an expert in this area, I've been in plenty of
meetings with "Sili-wood" companies, and observe they all mix-and-match
off-the-shelf + home-grown software in various degrees.
I guess there is some overlap (the computer business has them these
days), but not too much.

3) Just having the tools is nowhere near enough to make appealing movies;
it also takes serious artistic & other talents that do not just appear
because one obtains good tools.  What's clear is that some cool things
happen when talented folks combine with good tools, and it is indeed
exciting to see the progress over the last few years.

4) Just to be clear: if SGI thinks Pixar is a competitor, this is not
obvious to me.  Pixar looks to me like a valued customer 
[SGI's Silicon Studio had a private showing of Toy Story last week attended by
a PIXAR-V.I.P. to talk beforehand and answer questions.  It was great.]
When Pixar wants to come over to visit and talk, they get senior technical
people to interact with.

-john mashey    DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP:    mash@sgi.com 
DDD:    415-933-3090    FAX: 415-967-8496
USPS:   Silicon Graphics 6L-005, 2011 N. Shoreline Blvd, Mountain View, CA 94039-7311

Subject: Re: SGI slipping?
From: mash@mash.engr.sgi.com (John R. Mashey) 
Date: Dec 07 1995
Newsgroups: misc.invest.stocks,comp.graphics,mex.red,mex.general

(catching up):
In article <30c2ffcd.2517560@news2.cts.com>, mikem@alike.com (Mike
Morrison) writes:

|> I want to know how your company plans to compete against the Pentium
|> Pro? Especially 200 Mhz (and above) Quad (and above) systems that are
|> blindingly fast running true threaded apps in NT. I know you can
|> continue to drop prices, but it will be impossible to beat the
|> existing hardware infrastructure of the PC once more "SGI traditional"
|> graphics apps start porting to NT. And can those few high-end CGI
|> studios really keep SGI in business?

I'm not sure how to answer this, but I'll try:
        a) As I've said numerous times, low-end SGI's overlap with PCs,
        Macs, and plenty of other things, and have done so for a while;
        much of the entertainment business has always used other solutions
        at some price points, and this hasn't changed much. Studios with
        rendering farms have always used whatever gave them best price/
        performance on boxes with little or no graphics in them.
        
        b) (Must be careful here to not say things I'm not allowed to):
        *most* of SGI's entertainment-industry business is mid-range and
        up.  Here's a useful question whose exact answer would shed light
        on the above question:

                Of SGI's total revenue, what percentage is Indy's
                sold to the entertainment business?
        Is it:  50%?
                10%?
                a lot less.
        A: a lot less.

Hence, the question above doesn't make a lot of sense to me ... that is,
somebody may think that low-end SGI's in entertainment are the bulk of
SGI business, and that if that is attacked, SGI is out of business ...
but this is just plain silly.  Of course, even at the bottom end of
SGI's range, the comparisons are not as simple as they sometimes are made to
look:

Without spending a ton of time on it, observe that:
        a) Some tasks are I/O-bound.
        b) Some tasks are CPU-bound.
        c) Some tasks are graphics-bound.
SGI sells systems that tend to be fairly good at all 3, including all
3 at once, which is sometimes hard, requiring multiple hardware resources.
SGI does know something about using multi-processors for graphics, having
done so since 1988, and it turns out that for many interactive graphics
tasks, the fastest available CPUs are nowhere near close to the
performance you need, and hence you want serious graphics device(s)
connected up with the CPUs via high-speed channels; CPUs alone just do
not hack it.
[I'll post a serious discussion of this a bit later,
when I get some time, but within next week or so, for those who care.]
But, just taking a simple case:
        Consider the I/O graphics bus used by various systems; it may, or
        may not be a bottleneck, depending on what you're doing:

        New PCs: 133MB/sec (peak) PCI bus, a big improvement, everyone is very
        excited ... and this is the same speed bus as an SGI Indigo, which of
        course, hasn't been sold for a while.

        LOW-END SGI: Indy: 267MB/sec (peak) GIO64 ... By 1993, SGI had
        abandoned the previous GIO32 bus (133 MB/sec) as insufficiently fast...

        Without pre-announcing anything, I will note there are those people
        who think 267MB/sec is Awfully Slow...


|> Also, how much does SGI really stand to benefit from your Nintendo
|> license? It looks like Sega and Sony already have a major jump on that
|> platform. Not flaming, just curious.

As I've posted before:
        1) The financial arrangements with Nintendo are not public.
        2) The annual report said something like "fiscal 1996 revenues from
        Nintendo not material".

Also, just as a reminder ... Sony Playstations have MIPS chips as well.

