It is amazing how much of the pro drivers rely on the host CPU. I’ve seen several benchmarks where the test result is within 2% across several cards within the same generation or even similar to previous generation. The performance does vary greatly if the task is more like a 3D game not just a wireframe. It still leaves a lot of potential improvements.

The extra work to add the app accelerators/OpenGL and certify compatibility with pro apps is added to the price of the pro cards. That way they can reduce the cost to average consumers who do not use the pro software and recoup costs from the pro users.