# gpu-yield 1.13.0 adds BC-250 support

> The gpu-yield script, which frees the GPU from a local AI model when a game starts, adds BC-250 support in version 1.13.0.

- Published: 2026-11-03T06:34:29Z
- Score: 7/10
- Tags: bc250, ai, tool
- Port / platforms: - -> PC (Windows), Linux / Steam Deck (Tool)
- Web page: https://emu-radar.customfw.xyz/news/gpu-yield-bc250-support/
- Sources: https://github.com/AnonyMap/gpu-yield/commit/26818541ebcb349528e1703e66f59ba20b12050c

gpu-yield, a single Python script for Linux gaming PCs, supports the BC-250 board in version 1.13.0. The commit history for this version adds the board and shared-memory chips to the tool. The commits include review fixes, indicating ongoing work.

The tool serves users who run a local AI model on their gaming PC. Ollama keeps the model loaded while the PC is idle. When a game starts, gpu-yield stops Ollama within about 2 seconds to free VRAM and RAM. After the game closes, it waits 45 seconds by default before reloading the model. It runs on [Bazzite](https://bazzite.gg), requiring Python 3.11 or newer and an AMD GPU.

To install it, run these commands in a terminal:

1. `curl -fsSL https://raw.githubusercontent.com/AnonyMap/gpu-yield/main/gpu-yield -o /tmp/gpu-yield`
2. `python3 /tmp/gpu-yield setup`

The setup script installs files into the home folder and uses sudo only for firewall and Wake-on-LAN configurations. The README section covering the BC-250 is incomplete in the available source material, so board-specific setup details remain unclear.
