Running Qwen 27B on RTX 3060+2060 Yields Only 5-6 TPS
sheriffoftiltover · reddit · 2026-08-26
A user ran Qwen3.8-27B (Q4KL) on an RTX 3060 + RTX 2060 setup with 52GB DDR4 RAM, getting only 5-6 tokens/s. Memory headroom is too small for reasonable context lengths, making it fun for experiments but impractical. They ask whether specialized engines or optimizations exist for older GPUs and low-VRAM setups, mentioning ninfer discussions around the RTX 5090.
More from Infra
- How Microsoft optimized Windows for Nvidia's Arm-based RTX Spark AI laptops — tomwarren · 2026-10-15
- How Microsoft optimized Windows for Nvidia's Arm-based RTX Spark laptops — The Verge AI · 2026-10-15
- Global smartphone shipments plunge 22% in worst quarterly decline on record as AI demand fuels memory costs — Polymarket · 2026-10-15
- SK Hynix drops 11% again, now below 3x FY2027 earnings as AI memory selloff deepens — JOBhakdi · 2026-10-15
- ASML Sells Zero Lithography Machines into Europe as US Hosts 75%+ of AI Compute — NinaDSchick · 2026-10-15
- Nvidia quietly reclassified SPCX as 'Hyperscale', signaling deeper neocloud reliance — RihardJarc · 2026-10-15