Upcoming Events
Unite 2010
11/10 - 11/12 @ Montréal, Canada

GDC China
12/5 - 12/7 @ Shanghai, China

Asia Game Show 2010
12/24 - 12/27  

GDC 2011
2/28 - 3/4 @ San Francisco, CA

More events...
Quick Stats
82 people currently visiting GDNet.
2406 articles in the reference section.

Help us fight cancer!
Join SETI Team GDNet!
Link to us Events 4 Gamers
Intel sponsors gamedev.net search:

Putting Brute Force to the Test

I have put brute force to the test against a simple ROAM engine. I used Bryan Turner's ROAMsimple demo for this. Here's the technical info on the test:

Computer:
Pentium III 600 Mhz
128 Mb RAM
nVidia TNT2 32 Mb Video
800x600x32 resolution

Basis for comparison:
The detail level at the near clipping plane, defined as the size of the quad drawn between 4 adjecent points on the heightmap. This is the most interesting for game developers as it determines the quality of the scene. In both renderers I set the far clipping plane at 100.0f and increased the map size so it would fill in all directions beyond the far clipping plane.

The results:
I put the results in the graph below. The data is interpolated to find the crossing point which lies near the detail level of 2.0f. This means that if you need better detail than 2.0f you'd better use a CLOD engine. If you don't need this high a detail level, brute force can be a good alternative.

fps per detail level (interpolated)

Of course this is by no means a correct scientific experiment but I think it gives a reasonably good representation of the two methods.

Evaluation

The graph above shows that, when the detail is equal or worse than 2.0f, brute force can be used. But there are other differences that should be taken into consideration. For example, brute force can render maps with less memory usage because no triangle lists are built every frame. Finding the exact (interpolated) height with brute force is easier than with CLOD. Maps can be bigger, and finally, brute force is easier to implement (which is very good for hobby coders like me). On the flip side, CLOD simply gives higher quality and more detailed levels, and 3D accelerators need to become much faster before brute force can do the same. In the end, the choice totally depends on the needs of your engine.

Conclusion

The power of 3D accelerator cards has increased so much over the last few years that, under the right conditions, old fashioned brute force terrain rendering can be faster than CLOD algorithms. The extra CPU overhead of reducing triangles can sometimes be higher than the speed increase that you get out of it. With current technology, brute force can become faster than CLOD when the size of the triangle quads rises above approximately 1/50th of the distance to the horizon, but this value will become smaller as technology advances.

Sander "Tweety" Maréchal
Lone Wolves Production
s.marechal@jejik.com




Contents
  Introduction
  High Speed Brute Force
  Putting Brute Force to the Test

  Printable version
  Discuss this article