Nvidia's fortunes have surged in recent years, largely due to the immense value of its AI-accelerating GPUs. While many users employ Nvidia GPUs primarily for gaming, there’s potential for dual usage. Nvidia has recently launched its experimental G-Assist AI, which can run concurrently to assist users in optimizing their PCs and enhancing their gaming experiences. Although it boasts some impressive capabilities, Nvidia emphasizes that this tool is still experimental. G-Assist can be accessed via the Nvidia desktop app, featuring a floating overlay. Once activated, users can interact with G-Assist by typing or speaking to inquire about system statistics or adjust settings. You can pose fundamental questions, such as "How does DLSS Frame Generation work?" and it can also manage certain system-level configurations. By invoking G-Assist, users receive insights into their system's performance, complete with custom-generated data charts. Additionally, the AI can modify settings, such as optimizing configurations for a specific game or toggling specific features. G-Assist even has the capability to overclock your GPU, accompanied by a graph illustrating potential performance gains. Last year, Nvidia showcased G-Assist with impressive features related to active gameplay. That version could analyze your activities and offer suggestions for achieving your next objective. Unfortunately, the public version's game integration is limited, supporting only a few titles, like Ark: Survival Evolved. On a positive note, G-Assist is compatible with several third-party plugins, allowing it to control peripherals from brands like Logitech G, Corsair, MSI, and Nanoleaf. For example, it can adjust thermal profiles via your MSI motherboard or modify LED settings through Logitech G. As PC manufacturers rush to introduce AI-powered laptops, Nvidia often underscores that computers equipped with genuine GPUs are the premier AI machines. However, opportunities to leverage this capability have been scarce, with most AI applications relying on cloud processing.

While Nvidia previously launched the general-purpose ChatRTX app, G-Assist is specifically tailored for gamers, who are likely to possess robust GPUs. Nvidia states that G-Assist utilizes a Small Language Model (SLM) optimized for local operation. The default text installation occupies 3GB of storage, while adding voice control increases this to 6. 5GB. To use G-Assist, you need a GeForce RTX 30, 40, or 50 series GPU with a minimum of 12GB RAM. The performance of G-Assist improves with more powerful GPUs. Although support for laptop GPUs is planned for the future, many may not be sufficiently powerful for effective use. While G-Assist operates on your GPU rather than in the cloud, that same GPU is likely being used for gaming. We tested G-Assist with an RTX 4070, and interaction with the model significantly spiked GPU usage. Generating outputs from the model can slow down performance, which may hinder gaming experiences. Outside of gaming, G-Assist performs much faster, but if you plan on using it frequently, a very powerful GPU is advisable. Currently, G-Assist’s performance is slow and somewhat glitchy, making it less practical to rely on for immediate needs. Tuning most system and game settings manually is generally quicker. Nevertheless, it represents an intriguing step towards leveraging AI models on personal hardware. Future GPUs may reach a point where they can simultaneously run games and language models effectively, but for now, G-Assist remains an interesting experimental tool.