Benchmarking GPU pipeline stages

Content here was originally posted on linkedin.

Here is the PDF with the slides to download

For the benchmark I used
Unity 6000.2.9, custom render pipeline, custom engine loop, DX12, DXC shader compiler
Nsight Graphics 2022.7.0 and Nsight Graphics 2025.4.1
RTX 3060 and GTX 1050 Ti

Benchmarks were done on Windows 10, Unity release build with DX12 and IL2CPP:Master.

Enjoy the slides:

Download the PDF slides here

See it being used in practice

In the articles below I use this knowledge about the GPU hardware in practice:

I made this shader run 5x faster

Implementing the tornado VFX and making it render 13x faster

How to use native GPU profiler with Unity editor