Modded CMP 170hx runs Qwen 27B at 100+ tps — owner weighs whether selling for upgrades makes sense
JustinPooDough · reddit · 2026-11-04
The poster runs a modded mining card CMP 170hx alongside a 3090 on an old platform (Ryzen 5 1600, Asus B350-F, 64GB DDR4), getting 100+ tps on Qwen 27B locally. The card has doubled in price, and he's weighing selling it to upgrade CPU/motherboard/RAM for RAM+SSD-based inference of models like Strata or Maya. He leans toward 'no setup from the sale proceeds would outperform what I have now' and asks for input. Practical reference for local LLM hardware builders.
More from Infra
- Concurrency limits are the hidden tax in voice AI — how Phonic solved it — graceisford · 2026-11-04
- AMD beats guidance by just 2.9% as blowout era peaks; supply chain fungible between PCs and servers — tengyanAI · 2026-11-04
- Fireworks AI VP: Nexus Router Is Itself a Trained Model, Not Rules — sophiamyang · 2026-11-04
- Sipeed's 10cm MaixStation Dock claims DGX Spark-class local AI inference, runs off a power bank — vincent_koc · 2026-11-04
- vLLM, Red Hat and NVIDIA to host open-source inference meetup in Madrid on Nov 12 — vllm_project · 2026-11-04
- vLLM v0.33.0 ships with 927 commits, 1.5–2x faster decode on GB300 — vllm_project · 2026-11-04