"Samuel Benge" <sbe### [at] caltelcom> wrote:
> Are you joking? I don't know much about hardware engineering, so I
> really can't tell...
>
> Mike Blakely wrote:
>
> > See http://www.irtc.org/ftp/pub/stills/2002-06-30/povcard.jpg   first...
> >
> > To make this card is not technically challenging. I've been discussing
> > this with some fellow engineers for 4 years now. The only issue is cost,
> > which is related to quantity.
> >
> > How many people do you think would want this and how much would they be
> > willing to pay?

Of course he is joking, 96GB is already hard to have on hard disk, so think
of this in Ram...

Plus at 800 million pixels/s *and* 90 frames/s, you got roughly 9 million
pixels per image, ie. for example a 3442x2582 image, far from the
65536x65536 promised resolution. Unless they deliver either the resolution
*or* the speed...
At this speed, you no longer need meager OpenGL or DirectX accelerators...

As some others mentioned, if it was technically possible, ILM, Pixar and
others would had put a lot of money on the project...

A Windows 3.11 support (not to mention Dos) is... funny. I see no support
for AmigaOS not Tos (Atari)...

Is DKBTrace still alive? I used it on my Atari ST 520, and even recompiled
it on a Sun workstation.

-- #=--=#=--=#=--=#=--=#=--=#=--=#=--=#=--=# --
Philippe Lhoste (Paris -- France)
Professional programmer and amateur artist
http://jove.prohosting.com/~philho/

Post a reply to this message

Mike Blakely wrote:
> 
> To make this card is not technically challenging.

Well, getting 96Gigs of RAM onto a PCI card will be technically
challenging for several years.

> I've been discussing this with some fellow engineers for 4 years now.
> The only issue is cost, which is related to quantity.

Have you identified the most common calculations that are done by
POV-Ray in the course of a render?

> How many people do you think would want this and how much would they
> be willing to pay?

On a budget of $10000 I can buy an Athlon XP render farm that will
enable me to render a feature-length Rusty animation in about six
months (and that includes making frames that would take 7.5 hours to
render on my current machine).

Regards,
John

Post a reply to this message

Philippe Lhoste wrote:
> 
> "Samuel Benge" <sbe### [at] caltelcom> wrote:
> > Are you joking? I don't know much about hardware engineering, so I
> > really can't tell...
> >
> > Mike Blakely wrote:
> >
> > > See http://www.irtc.org/ftp/pub/stills/2002-06-30/povcard.jpg   first...
> > >
> > > To make this card is not technically challenging. I've been discussing
> > > this with some fellow engineers for 4 years now. The only issue is cost,
> > > which is related to quantity.
> > >
> > > How many people do you think would want this and how much would they be
> > > willing to pay?
> 
> Of course he is joking, 96GB is already hard to have on hard disk, so think
> of this in Ram...
> 

No joking, just dreaming!
Remember, 64K is enough for data page,
640K is enough for computer's memory,
512M is enough for a hard disk,
2 G is enough for a partition and a video file,
4 G is enough for file,
49 days and a few more is enough for a system clock to roll up...

> Plus at 800 million pixels/s *and* 90 frames/s, you got roughly 9 million
> pixels per image, ie. for example a 3442x2582 image, far from the
> 65536x65536 promised resolution. Unless they deliver either the resolution
> *or* the speed...

Well, you have the problem of the available bandwith anyway at the interface ;->
That's probably the biggest bottleneck.


> 
> A Windows 3.11 support (not to mention Dos) is... funny. I see no support
> for AmigaOS not Tos (Atari)...
> 
There was some in the early version (including also some older computer,
including a ZX80 (ancestor of ZX81 !), but it goes out to reduce the list :-)
May be I should have kept it in ?

> Is DKBTrace still alive? I used it on my Atari ST 520, and even recompiled
> it on a Sun workstation.

Latest version is now called Povray 3.5, I believe...

-- 
Non Sine Numine
http://grimbert.cjb.net/
Etiquette is for those with no breeding;
fashion for those with no taste.

