Nvidia introduced its own alternative to popular chatbots based on artificial intelligence – Chat with RTX. As the name suggests, it uses the computing capabilities of GeForce RTX graphics accelerators. The main feature is that it works directly on the user’s computer, along with the ability to personalize a large language model.
Chat with RTX uses Nvidia’s TensorRT-LLM library and RAG (Retrieval Augmented Generation) to fill knowledge gaps in a large language model. Users can add their own notes, documents, videos and other data, such as getting an answer based on information from YouTube. “And because the application runs locally on a Windows PC or workstation with RTX, results are generated quickly and securely,” Nvidia emphasizes.
You can download Chat with RTX from the official Nvidia website. The system requirements mention GeForce RTX 30- and 40-series adapters equipped with at least an 8 GB video buffer, that is, the recently released GeForce RTX 3050 6GB will not work. You will also need 16 GB of RAM, the Windows 11 operating system, and driver version 535.11 or later.