Only did and do floating point for image processing. In fact, looking into my logs, I registered 5 bugs with NVidia in the last 2 years, none with AMD.