1. This forum section is a read-only archive which contains old newsgroup posts. If you wish to post a query, please do so in one of our main forum sections (here). This way you will get a faster, better response from the members on Motherboard Point.

high-framerate videos of PS3 simulations (not games or game-demos)

Discussion in 'Nvidia' started by Dan, Apr 3, 2006.

  1. Dan

    Dan Guest

    oh wow, simulators! LOL
     
    Dan, Apr 3, 2006
    #1
    1. Advertisements

  2. Dan

    Air Raid Guest

    http://www.playsyde.com/news_2734_en.html


    Raycasting crater-volcano demo ~
    http://www.playsyde.com/leech_2410_en.html
    (uses 1 GB RAM)

    'Chameleon Fish' ~ http://www.playsyde.com/leech_2411_en.html


    'Crowd Demo' (more Fish) ~ http://www.playsyde.com/leech_2412_en.html




    the fish demo ran at 30 fps / hz because of the type of graphics
    rendering used, but the underlying simulation could run at 60 fps / hz


    these are all more about showing the massive computational performance
    of the CELL CPU than it is about graphics rendering of the RSX GPU,
    but of course RSX was used extensively for the Chameleon Fish demo.
     
    Air Raid, Apr 3, 2006
    #2
    1. Advertisements

  3. Dan

    Sir Daniel Guest

    Man....this Demo sucks.

    My old 500MhZ PIII with a TNT2 could this better.
     
    Sir Daniel, Apr 3, 2006
    #3
  4. was this the internal mail from sony?

    'hi devs,

    please create 3 technology demos that will SEND THE AUDIENCE INTO A DEEP
    SLEEP. im thinking bulk fish and some sort of landscape demo like that
    awesome one from 1993 with the voxels and the helicopter... except
    without the helicopter

    there is a spare 386SX/16 with 1mb ram a 512k trident VGA card available
    for you to run the demos on, that should suffice'
     
    furious gibbon, Apr 3, 2006
    #4
  5. Dan

    Chris F Guest

    i dunno, the chameleon fish one was probably the most impressive of
    the three, but thats not saying much.

    the landscape one didn't really seem to show off anything at all, i'm
    guessing it was to highlight the lighting effects on the ridges, etc,
    but it just looked like they bump-mapped a google earth jpeg or
    something.

    the fish crowd one, while impressive that it can calculate and control
    10'000 individual fish, the logic wasn't exactly high tech, they
    didn't show much differentiation in their movement patterns.

    all in all, i'm a bit worried that its 8 months til the proposed
    launch, and this is the best they can show off at the GDC?

    i guess that since these aren't really about the GPU, but rather the
    processing power of the Cell chip you can't really judge them
    visually, but you would expect them to be showing at least something
    that does utilise the GPU to some extent, because its one thing going
    "look, it can calculate the movement patterns of 10000 individual
    items" but how that translates into being something useful for games,
    i'm not too sure.

    still, the next finding nemo game should look good.........
     
    Chris F, Apr 3, 2006
    #5
  6. Dan

    Dreamcasted Guest


    Oh wow a dumbass nigger! LOL

    --
     
    Dreamcasted, Apr 3, 2006
    #6
  7. Dan

    spaceman Guest

    oh wow, framerate problems in the terrain demo, and 10,000 fish at 10 polys
    each LOL. so it's moving 100,000 polys a sec eh? mighty impressive!
     
    spaceman, Apr 3, 2006
    #7
  8. Dan

    Dreamcasted Guest

    What a stupid bunch of fucks.

    Do you even know what Raytracing is?

    --
     
    Dreamcasted, Apr 3, 2006
    #8
  9. Dan

    spaceman Guest

    Raytracing or not, those demos are completely underwhelming. Why don't they
    quit wasting their time on bullshit like this when they could have designed
    an actual level of gameplay, or anything similar to real world, interactive
    realtime graphics to show what these things are going to look like and
    perform like in a practical environment? You put a bunch of high poly
    characters and environments onscreen that are normal mapped & bumpmapped,
    parallax mapped, with soft dynamic shadows and lighting, raytracing(or so
    they hope), character AI, physics, particles, some complex animations for
    characters and objects(maybe even procedural), precision collision
    detection, and so on... and then come back and show me something worth
    seeing. Yeah, i'd say this stuff is a bunch of bullshit and people tire of
    it. And you're the kind of dumbfuck that falls for these kinds of tech
    demos. I bet you own a PS2, and it's your favorite system. Why don't you go
    watch the Crysis tech demo and tell me which is more impressive? Because
    that's a very good indication of what we're going to see on our screens, not
    a school of 10,000 fish made of 10 polys each floating around in empty
    space. And who cares about Cell? It's still nothing without the sum of parts
    as a whole..
     
    spaceman, Apr 4, 2006
    #9
  10. Dan

    Guest Guest


    Stop cross posting this rubbish to comp.arch
     
    Guest, Apr 5, 2006
    #10
  11. Are you confusing ray casting with ray tracing? The original poster
    said that they were ray casting, a cheaper rendering algorithm than ray
    tracing.

    Before I past judgment on the demos and decide whether I want to buy a
    PS3 or XBOX 360, is there a website that accurately describe the
    hardware configuration of the system?

    Derek
     
    Derek Simmons, Apr 5, 2006
    #11
  12. I thought the original poster said they were ray casting, not ray
    tracing. Ray casting is a much cheaper method.

    Derek
     
    Derek Simmons, Apr 5, 2006
    #12
  13. Dan

    buyanovsky Guest

    Ray-casting provides the best overall rendering quality for volumetric
    data "volume rendering" (not 2d polygons hanging up in 3D space).
    The examples of ray casting of volumetric data can be found on
    www.fovia.com

    It is relatively simple to do brute force ray-casting on GPU or CPU. To
    obtain high quality ray-casting 8 samples per voxel has to be
    maintained (case one voxel is projected to one pixel on screen). To
    render by brute force 1K-cube on 1K-projection plane with 30fps will
    require 1000 X 1000 X 1000 X 7 X 8 X 30 = 1.680.000.000.000 samples. 7
    is number samples per Phong (6-gradient + 1 classification sample) so,
    number of Phong calculation is 240.000.000.000. This numbers are way
    above modern GPU capability. Thus, the fps shown by cell in original
    post demonstrates that cell is quite good performer. However the
    examples of rendered volumes are highly homogenous and adaptive
    ray-casting technique may provide similar performance on dual core dual
    Opteron machine.

    Regards,
    George
     
    buyanovsky, Apr 22, 2006
    #13
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.