_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
news videos cs maps articles software music forum
LIGHT | DANK

PixelMachine Stress Tests

This page is about testing the upper limits of PixelMachine with lots of objects, reflections, effects, and at very high resolutions.

Enjoy!

Needs More Spheres! A stress test with 4225 spheres. All spheres can potentially reflect all other spheres. Limit set at 200 bounces.
  • Resolution: 1280 by 960 (1.2MegaPixel)
  • Rendering Time: 1 hour on 3.4 GHz Pentium 4
      • This was sped up by placing the spheres in the hit-detection cubes
      • Also slowed down by some glitches in the reflection code -- where you see solid black areas
Errors in red Similar to the above but from an angle where the reflection code glitch shows up a lot. The red shows where a ray couldn't stop bouncing around. Obviously that shouldn't be happening so much.
Photon Not-Mapping This image was rendered using a similar method to "photon mapping" but without the map. Source code is coming soon, I have to clean it up.
  • Resolution: 800 by 600
  • Photon count: 300 Million
  • Rendering Time: 21 minutes on Athlon 64 X2 5000+
  • Photons were shot from a rectangular lightsource on the ceiling (not visible)
  • Diffuse lighting is too weak in this image
Photon Not-Mapping w/Diffuse Unmapped photons with better diffuse lighting.
  • Resolution: 800 by 600
  • Photon count: 3 Million
  • Rendering Time: ~40 seconds on Athlon 64 X2 5000+
  • Photons were shot from a rectangular lightsource on the ceiling (not visible)
  • Diffuse lighting more apparent here
1 Billion Photons This really isn't all that stressful. An increase in photon count increases rendering time linearly. There is no additional memory requirement for more photons either.
  • Resolution: 800 by 600
  • Photon count: 1 Billion
  • Rendering Time: ~1 hour on Athlon 64 X2 5000+
  • Same light source as previous
  • Accidentally rendered in black and white :(
  • The line artifacts are due to the not-so-randomness of rand()
Unfair Comparison Here is a comparison between the simple ray-tracing and the new photon mode. Not quite a fair comparison because the lighting is directional in the photon mode.
  • Resolution: 800 by 1200 (2x600)
  • Photon count: 100 Million
2 Billion Had to do it. A 2 billion photon shot. Billions of photons causes the dominant color to drown out everything else. That's why it's red. This needs to be fixed.
  • Resolution: 800 by 600
  • Photon count: 2 Billion
  • Rendering Time: ~8 hours on Athlon 64 X2 5000+
  • Light source is finally directly visible!
Caustics! Caustics is when light is bent, or focused as it enters and exits materials like glass, water, and diamond. Shine light thru a glass of water and you'll see it.
  • Resolution: 800 by 600
  • Photon count: 1 Billion
  • Rendering Time: ~5 hours on Athlon 64 X2 5000+
  • Sphere Material: Glass (Refraction Constant = 1.50)
Mersenne Twist Switched to Mersenne Twister random number generation. Most random number generators are crap. With a good one (MT) we can get much better quality with fewer photons because they don't clump together.
  • Resolution: 1920 by 1200
  • Photon count: 1 Billion
  • Rendering Time: ~2.5 hours on Intel Quadcore 3.0 GHz
  • Sphere Material: Reflective/Metal


Back to main PixelMachine Page

Copyright © 2000-2017 SuperJer.com. All rights reserved.