Low fps problems with many objects in OpenGL

WIP games, tools & toy projects Cube World Projects
#1

Hi everyone, I wrote a simple “game” using LWJGL that allows me to modify a chunk of 32x32x32 blocks ust like in minecraft. It has ray picking and lighting. Here’s a screenshot of what I can do with it :

Alright, now here’s my problem. My code is fairly good, I’m using vertex buffer objects, one for each block. The chunk is never really re-rendered, because each block is an individual object and can be rendered alone. The blocks are also only rendering faces that are visible, and they tell the 6 blocks (if existing) to hide a face for faster rendering.

Now, I’m only getting ~100fps with a 32x32x32 block that only contains 32x8x32 blocks really, the rest is air. Why am I getting so low fps? Anyone has proposition?

Here are the system specs :

i5 760 2.8ghz
8gb ram 1333
gtx 580 1.5gb
intel 80gb ssd

I can post parts of my code if anyone wants it to help me!

Thanks a lot!

#2

Wouldn’t it be better if you used a VBO to hold more blocks? Single blocks seems like too litle :confused:

#3

I know minecraft uses something like 16x16x16 blocks chunks, but won’t it be slow to re-render if a modification happens? maybe a thread? I’m lost ! but thanks for the idea, gonna try that for the next few days

#4

Store all the blocks in one chunk in one VBO.

Look through the recent posts in LWJGL Blockworld and Voxel - A start for more info.

#5

And no, it is not slow to rebuffer the chunks (if they aren’t too big)

#6

Alright I’ll take a look, is this a board on this forum?

#7

Go to the index page and scroll down to the recent posts section.
Then find the 2 topics I mentioned.

#8

that is a pretty beefy system so you should get more fps, but yeh one VBO per chunk of blocks is probably the way to go. VBO is only updated if there is some change to the chunk.

Also curious, what type of lighting are you using simple per pixel/vertex spotlight? SSAO? other?

#9

Ok I’ll check out these posts… Last question if someone can help, the only thing I’m not entirely sure about is the way my chunk is supposed to update… I’m gonna be doing 16x16x16 chunks, but I’m working with FloatBuffer and the predefined .put(float values) function… I have to tell my FloatBuffer what size it’s gonna be before even knowing it! is there a workaround? Perhaps I should be using float[][] and then getting float[] from each block, and then put them manually in the FloatBuffer? That seems slow to me!

(For lighting, I use the OpenGL predefined glLight set to a color! I love it and it’s simple… no shadows though, I’m gonna have to work on that some time soon!)

#10

The buffer only needs an max size, not an exact size, so you could just make it the maximum size of possible vertexes (or precalculate it).
You dont have to worry about functions benig slow if you only use them when updating, so just make it something good to understand and stable as possible.

#11

Wow thanks ok I’m gonna try that out right now and I’ll give you some feedback once I’m done!