Creating a Stencil Mask to prevent pixels to be rendered behind them

[JulienActiveMe]

Hello,
In the constant fight for optimization, we came across a potential apporach that would maybe benefit us.
We noticed when doing screenshots that there is a rather important number of extra pixels rendered in the FOCUS3 that are not visible to the user.
It seems like is about an extra 10% pixels rendered that would not be "necessary".
Is this correct ? Is there 10% of rendered pixels that are not visible to the users eye ? 
If so, why is there so much headroom ? 

To be able to work around this, I'm trying to create an "elliptical mask", inside of which the pixels would be rendered, outside of which nothing would be rendered.
It should be slightly bigger than the real FOV of the player,  so that the player does not see is, but small enough to help lighten the GPU workload by having less pixels rendered.
So far, I'm not able to achieve this.
My first approach was just a plane at the near camera's clipping plane, with a custom shader on it that would draw an ellipse that would be transparent in the center and the rest would be opaque and black.
I do not think it stops the objects behind the mask to make draw calls to be rendered.

 

I also tried having two cameras, one rendering only this mask, and the other on top rendering the whole scene.

This does not work at all, It's either all black or either all clear. I do not seem to be able to blen the image of the two cameras together in one.  

Now I'm trying to use a scriptable Renderer Feature, but it doesn't seem to work. I think the result is about the same as my first approach, since when I lower the alpha on the mask, I see the objects behind the mask. So they are rendered and the first intent of the feature (lighten the GPU workload) is not met.

Has anyone ever achieved this ? 
Is this a valid approach ? Is there something else I could try ?

Thanks in advance ! 

 

Edited October 28, 2024 by JulienActiveMe
hit enter accidentally ;)