New comment by DAFtwinTurbo in "AI on a Budget: Recompiling Llama.cpp for Qwen3.5 Inference on an HP Z440"

DAFtwinTurbo — Tue, 10 Mar 2026 07:54:35 +0000

Hi hackernews,

I wrote a small blogpost on a little experiment I did last week-end. The goal was to see if I could get more tok/s performance from llama.cpp running the latest Qwen3.5 models. A 5.5x perf increase was achieved by 1. recompiling with optimization flags 2. using ik_llama.cpp!

The build I'm using is the 750 USD rig from Digital Spaceport.

Very cool to see more than usable speeds on a cheap setup like this.

AI on a Budget: Recompiling Llama.cpp for Qwen3.5 Inference on an HP Z440

DAFtwinTurbo — Tue, 10 Mar 2026 07:54:35 +0000

Article URL: https://jeanbaptistefleury.neocities.org/importance_of_inference_engines

Comments URL: https://news.ycombinator.com/item?id=47320244

Points: 3

# Comments: 1

Hacker News: DAFtwinTurbo

New comment by DAFtwinTurbo in "AI on a Budget: Recompiling Llama.cpp for Qwen3.5 Inference on an HP Z440"

AI on a Budget: Recompiling Llama.cpp for Qwen3.5 Inference on an HP Z440