Optimal MSAA technique for a deferred renderer

I got multisampled textures working with my deferred renderer. In this thread, Pierre Boudier says the following:

if you are never going to use a surface as a texture, we recommend that you use render buffer. this allows the implementation to make better optimizations for those surfaces (compression or hiz for depth, compression for msaa surface


I don’t know what he is referring to as a “surface” or “renderbuffer”. Is there another more optimal MSAA technique I should look into besides multisampled textures?

I don’t know what he is referring to as a “surface” or “renderbuffer”.

By “surface,” he means “image”. IE: what you’re rendering to.

A Renderbuffer is an OpenGL object that can be used as a render target, but cannot be used as a texture. Since you’re using deferred rendering, and therefore need to use your render target as a texture, this seems unlikely to help.

Renderbuffers are useful for things like depth buffers (that you don’t intend to read from. Strictly for FBO rendering that needs a depth buffer) or for render targets you intend to use as source data for a blit operation.