I think I call it Pixel Meshing, or how to 2miljon voxels 60fps. [new v3 update]

Maybe they should long triangular wedges of the ground for each scene.Sort of like a pie chart.

The worst case scenario would never happen with this system because it is relying on many scenes
that are switch to to display the ground.

And how much ground are these cameras displaying?
The cameras still need to render the voxels, which is the worst case with multy scene/camera overhead, so even worse.

Which are displayed on a plane from a camera in the scenes.The overlay scene camera in every scene is essential.

So you are surrounding the player with video texture screens?
Thats atleast 3-6 times more rendering needed than just displaying anything.

Giving the illusion that all the ground is displayed at ounce.And it is my plan for the birds eye view of the map.

this boils down to

  1. Create scenes with cameras
  2. Render a lot of things
  3. ???
  4. Profit

It kinda seems like a troll profit joke from 2010…

At no point are you saying how you represent MORE voxels with LESS performance.

Maybe start your own thread where you give examples and explanations in more coherent manner.

It is just a plane that has four vertexes parented to the camera with videotexture of the scenes.It combines the images of all the scenes together.I will start a new thread.

3036 cubes for every scene.With hills a little more.Dynamically load all 3036 in the direction
your looking at then switch scenes one after the other.And keep the camera letter box.
Here is a screenshot of one of the scenes.


What?
Do you even…

First of all the game engine allready hides all the objects that are not in view.
I mean when you look away from the cubes, the framerate goes up.

Each object has a bounding box which when not in view, disables the rendering.
You can even move the bounding box to see how the mesh disappears when the box is not in view.

That your solution done better since the dark ages. And it runs on the gpu!

And what do you do when I would want to see further than 3000 units/cubes?
This would be too much load for a single camera, not to mention multiple.

With that method you are still crippled by the fact that each of the 3000 cubes needs a draw call.
Its not even the polygon count, its the number of draw calls that kills blender.

And i asked to open your own thread, I don’t really think this discussion is productive in shadow of this thread.

Check out main post.
New features and examples.

Its amazing how much this approach is easier to manage than other voxel optimizing options I have tried.
#pineapple60fps

If it weren’t for bge horrible allergy to abuse, this would make a decent resource.
(libload async crashes when I dont run spawner function( which libloads and creates new textures etc), go figure :spin:)

When i look in certain direction with the camera the framerate goes down to 39.When i look at the ground the framerate goes 60.I guess it would be best to play it in letterbox.If you put occluder cube around the area you see out of.I suspect that if you added the ability to dig the framerate would go down even worse.

You mean when you look at a part of a map that is more complex your framerate changes?
Sounds totally obvious to me.

The idea of letterbox is a new level of #pineapple60fps.
Its the year 2016, with people having 2k 165Hz ultrawides and you suggest going back in time 10 years to letterbox.
Good job.
Its not the shape that gave you more framerate, it was the reduction of resolution…

Stop suggesting occluders when you don’t understand how they work.
There is nothing fully obstructed -> no use for occluders.

But as I am so nice, here is an actual article talking about custom occluding in Minecraft:

It is the 2nd article, the first doesn’t have as nice of interactive demonstration.
So basically pathfinding for chunk visibility. Seems logical to me.

Half of this video is actually frustum culling which bge already does.

I think you would be better off 4000 planes 2000 planes vertical and 2000 planes horizontal.And just making them transparent where there arent any cubes in game.And where there are cubes make them opaqe in game.Then just spawn the ground near the player.You would get better framerate.That would represent 40,000 cubes.

First 40 000 is 0.8% of 5 million, which I can currently push out.
Thats literally 100 times worse…

Now did you even understand that I represent ‘faces’ as pixels on a texture.
I don’t need to have 2000 faces for 2000 cubes (top faces), I only need 1 face!
If I want to hide a ‘face’ I simply make that pixels alpha 0.

Finally, again, no idea what you are blabbering about.
Do you mean having instead of chunks having huge planes for the whole map?

Currenlty each chunk has a 128x128 texture that needs to be refreshed every time its voxel is changed.
That is fine.

But for your idea I would have to refresh a texture of 2kx2k which would not fit into 16ms.
Also what would happen if multiple 4k planes are visible? I think the engine can more easily (frustum) cull the visible textures when they have contained dimensions, or in other words chunks.

I mean having huge planes for the whole map.
For my idea that would be 1,600 vertexes not including the ground near the player and the ground near the enemies.The same that would happen with 2000 cubes visible.You would get a framerate 60.

You didn’t read, you didn’t think, and you #pineapple60fps -ed again.

Learn to express yourself, please!
You are leaving half of what you ‘think’ out while writing everything into one long cluster flock of a sentence.

What are the planes doing?
Is one plane representing 1 voxel face or is it shared over multible?
How are you turning the represantation of voxel visible/invisible?
By changing the texture, a vertex or the whole plane?

How much data needs to be processed for each plane?
How do you plan to enable frustum culling when the planes are almost always in the view because of their SIZE.

If you are not talking about using pixels as faces, then this is the wrong thread!
My bad if it seemed otherwise.

Is one plane representing 1 voxel face or is it shared over multible?
It shared over multiple.
How are you turning the represantation of voxel visible/invisible?
Drawing on the plane and erasing on it.
By changing the texture, a vertex or the whole plane?
By changing the texture or color from
transparent to opaque in some area’s.
How much data needs to be processed for each plane?
A plane’s worth of data plus its texture.
How do you plan to enable frustum culling when the planes are almost always in the view because of their SIZE.
No fustum culling just oclussion culling.
Here is my thread for it.https://blenderartists.org/forum/showthread.php?404076-This-would-be-great-for-a-openworld-game-with-digging-in-it-to-keep-the-framerate-60