Minecraft/voxel - performance

This question as about modern hardware capabilities and voxel algorithms.

I neved did voxels, also I was doing 3D almost 10 years ago. I look at Minecraft and I wonder. How it is done? I read a tiny bit about voxels and they say you should somehow exclude the interior and leave only the outer layer, but when I see these interesting errors in Minecraft (like seeing all caves below me) I get an impression they simply render all cubes that are within the “cone of view” of the player, without any exclusion of what is behind what… But that would mean at least 100x100x100 cubes (which is 1 million, and there is minimum 7 vertices per cube). Can you really push 7 million vertices on modern hardware (with texture coords and colour)? Or have they reduced it first via algorithms?

The first, and most important optimization:
Any Block wich is encloded by 6 solid (non transparent) blocks does not need to be rendered.
(-> the 6 sides of a cube)
As it can never be seen

This cuts down the number of voxels by about 90% already.

Then of course dont render Polygons / sides wich are attached to another solid block.
(wich implies the above naturally)

the 100x100x100 block chunk in a perfect scenario can be rendered with these simple optimizations with just
2100100*3 Triangles -> 60000 tris