Recently, NVIDIA introduced Project G-Assist as part of an update to its NVIDIA App. As the company explains, G-Assist is a local small-scale artificial intelligence model that runs locally directly on RTX graphics cards. The first reviews of this innovation have already appeared online.
According to NVIDIA, G-Assist is supposed to be an assistant that diagnoses problems in real time, offers optimal settings for games, helps to overclock the GPU, and even controls fans. The company assures that the assistant understands natural language (currently only English), so you can ask questions and get answers.
Here is how NVIDIA describes the process:
«Under the hood, G-Assist now uses a Llama-based Instruct model with 8 billion parameters, packing language understanding into a tiny fraction of the size of today’s large scale AI models. This allows G-Assist to run locally on GeForce RTX hardware. And with the rapid pace of SLM research, these compact models are becoming more capable and efficient every few months.
When G-Assist is prompted for help by pressing Alt+G — say, to optimize graphics settings or check GPU temperatures— your GeForce RTX GPU briefly allocates a portion of its horsepower to AI inference. If you’re simultaneously gaming or running another GPU-heavy application, a short dip in render rate or inference completion speed may occur during those few seconds. Once G-Assist finishes its task, the GPU returns to delivering full performance to the game or app.»
But, as is often the case with AI, everything sounds great in words, but in practice, it’s a completely different story.
To use the new tool, you need to:
Therefore, NVIDIA Project G-Assist requires at least 12 GB of video memory. Although, during the recent CES 2025, NVIDIA stated that the video memory component is no longer critical due to AI optimization. This may be true for games, but not for their own AI tools.
On paper, the capabilities of NVIDIA Project G-Assist look good. The first users of the new tool tell us what happens in reality.
For example, one of the gamers on Reddit successfully used AI to underwrite a video card.
But there are other examples. Richard Devine tested G-Assist on a gaming PC with an RTX 5080 Founders Edition graphics card (with 16 GB of memory), and the results were simply terrible. Although this is one of the most powerful graphics cards available today, G-Assist was completely unusable while playing games. When you turn on the assistant, the FPS drops from triple digits to single digits. In the worst case, the system freezes altogether.
In Call of Duty: Black Ops 6, Atomfall, and Avowed, performance crashed as soon as G-Assist was launched.
To make matters worse, if you asked G-Assist a question or gave it a command, the system simply locked up while the «artificial intelligence worked on it. And that was too long. Even a simple question like «Is G-Sync enabled?» almost causes the game to freeze.
Richard Devine suggested that part of the problem might be due to the fact that G-Assist requires an NVIDIA overlay to work, which in itself can degrade performance. However, the actual performance drop is too great.
If even the RTX 5080 can’t pull off G-Assist while gaming, what can we say about weaker graphics cards?
NVIDIA says that G-Assist will improve as more people start using it. But how can you get positive feedback if the assistant doesn’t work properly? The only good news is that it can be removed quickly. If you don’t like G-Assist:
That’s all.
Source: windowscentral