Inquiry about DLSS efficiency at ultra-low resolutions (300x300)

At a target resolution as low as 300x300 , does DLSS provide a meaningful performance or VRAM advantage? I am concerned that the internal overhead of the DLSS plugin (AI models, buffers, etc.) might actually result in a net loss compared to the gains from lower internal rendering.

In summary, I’m questioning whether applying DLSS to ultra-low resolutions is truly beneficial or even practical.

Somebody did that experiment before:

https://youtu.be/DKCyk3CeUFY?si=Ipy2iRwd7kd-16Ju