Post a reply to this message

A serious effort to make a POV-card would have to resort to many parallel
CPUs with their own cache of RAM, and a small segment of the visual scene
dedicated to it.  Imagine splitting the image field into tiles, then having a
pre-processor that would divide the task into just what each tile would contain.
    Now, each tile gets its own 2+ GHz CPU, ends up only having about 64 pixels,
and blazes through a special RISC high speed processor.  The microcode would be
dedicated to image processing, the textures and pigments would be flashed into
memory and could be updated, and then the output would be written into a video
frame buffer.
    Now you have a chance to actually make this thing.  This could be the
rebirth of the transputer, only with new technology.

Cheers!

Chip Shults
My robotics, space and CGI web page - http://home.cfl.rr.com/aichip

Post a reply to this message

In article <3d448d5b$1@news.povray.org>,
 "Sir Charles W. Shults III" <aic### [at] cflrrcom> wrote:

>     A serious effort to make a POV-card would have to resort to many parallel
> CPUs with their own cache of RAM, and a small segment of the visual scene
> dedicated to it.  Imagine splitting the image field into tiles, then having a
> pre-processor that would divide the task into just what each tile would 
> contain.

This is the part that isn't workable. Each pixel depends on the entire 
scene, unless you remove shadows, reflections, refraction, radiosity, 
photon mapping, etc...

Each processor would have to work with the entire scene. And to keep 
things like radiosity consistent over the image, you couldn't just give 
each processor its own copy of the scene data, all the processors would 
have to share it and add to it on the fly.

-- 
Christopher James Huff <chr### [at] maccom>
POV-Ray TAG e-mail: chr### [at] tagpovrayorg
TAG web site: http://tag.povray.org/

Post a reply to this message

Thanks, Chris.
    Still, the real world is really made of little parts and no part knows what
the others are really doing except by the four forces.  I know that the optics
would be complex and yes, that everything is interdependent.
    Couldn't a sort of "illumination sphere" be derived by a front end processor
and made available to all the sub-processors?  I know that the light from other
parts of the scene are important to the realism, but overall it tends to be
rather blended together for many scenes, or at least that was my impression.
    I may be way off track here, but it should be possible to have a processor
that calculates each light source and renders partial results to the tile
processors.  And another processor that renders a sort of "light density and
direction" map that would be passed down.  Even if it only used an approximation
based on boundaries or something, it would be pretty fast.
    I once worked with a thing called the GAPP (geometric array parallel
processor) a little, and some sort of massively parallel pipeline for
illumination/shadows/photons may be possible, in concert with another parallel
that splits the scene into the tiles.  Then the pool of data would be present
for each part of the scene.
    Well, my little outlandish bit of nonsense for now.

Cheers!

Chip Shults
My robotics, space and CGI web page - http://home.cfl.rr.com/aichip

Post a reply to this message

"Le Forgeron" <jgr### [at] freefr> wrote:
> Philippe Lhoste wrote:
> >
> > "Samuel Benge" <sbe### [at] caltelcom> wrote:
> > > Are you joking? I don't know much about hardware engineering, so I
> > > really can't tell...
> > >
> > > Mike Blakely wrote:
> > >
> > > > See http://www.irtc.org/ftp/pub/stills/2002-06-30/povcard.jpg
first...
> > > >
> > > > To make this card is not technically challenging. I've been
discussing
> > > > this with some fellow engineers for 4 years now. The only issue is
cost,
> > > > which is related to quantity.
> > > >
> > > > How many people do you think would want this and how much would they
be
> > > > willing to pay?
> >
> > Of course he is joking, 96GB is already hard to have on hard disk, so
think
> > of this in Ram...
> >
>
> No joking, just dreaming!
> Remember, 64K is enough for data page,
> 640K is enough for computer's memory,
> 512M is enough for a hard disk,
> 2 G is enough for a partition and a video file,
> 4 G is enough for file,
> 49 days and a few more is enough for a system clock to roll up...

Nothing wrong with a little dreaming... I still think it would have been a
good April Fool joke.

