In short terms: it isn’t slow 🙂
Compared to an native application it is slightly slower but there should be not that much difference. (This is true at least since version 0.42 of JWildfire at which I have tried to optimize the hell out of it ;-)) For the maximum performance you should use Java 7.
I have created some flames for benchmarking. I used Apophysis7X15 as reference because it is very popular. Below are the results (the used *.flame files are linked inside the table):
Render settings: Resolution: 1920×1080, Filter radius: 0.25, spatial oversample 1
- Benchmark01Â
quality 100 quality 500 Apophysis7X15 14.99s 72.61s JWildfire V0.42 18.27s 88.53 - Benchmark02Â
quality 100 quality 500 Apophysis7X15 5.99s 28.04s JWildfire V0.42 8.86s 41.632s - Benchmark03Â
quality 100 quality 500 Apophysis7X15 10.69s 51.06s JWildfire V0.42 12.80s 62.51s
Hi Andreas,
Tom Allen has sent me to your page. I am a software developer myself (15+ years C++ and 7+ years of C#) and besides being interested in fractals I was also interested in how my new dev system I just assembled would perform. It’s got a i7 3930K running at 4.2GHz and more than enough RAM (32GB). I was curious to see if J-Wildfire would make use of all 12 threads. And indeed it did! I don’t quite understand the conditions for the times you cited, tho.
Here’s what I did (all today):
Downloaded Java 7.
Installed J-Wildfire. I didn’t change anything else. All default settings.
Copied content of Benmark01.txt into clipboard. Hit FromClipboard button and the outline of the image appeared.
I hit “Render Image (normal)” and a 800×600 image appeared in maybe 3.2 seconds. Then I hit “Render Image (high quality)” and now it took 165.4 seconds, which shows a vastly different ratio to your 88.53 seconds than your 18.27 to my 3.2. I couldn’t find a setting “quality 100” or “quality 500” and I assumed that’s what the 2 Render button were for.
Could you please clarify?
Hello Harald,
the two buttons are just for choosing between two render sets you define in the preferences window. The difference you measured depends on the following properties: spatial oversampling (improves quality and increases rendering in time in quadratric terms), color oversampling and quality (both with linear dependency).
The “normal” setting is intended for rendering images which give a good overlook with most of the details, the “high” setting is for final images you want to publish or want to use to impress your girlfriend 😉
In the case of the benchmark I just used the “normal” set and switched the “quality” property between 100 and 500.
Hope this helps and you enjoy the software 🙂
Best regards,
Andreas
Hello,
You can also boost the application by setting Jwildfire with highest priority.
Under a mac or linux:
sudo nice -n -20 sh start_XXX
or change the java call in the start file:
sudo nice -n -20 java …
Under windows:
it’s via Task manager.
++
Andreas,
The facts speak for themselves and you can probably squeeze some more out of that. But the quality of the rendering with TINA is better and that has to count for something.
tom allen
Facebook J-Wildfire Users Group
Thank you very much and I hope you enjoy the new release 🙂