# Upgraded Your GPU? Don't Sell the Old One. Put Both to Work Running Local LLMs. — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/dbae7d3c815f2f5d

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-11-08 16:33 ingested 2026-11-09 02:37

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1xr2ici/upgraded_your_gpu_dont_sell_the_old_one_put_both/)

## retained text



```
Friendly reminder: If you upgraded to a newer GPU, don't let your old one collect dust! You can run both GPUs in the same rig and split LLM inference workloads across them using tools like llama.cpp. I'm doing exactly that with my RTX 5080 and 3080, combining their VRAM to run larger local AI models.
```



document fetched 2026-11-09 00:36 · canonical https://reddit.com/r/LocalLLaMA/comments/1xr2ici/upgraded_your_gpu_dont_sell_the_old_one_put_both