-john mashey    DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP:    mash@sgi.com 
DDD:    415-933-3090    FAX: 415-967-8496
USPS:   Silicon Graphics 6L-005, 2011 N. Shoreline Blvd, Mountain View, CA 94039-7311

Subject: Re: SGI slipping? [really: "rape" is pretty serious accusation]
From: mash@mash.engr.sgi.com (John R. Mashey) 
Date: Dec 07 1995
Newsgroups: misc.invest.stocks,comp.graphics

In article <30c46176.2702234@news2.cts.com>, mikem@alike.com (Mike
Morrison) writes:

|>>Note, a slightly strange thing here, is that if you look at:
|>>http://www.toystory.com/about/history/history.htm
|>>i.e., excerpts from Mr. Morrison's "Becoming a Computer Animator",

|>>Listerine, Lifesavers, Tropicana, Toy Story
|>>Terminator 2, Beauty and The Beast, Jurassic Park, Forrest Gump, The Mask
...
|>>While some companies may have the market clout to nearly force the
|>>purchase of their products, I do not believe this applies to SGI, as
|>>there are always alternatives.
|> 
|> What are the alternatives?

At various times & in various market segments: Macs, PCs, Suns, IBMs,
Amigas, all sorts of things as rendering farms, and often, near-giveaway
systems by various people wanting to be in this business.

|> >In addition, compare:
|> >1) profits SGI has made selling to the animation community
|> >2) the profits made by selling the results, i.e., movies and commercials.
|> >The latter is substantially higher.
|> >This is not always the case of course, when people sell technology to
|> >other people :-)
|> 
|> I don't think this is a fair comparison. It would be better to look at
|> the average lifespan of a CGI house. Only a handful have survived over
|> the years. I'm not saying that is due to SGI, but just because a few
|> select companies have made good money on a few movies/commercials
|> doesn't really say anything concrete about your prices. Maybe they
|> were low and CGI would never have happened. Maybe they were high and
|> every desktop today would have an SGI chip in it for real-time 3D.

Re: fair comparison & select companies: I don't understand this.
*You* provided a nice list of great things that had happened in CGI over
the last few years ... and I just pointed out that many of them used
SGI gear for at least part of their work.  Plenty of other systems
were available to do this work, but by your own list, some people *chose*
to buy SGIs ... and pay for R&D that runs 10-12%, rather than the
2-4% (or even nearly-0%) found in PC-hardware companies.
It does cost more to push hard on the state of the art, and to
provide support that includes time spent by senior engineers working
with customers ... which also doesn't come free.        

|> >Now, "rape" is a pretty serious term. So, want to explain?
|> 
|> CGI Houses have been forced against their will to pay outrageous
|> prices for SGI components & maintenance. Compared to the existing
|> hardware market. But, you're probably right, it was too strong of a
|> word.

One more time: there is surely an emotional mindset going on here;
I don't know what the generic term "CGI houses" means; I suspect it
means that some people wished they had the money to buy SGIs, and
didn't ... and I'm sorry, I wish that doing aggressive R&D cost nothing,
and that building software that persists a while (rather than having
interfaces trashed pretty often) cost nothing, and that having senior
engineers spend a lot of time with leading-edge customers cost nothing,
and that working on open standards that promote advances in the state of
the art (like OpenGL, or 64-bit C, or DMIG, or...) cost nothing ...
etc, etc, etc, etc, and all sorts of things that need to be done that
aren't obvious...

...but they don't.  It is the nature of this business that things that
used to be high-end fall down into the commodity end sooner or later,
and that it is *good* ... but the leading-edge always costs more than the
sum of its parts, and if it's worth it you pay for it, and if it isn't,
you don't.


"Forced aginst their will" is awfully strong, and I'm glad to hear that I
was probably wright in saying "rape" was a bit strong?:

        a) Did SGI hold a gun to their heads?
                Unlikely, guns are not standard equipment for our salesforce.
        b) Did SGI threaten blackmail?
                Not too likely.
        c) Did SGI use overpowering dominance of the computer market?
                Not likely; SGI has never been IBM, Intel, or Microsoft.
        d) Did SGI threaten (illegal) tie-in sales?
                ("You need this Onyx, but if you don't also buy these other
                SGI machines for your rendering farm, we won't sell you the
                Onyx.")  Sure, that's *real* likely :-)
        e) The reality is that many CGI shops have no SGI gear, and very few
                (if any) are 100% SGI.

-- 
-john mashey    DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP:    mash@sgi.com 
DDD:    415-933-3090    FAX: 415-967-8496
USPS:   Silicon Graphics 6L-005, 2011 N. Shoreline Blvd, Mountain View, CA 94039-7311

Subject: Re: SGI slipping? [really: "rape" is pretty serious accusation]
From: mash@mash.engr.sgi.com (John R. Mashey) 
Date: Dec 19 1995
Newsgroups: misc.invest.stocks,comp.graphics

In article <30C49EC9.45A0@gte.com>, Yechezkal-Shimon Gutfreund
<sgutfreund@gte.com> writes:

|> While we are still on the topic. Did anyone see the
|> article in SiliconGraphicsWorld about how SGI SiliconStudio
|> is segments of significant clients (DreamWorks, BOSS, etc.)
|> to farms of Dec Alphas running NT and Lightwave?
|> 
|> Believe, me I am not trying to knock SGI. Probably a
|> fair ammount of hardware acceleration techniques and
|> knowledge about how to tune and balance a graphics system
|> was developed there and at UNC. 
|> 
|> But it appears to me that SGI as a business is going to
|> have to come up with a new strategy other than: More of
|> the same - to deal with the new threat.

