Have you got any comparable benchmarks of JME3 vs JME2 ??
benchmarks are kinda useless because:
- usually speed depends on video card
- jme3 has better graphics and will drain more fps anyway
Momoko_Fan said:
2) jme3 has better graphics and will drain more fps anyway
jme3 magically makes your scene look better? awesome! :)
MrCoder said:
jme3 magically makes your scene look better? awesome! :)
Ardor doesn't have that button? :P
jme3 magically makes your scene look better? awesome!
Yeah it has per-pixel lighting built-in and with a few clicks here and there you get HDR for free too.
thumbs up 8)
MrCoder said:
*thumbs up* 8)
Glad you like it :P
Sounds like the kind of engine you would want to work on, doesn't it?
I've noticed that fps flop down when in fullscreen mode (from 50/60 fps to 10/15 fps) raising screen resolution (from 640x480 to 1280x800)
From what does it depend?
truman said:
I've noticed that fps flop down when in fullscreen mode (from 50/60 fps to 10/15 fps) raising screen resolution (from 640x480 to 1280x800)
From what does it depend?
Raising resolution will cause an increase in fillrate (e.g # pixels to process), which your video card might not be able to handle as well. What test are you running exactly and on what video card?
Momoko_Fan said:
truman said:
I've noticed that fps flop down when in fullscreen mode (from 50/60 fps to 10/15 fps) raising screen resolution (from 640x480 to 1280x800)
From what does it depend?
Raising resolution will cause an increase in fillrate (e.g # pixels to process), which your video card might not be able to handle as well. What test are you running exactly and on what video card?
I'm testing my car racing game on a nVidia geforce 6200 (agp bus)
bye
Momoko_Fan said:
benchmarks are kinda useless because:
1) usually speed depends on video card
2) jme3 has better graphics and will drain more fps anyway
1. Immaterial as tests will be on same system ( propose a modern card )
Momoko_Fan said:
jme3 magically makes your scene look better? awesome!
Yeah it has per-pixel lighting built-in and with a few clicks here and there you get HDR for free too.
2. Can per pixel lighting be disabled - presume you would want to for a HUD. Hdr, surely this also is configurable - eg for far away objects disable it.
You need to compare it to JME 2 to show you have actually improoved it and even beaten JWatt's threaded implementation. As soon as you can show its a marked improovement expect to have many projects converted.
Can per pixel lighting be disabled - presume you would want to for a HUD
Only in-game objects that you select will have lighting enabled, but lighting is, by default, per-pixel. The shader has a define-param called m_VertexLighting which can be used to disable per-pixel lighting but normal/specular maps will not be used.
Hdr, surely this also is configurable - eg for far away objects disable it.
It's not possible to disable HDR to specific objects because it is a screen space effect.
You need to compare it to JME 2 to show you have actually improoved it and even beaten JWatt's threaded implementation. As soon as you can show its a marked improovement expect to have many projects converted.
Not sure what you're talking about here. Never heard of "JWatt's" implementation and don't think anyone else did. A google search comes with some band singer :/
Momoko_Fan said:
Not sure what you're talking about here. Never heard of "JWatt's" implementation and don't think anyone else did. A google search comes with some band singer :/
Perhaps it was some of vears tests from a year or so that beat the out of the box JME2 version on performance. It wasnt J Watte at all, from searching back I now remember that he was persecuted for criticising JME2 although he had some valid points.
What about the quake level loader as a comparison between Xith and JME3 ?.
Just wondering, what does the prefix m_ stand for?
Just wondering, what does the prefix m_ stand for?
The prefix is used for the material scripting in jME3. "m_" stands for material parameter, "g_" stands for global parameter and "in" stands for input/attribute parameter.
What about the quake level loader as a comparison between Xith and JME3 ?.
Sure, if you have some free time feel free to port it to jME3. Although performance strongly depends on how the BSP data is handled. If all tris are merged into one mesh then fps is likely to be higher on new gpus. If lighting or advanced shading is used then fps will be lower. jME2 with VBO enabled should have the same performance as jME3 because you will be GPU bound, I guess Xith will be in that camp as well if they used standard VBOs. In jME3 there's a test that renders 400 teapots, I tried some optimizations on the mesh data and found that simply rendering as-is, will gain the highest fps. This is not the case with Android however.
Once again I will mention that CPU usage while simply rendering a static scene will be very low and entirely insignificant compared to GPU usage.