llamaperf

Qwen3.8 27B

on 2× NVIDIA RTX 3080 20GB · 180,000 ctx

Tone: positiveNVIDIA hardware
Sep 12, 2026
Throughput
39.0 t/s gen
Quant
Q4
System RAM
64 GB
VRAM reported
40 GB

Summary

User reports Qwen3.8 27B at 39 t/s on dual RTX 3080 20GB GPUs. Setup uses Q4 quantization with 180k unquantized context on a home server built for just under $2000. The user notes the cards have coil whine during decode and that the motherboard could support up to 4 GPUs at 16 lanes or 7 GPUs at 8 lanes.