correction:
float2=8Byte
128*128*128*6*8=96mb.

with voxels maximum voxel count reach if you put voxel in every other position and skip one, that will bring the buffer sizes to:
128*128*128*6(vertices for a quad)*16(size of a float2)=192mb for 2,097,152 voxels.
If you have read this far check Voxy at:
www.fab.com/listings/4ce...

I am using for all my compute shaders.
For the position buffer, I am using a float4 for now, xyz is the position, and w is face and neighbor data, but since my max volume size is 256x256x256, I can pack the position data into a float2 and still have 16bit extra for more../🧵

and IndirectArgs that is been calculated in the neighbor count compute shader will set the max vertex count for me.
As for optimizing the buffer sizes, now I skip passing the Scale and Color vertex buffer and directly sample voxel data from the volume texture that I am../🧵

Previously I was calculating number of polygons then stall the frame to readback that value, and reallocate the vertex buffers I need, stalling frame cost about 5ms to 10ms which is a lot, but now since I using indirect draw I pre-allocate the maximum size for a volume and ../🧵

check this playlist to see other features I have added to Voxy
www.youtube.com/playlist?lis...

FYI: this specific feature is coming on next update soon.

yes, if you pre-create the asset (not realtime generation) it already has the collision, because collision generation takes about 40ms which make hiccup if you do it realtime, but plan is to fix that soon, and this is a single volume you can have as many as you want.

also I am new here, help me grow!