Dual Purpose: Optimizing a 7900 XTX Gaming PC as a Low-Power LLM Node

I recently built a new rig intended to serve two distinct functions. It needs to be a high-end gaming machine capable of crushing titles at 4K, but I also wanted it to serve as a headless inference node for my local LLM stack, powered by Ollama. The challenge is balancing Power and VRAM. My GPU, the Radeon RX 7900 XTX, features a massive 24GB of VRAM—excellent for local LLMs. However, high-end AMD cards can have high power draw at idle, and running an LLM server 24/7 on a “High Performance” power plan is inefficient.

• llm
• ollama
• hardware
• powershell
• homelab
• amd