Realtime Raytracing benchmark!

alexsok

Regular
Guru3D reports the following:
The people from realtimeraytrace released a benchmark based on a 3D-Realtime Raytrace engine. I just installed it and gave it a try, pretty dandy .. it will bring down your system to it's knees and has excellent options. I just became aware of the fact that the engine does not use your 3D Graphics card though.

This benchmark calculates Processor Speed and Memory performance, it does ->NOT<- use the performance of your 3D-Accelerator.
You can grab the benchmark from the respective site which is right here. And when you run it, trust me .. it's not a pretty sight, but it is realtime raytracing.

Quote from the offical site:
So we started about 3 years ago with "Realtime Raytracing". Of course everything is software rendering, where we come to the main problem of this engine, the processor speed (and memory speed) has to be very high, for a good speed at a resolution of 512x384 you would need at least and AMD-Athlon at 1300Mhz. But processors will get faster and faster, so i think the time works for us.

Of course 3D-Accelerators get faster too, and have new features, but one day a polygon would be smaller than a pixel on your screen. When that day will come, there is absolutely no reason to use polygon routines. So we decided to use platonic solids combined with boolean algebra and raytracing.

You can snag it from the offical site here:
http://www.realtimeraytrace.de <-- click there on downloads

P.S
Hmm... 320x240@antialias@depth of field is the max I can go for acceptable fps... :LOL:
all understandable but pretty sad :cry:
 
640x480
- Shadows
- Reflections
- Anti-alias
- Depth of field

Average fps: 7.86
Minimum fps: 4.22
Maximum fps: 19.23

Running on an 1.33GHz Athlon :) The pixelation sucks though :-?
 
Real time ray-tracing was the star of 4k intros for some time now.
IIRC Pentium systems was just getting to spread around that time.

But the improvement of the technique since then is not so impressive.
Ok. there's texturing (no mipmaps, point sampling), volumetric lighting (slow as hell).

Compare that with the improvement of raster rendering, during the same timeframe!

Also, this demo (benchmark) has very agressive optimizations - causing terrible artifacts all along. (Part of the objects/shadows dissappearing and reappearing again between frames.)
The "antialiasing" is really a selective blur performed on edges only. DOF is a similar (unimpressive) hack.

What will bring RT ray-tracing back to my interest, when it'll be possible to do is on GPUs.
Not because of the speed.
Because of the improved visual quality.

Looks like it will be soon. :)
 
Hyp-X said:
What will bring RT ray-tracing back to my interest, when it'll be possible to do is on GPUs.
Not because of the speed.
Because of the improved visual quality.

Gotta argue with you again: 95% of feature film CGI effects does not use any raytracing. Does it have an effect on the quality of their visuals? No :)
 
My question is would they use more ray tracing in movies if it didn't take so long to render? I'm sure they would at least in any area where accurate reflections are necessary.
 
Laa-Yosh said:
Hyp-X said:
What will bring RT ray-tracing back to my interest, when it'll be possible to do is on GPUs.
Not because of the speed.
Because of the improved visual quality.

Gotta argue with you again: 95% of feature film CGI effects does not use any raytracing. Does it have an effect on the quality of their visuals? No :)

It seems I wasn't really clear...

What I meant is the improved visual quality of doing RTRT on graphics hardware instead of doing it in software (on CPU).
I wrote this because a lot of people might think the big advantage of running on graphics hardware will be speed.

I didn't talk about whether it is better than raster rendering.

On the other hand, to answer your question: Yes I think it does. :)
 
3dcgi said:
My question is would they use more ray tracing in movies if it didn't take so long to render? I'm sure they would at least in any area where accurate reflections are necessary.

On of the problems of traditional raytracing is that it doesn't calculate accurate reflections. It can only handle perfectly shiny surfaces. A lot of the real-world surfaces blurs the reflection.

The other problem is having area lights or the fact that a lit surface is emits lights of its own. The latter is called "global illumination" (in the diffuse case) or "caustics" (in the specular case).

Distributing raytracing methods can solve these problems but they're way slower than the traditional method.

A radiosity algorithm can solve the diffuse case only. (And it is slow too.)
 
Nice to see FAN in action again. It's been a while...

My 933MHz Athlon TB does not really like this benchmark =)
 
2089 Raymarks...

640x480x32
- Shadows
- Reflections
- Anti-alias
- Depth of field

Average fps: 9.07
Minimum fps: 4.88
Maximum fps: 25

Running on an 1.59GHz Athlon TBird (145Mhz FSB).

Makes me lust for a dual AMd Hammer system even more...

--|BRiT|
 
More numbers to compare to the ~1200Mhz base numbers...

320x180x32
- Shadows
- Reflections

1884 Raymarks...

Average fps: 32.71
Minimum fps: 18.18
Maximum fps: 111.11


512x288x32
- Shadows
- Reflections

2260 Raymarks...

Average fps: 15.33
Minimum fps: 7.58
Maximum fps: 55.55


Running on an 1.59GHz Athlon TBird (145Mhz FSB).


--|BRiT|
 
Back
Top