Nvidia Releases Project G-Assist AI Gaming Assistant That Runs on GeForce RTX GPUs

Nvidia announced the release of Project G-Assist on Tuesday. The new feature is an artificial intelligence (AI) assistant that runs locally on GeForce RTX GPU-powered desktop PCs. The current release is an experimental version of the chatbot, and it does not support RTX GPU-powered laptops. Focused on gaming, the AI assistant can control a wide range of PC settings, such as optimising game and system settings, charting frame rates, and controlling peripheral settings via text and voice prompts. It was first unveiled as a tech demo during Computex 2024 in June.

Nvidia's G-Assist Debuts

In a newsroom post, the tech giant announced the release of an experimental version of G-Assist. The AI assistant is a personalised gaming assistant for GeForce RTX AI PCs and can run locally, so the user data never leaves their device. It can answer queries about video games and can share information, such as gaming strategies. It can also analyse replays of multiplayer games and offer feedback to the user.

G-Assist is powered by a specially tuned small language model (SLM) that can interpret natural language commands as well as call a variety of Nvidia and third-party PC application programming interfaces (APIs) to execute specific actions on the user's behalf. It accepts text and audio inputs and can also process visual information via computer vision capability.

Coming to system-specific capabilities, the AI assistant can run real-time diagnostics and suggest recommendations to fix issues. It can also handle complex tasks such as improving power efficiency, optimising game settings, and overclocking the GPU.

Project G-Assist can also monitor and display performance metrics of the device, such as frames per second (FPS), latency, GPU usage, system temperature, and more. Additionally, it can also handle peripheral tasks such as running benchmarks, adjusting fan speeds, or changing lighting (on supported devices).

The AI assistant is built on a Meta Llama-based Instruct model with eight billion parameters. Nvidia said that despite the size, Project G-Assist is fully localised within the device and can be powered by a single RTX GPU. This also means that the chatbot comes without latency issues, data privacy risks, or any subscription fee.

It is currently available via the Nvidia app's Home tab in the Discovery section. It currently supports only the English language. Once it has been installed and activated, users can access the shortcut Alt+G to summon the chatbot quickly.