NVIDIA has made a significant announcement, enhancing its RTX Video Super Resolution application while providing additional capabilities for the LLaMa 2 and Stable Diffusion models.
To start, RTX VSR version 1.5 has been upgraded with a better-than-ever algorithm, resulting in improved artifact removal and upscaling quality. What’s even better is that it now supports RTX 20 series GPUs. Gone are the days when only specific cards could utilize this feature – now it’s accessible for all RTX cards.
Moreover, the latest version permits the model to operate at the display’s native resolution rather than the video source input. This significantly reduces instances of fuzzy and blurry images.
This update is readily available to all users through the latest GeForce Game Ready Driver update, with the Studio Driver set to receive it next month.
The official video has more side-by-side comparisons so go check that out.
Meanwhile, NVIDIA’s TensorRT-LLM acceleration software stack for Windows will soon be accessible on the NVIDIA Developer website. It also supports local acceleration, delivering up to 4x performance.
On the other hand, the image-generating algorithm, Stable Diffusion, will now operate at twice the speed, thanks to TensorRT’s enhancement, achieved through layer fusion, precision calibration, kernel auto-tuning, and other capabilities that significantly boost inference efficiency and speed.
Interested personnel in the AI image generator scene may try it out over here.