I meant, of course, that today it cannot be done.
"To make this card is not technically challenging." Really? Today? :-)

> > Plus at 800 million pixels/s *and* 90 frames/s, you got roughly 9
million
> > pixels per image, ie. for example a 3442x2582 image, far from the
> > 65536x65536 promised resolution. Unless they deliver either the
resolution
> > *or* the speed...
>
> Well, you have the problem of the available bandwith anyway at the
interface ;->
> That's probably the biggest bottleneck.

Well, you can add a TV output. HDTV anyone?

> > A Windows 3.11 support (not to mention Dos) is... funny. I see no
support
> > for AmigaOS not Tos (Atari)...
> >
> There was some in the early version (including also some older computer,
> including a ZX80 (ancestor of ZX81 !), but it goes out to reduce the list
:-)
> May be I should have kept it in ?

To list all old systems would have been funny, but a bit too long. It's OK
this way.

> > Is DKBTrace still alive? I used it on my Atari ST 520, and even
recompiled
> > it on a Sun workstation.
>
> Latest version is now called Povray 3.5, I believe...

Yeah, just curious if it has evolved along a parallel branch, or if anybody
is still using it.
Does PoV-Ray is able to understand DKBTrace scenes? I can't remember if the
syntax has changed a lot.

-- #=--=#=--=#=--=#=--=#=--=#=--=#=--=#=--=# --
Philippe Lhoste (Paris -- France)
Professional programmer and amateur artist
http://jove.prohosting.com/~philho/

Post a reply to this message

See this http://www.realtime-raytracing.com/ ....

--
Luis
All I Need Is POV

****************************************************
Lightning Generator
http://www.lightning-generator.fr.st
http://LG666.ifrance.com

http://www.pov-wip.fr.st

http://www.pov-monde.org
Chan IRC sur eu.undernet.org : #povray-fr
****************************************************

3D405CAE.BF7D063F@cts.com...
> See http://www.irtc.org/ftp/pub/stills/2002-06-30/povcard.jpg   first...
>
> To make this card is not technically challenging. I've been discussing
> this with some fellow engineers for 4 years now. The only issue is cost,
> which is related to quantity.
>
> How many people do you think would want this and how much would they be
> willing to pay?
>
> Mike
>
>

Post a reply to this message

"Philippe Lhoste" <Phi### [at] GMXnet> wrote in message
news:3d464f98@news.povray.org...
> > > > Mike Blakely wrote:
> > > > >
> > > > > To make this card is not technically challenging. I've been
> discussing
> > > > > this with some fellow engineers for 4 years now. The only issue is
> cost,
> > > > > which is related to quantity.
> > > > >
> I meant, of course, that today it cannot be done.
> "To make this card is not technically challenging." Really? Today? :-)
>

    The numbers are a bit 'out there', but the basic idea is feasible.
    As Mike says, the main problem is cost. You could develop a custom slab
of silicon that could easily manage up to current desktop resolutions (even
HDTV), but how many people would buy it, and therefore what's the unit price
going to be to cover the development costs for the custom silicon ?
    Bandwidth wouldn't be an issue - zap across your scene description &
source image-maps and all you'll need to be pumping back out, across the
bus, is each finished frame at whatever framerate you desire...

--
Pandora/Scott Hill/[::O:M:C::]Scorpion
Software Engineer.
http://www.pandora-software.com

Post a reply to this message

Luis wrote:
> See this http://www.realtime-raytracing.com/ ....

Site is far too vague on details for a serious investor.

-- 
Tim Cook
http://empyrean.scifi-fantasy.com
mirror: http://personal.lig.bellsouth.net/lig/z/9/z993126

-----BEGIN GEEK CODE BLOCK-----
Version: 3.12
GFA dpu- s: a?-- C++(++++) U P? L E--- W++(+++)>$
N++ o? K- w(+) O? M-(--) V? PS+(+++) PE(--) Y(--)
PGP-(--) t* 5++>+++++ X+ R* tv+ b++(+++) DI
D++(---) G(++) e*>++ h+ !r--- !y--
------END GEEK CODE BLOCK------

Post a reply to this message