User turns GTX 1060 laptop into Proxmox AI server with llama.cpp
The author converted an old gaming laptop equipped with a GTX 1060 into a local AI server running on Proxmox, using a Vulkan backend for llama.cpp to perform inference tasks. The setup efficiently runs the Gemma 4 E4B model at over 30 tokens per second, powering self-hosted FOSS services such as automatic tagging, bookmarked website summarization, and OCR generation. To prevent hardware degradation, the author implemented a Home Assistant automation to manage the laptop's battery charging state and kept the lid open to avoid overheating. Open article →