[GPU Computing] Fractron 9000: GPU (OpenCL) Fractal Flame Renderer

Fractron 9000: GPU (OpenCL) Fractal Flame Renderer
Microcosm sample

Fractron 9000 is a GPU accelerated fractal flame renderer. It can generate fractals using OpenCL or OpenGL+CPU. Fractals can be saved in PNG or JPG files. Fractron 9000 is based on the fractal flame algorithm:

The Fractal Flame algorithm is a member of the Iterated Function System (IFS) class of fractal algorithms. A two-dimensional IFS creates images by plotting the output of a chaotic attractor directly on the image plane. The fractal flame algorithm is distinguished by three innovations over text-book IFS: non-linear functions, log-density display, and structural coloring. In combination with standard techniques of anti-aliasing and motion blur the result is striking image variety and quality.

You can download Fractron 9000 v0.4 HERE.

I tested the tool on Win7 64-bit + R258.19 + GTX 480. Works fine!

Fractron 9000: GPU (OpenCL) Fractal Flame Renderer
Manta sample

Fractron 9000: GPU (OpenCL) Fractal Flame Renderer
Bubble wrap sample


8 thoughts on “[GPU Computing] Fractron 9000: GPU (OpenCL) Fractal Flame Renderer”

  1. applejack

    Fractron 9000 also support CUDA for GPU acceleration.
    the OpenCL implementation is a new feature.

    would be interesting to compare performances in Fractron 9000 between CUDA and OpenCL.

  2. WacKEDmaN

    concentric lite

    GTX 470..
    openCL 308.97M dots/sec
    CUDA 292.04M dots/sec

    openCL 16.12M dots/sec
    CUDA 13.00M dots/sec

    Fractron actually has a performance rating built in when you select the driver..

    OpenCL FTW!!

  3. Lokavidu

    this is awesome!!! power of opencl being unleashed! i was hoping that some gpu accelerated version of apophysis would appear, and here it comes.. awesome! looking forward new versions.. and btw strange that opencl is better than cuda since on nvidia, opencl runs on cuda, right?

  4. fellix

    HD5870 @ 900/1250MHz — 135M dots/sec.

    Looks like the OCL kernel in this beta is not vectorized.

  5. Mike Thiesen

    Hi, I’m the author of Fractron 9000. Thanks for the writeup, I’m glad people are noticing my app!

    I spent a bit more time performance tuning the OpenCL code than I did for CUDA, so that’s probably why the OpenCL mode is faster. In the future I will probably just drop support for CUDA so I don’t have as much code to maintain.

  6. Xzendor7

    I’ve been using Fractron 9000 for roughly 3 weeks now, and I am impressed this is the only fractal generating program I’ve come across that produces real time results.

    I can easily say that OpenCL is the future of graphic software especially when both the CPU & GPU are working hand in hand.

    Can’t wait to see more vendors port their graphic applications to run under OpenCL.

    Looking forward to seeing what Fractron 9000 will really do when Mike finishes with the Beta’s…

  7. Lokavidu

    @Mike thx for answer, awesome app anyway! :)
    btw i prefer opencl, but since got gts 360m (desktop gt 240) i tried CUDA too and this is interesting: opencl has rating 116 (or smt), cuda has rating around 111 but real life results are:
    cuda: 52M dots/s (and is getting to peak value faster)
    opencl: 48M dots/s (looks like not really significant performance drop, only 2x of core i7, haha)
    core i7 720QM: 2,3M dots/s (lol)

    but looking at results above.. WOW.. 308M dots/s.. that must render sooo fast :)

  8. Angus

    It’s shit. 3 GPU flame renderers appeared within a few months of each other and none of them do doodly squat and were never developed into anything properly usable. I’m starting think I’m the only person in the whole world who wants to render giant fractals in this lifetime.

Comments are closed.