Re: The Agamemnon Challenge - Which He Failed Miserably.
- From: imipak <imipak@xxxxxxxxx>
- Date: Fri, 8 May 2009 16:08:52 -0700 (PDT)
On May 8, 2:06 pm, "Nod" <N...@xxxxxxxxxxxxxxxxx> wrote:
"Agamemnon" <agamem...@xxxxxxxxxxxxxxxx> wrote in message
news:j-idnaZ_-axRDJnXnZ2dnUVZ8o6dnZ2d@xxxxxxxxxxxxxxxxx
|| "Nod" <N...@xxxxxxxxxxxxxxxxx> wrote in message
|news:vz%Ml.4973$oh6.4570@xxxxxxxxxxxxxxxx
| > The original challenge for those that missed it and those that want to
see
| > it again.
|
| You failed to supply the animated model of the Titanic used by Cameron or
| one of the same quality. YOU FAIL BY DEFAULT! I WIN!
|
That'll be because the challenge wasn't to pinch a scene from the film
Titanic. It was for YOU to do a 5 minute animation of the Titanic at sea
which YOU FAILED TO DO.
Stop crying like a little brat and stop trying to move the goalposts by
adding conditions which were never part of the challenge you accepted.
Even if it were to pinch a scene from the movie Titanic, his own link
shows that he could never have done it on any single computer, that
even one shoved full with multiple top-of-the-line graphics cards
simply could not do what Aggy claims it can. You can probably get
quite a lot of mileage from using Aggy's own supply of evidence to
show that his boasts simply aren't possible, that even the best Aggy
can come up with falsifies Aggy's position.
In fact, let us take a look for a moment at what Aggy supplied. The
graphics card in the machine he points to is the Quadro FX 4700 X2.
According to NVidia, this graphics card does the following:
"Unmatched performance and Shader Model 4.0-based realistic effects
for all next-generation OpenGL and Microsoft® DirectX® 10 professional
applications."
So it's shader-based, not ray-tracing. So Aggy's wrong about the
graphics card actually doing any of the ray-tracing in hardware.
(Error 1. It's not hardware ray-tracing.)
In fact, it's using OpenGL shaders, not Renderman shaders, and Cameron
used Renderman not OpenGL. Thus even the shaders on the card could
never have been used to accelerate the software Cameron used. (Error
2. It's not usable by Renderman or anything similar and therefore
could not have handled the scene files from the movie even if Cameron
supplied them.)
It has a memory transfer rate of 51.2 GB/s, which means it's double-
speed PCI Express. Regular PCI and even regular PCI Express isn't
capable of that kind of speed. Not only that, but you'd run out of
lanes if you were running PCI Express at that kind of speed to the
graphics card, which means you've no lanes left for transferring
instructions from memory to the CPU. This means each and every scene
file has to be pre-generated offline and pre-loaded. (Error 3. The
modeling isn't real-time, it's all pre-generated.)
The rendering in the article is one quarter movie resolution. If we
scale up to movie resolution, we get a whopping 7 frames per second
for this shading, even with two of these cards linked in parallel,
which falls under the 24 frames per second needed for movies. (Error
4. The animation cannot be done in real-time even on this massively
powerful system.)
The rendering in the article only handles three levels of direct
reflection and only with single rays. There is no attempt to do
diffuse reflection or refraction, both of which you need for modeling
the sea, no attempt at correctly handling edges and no attempt to
handle the fact that sunlight isn't from a point source of zero
radius. (Error 5. The software in the article is simply too primitive
to do what was done in The Titanic.)
http://www.linuxjournal.com/article/2494
Aggy claims the high-end PC of today is 200x faster than the entire
cluster farm of DEC Alphas used in the movie. If we were to slow down
the rendering to achieve the quality Cameron achieved, using ray-
tracing techniques (since that's what the article talks about), it
would take about a day for this card to do a single frame of the movie
and thus 24 days to produce one second of footage. In order to be 200x
faster than the cluster farm, this would imply it took 4800 days per
second of rendered footage in the movie (13 years real-time per second
of movie). According to the people involved, 30,000 frames were
generated (12,500 seconds of footage).
.
- Follow-Ups:
- References:
- Prev by Date: Re: The Agamemnon Challenge - Which He Failed Miserably.
- Next by Date: Re: The Agamemnon Challenge - Which He Failed Miserably.
- Previous by thread: Re: The Agamemnon Challenge - Which He Failed Miserably.
- Next by thread: Re: The Agamemnon Challenge - Which He Failed Miserably.
- Index(es):
Relevant Pages
|
