Post

@sudoingX Better than Mac Mini for running a local models ?
English

@sudoingX I got 7900 XTX , it’s a beast. 24GB vram, bandwidth almost same as RTX4090 for third of the price and everything is working in Linux. You don’t get CUDA but you don’t really need it to run QWEN models
English

@sudoingX looking forward to reading benchmarks on the Framework's Ryzen MAX 395
English

@sudoingX Ended up getting 7900 XT(used RTX 3090 is just too expensive in my area). Got Hermes running on it with Qwen 27b (Unsloth's Q4_K_M GGUF) with 64k context. Works fine, VRAM stays at 95%. Responses do take longer when Hermes compresses. Wondering what else I can tune up to fix that
English

@sudoingX We have AMD 192 Core CPU coming online early this week. AMD indeed ❤️🙌
English