It is fairly difficult to understand this comment, especially if you
have been talking to these sorts of customers over the years.
While not a graphics guru, I do talk to these folks.  They visit here
fairly often.  For some weird reason, people seem to think that SGI
"owns" this business, and that any presence of other vendors means
The End.... Nobody familiar with this business could think that....

I'd observe that there are at least 3 functions that such folks do,
and they include:

1) Interactive modeling/animation, i.e., real-time graphics.

2) Data management.

3) Rendering of the final images, and maybe some intermediate ones.

and, if they do a lot of their own software development

A one-person shop might do all of this on one system [probably a Mac];
bigger shops will have more systems, perhaps huge numbers.
In many shops, there is a wild mixture of systems, as:

1) Needs interactive graphics good enough for the creative folks to
do what they want to do at the speed they want to do it and be able to
pay for it.  This piece can be *very* graphics-intensive.
Commonly-favored platforms include:
        Very high end: SGI Onyx
        Mid-high-end: SGI Indigo2 Extreme (2 years ago), Indigo2 Impacts
        Mid: SGI Indy, maybe Suns, or high-end Macs or PCs 
        Low: Macs or PCs

2) If you are doing something big, may be a large server or servers.
        If SGI is in the shop, it's probably some flavor of Challenge SMP,
        but it could be other vendors' systems.

3) Rendering seems to be done on:
        supercomputers, if you have one.
        special-purpose systems like IBM's PVS
        SMP servers
        workstations not otherwise in use
        video toasters
        "render-farms", that is racks of (typically) workstations without
        graphics heads stacked up to give throughput of CPU, or more
        likely, server variants of the workstations, perhaps with everything
        stripped out.  Some rendering tasks fit this quite well, since
        each final rendering takes a long time, and they parallelize
        perfectly, and have predictable memory usage, so this works.
        

Category 3 is always a free-for-all, since:
        (a) Depending on one's choice of rendering software, you may often
        select systems with zero graphics hardware, and you pick whichever
        systems give you adequate turnaround time at lowest cost ... since
        it may well take 3 hours/frame to do final rendering for movies.
        There is not only no advantage for having good graphics boards
        in many cases, it is a cost disadvantage.  Also things like:
        video-input, good audio, video-cameras, high-performance disk I/O,
        etc, etc ... are either irrelevant or not very important.
        (b) Since interactive graphics performance is usually irrelevant,
        and if the code is developed internally [rather than needing
        specific applications that may not be available], vendors wishing
        to get into the movie/TV business may choose a shop to which they
        donate equipment and/or discount equipment incredibly in the hopes
        of getting further business.  This is easier than trying to attack
        category (1) and sometimes (2).

Examples:

==Jumanji - ILM
See: http://www.sgi.com/ and click on the "Creating Amazing Special
Effects in Jumanji".  or go to "What's New" and look around.
In this case, ILM is very heavy SGI customer, and SGI's were used for
all 3 functions.  One of the Showtime channels is running a piece on
the making of Jumanji.

==Toy Story - Pixar

(Disney channel has nice presentation on this one).

(1) 70+ Indigo2 Extremes for modeling, animating, shading, lighting,
compositing.  Alias|Wavefront's PowerAnimator + Pixar proprietary software

(2) Challenge network server

(3) A Sun rendering farm: 117 Sun headless workstations

See:
        http://www.Sun.COM:80/cgi-bin/show?951201/cover/cover.html
for a generally useful writeup, if you ignore several issues:
(a) Lots of films are mentioned ... most of which were done on SGIs.
(b) Some rather misleading price/performance figures, i.e., that compare
        "rendermarks" on properly-configured Suns (i.e., no graphics)
        versus SGI workstations with monitors, graphics boards & such.
        [When people use SGIs as render farms, they use bunches of
        special Challenge S's, i.e., Indys with no graphics and other
        things deleted ... since this makes a difference of ~3X-4X
        improvement in SGI price/performance for the way people actually use
        them in a render-farm.]
(c) And of course, that there were any SGIs involved anywhere in making this
movie of course got lost somewhere....  The real truth is that Toy Story
wouldn't have gotten made without all of:
        - The talent at Pixar
        - The SGIs that Pixar bought
        - And the Sun render-farm

==SeaQuest DSV

Uses VideoToaster; I don't think there are any SGIs; see:

http://www.omnipresence.com/Amiga/Hardware/VideoToaster.html

==Summary:
        SGI has *never* *owned* this business ... although SGI certainly has
        a major presence, and certainly many exciting things
        have been done using SGIs.   SGI presence is concentrated in
        categories (1) and (2), where SGI can add some value.
        Category (3) gets driven by cost/(render-unit), and sometimes by
        wild pricing/outright giveaways; SGI competes for this business,
        wins some, and loses some ... just like it's been for a while.

Index Home About Blog