Ninfer Benchmark: 5090 Doubles Throughput for Qwen3 27B
Rollingsound514 · reddit · 2026-08-28
The author reports benchmark results running Qwen3 8 27B (nvfp4) on an RTX 5090 using the Ninfer engine. The setup reportedly more than doubles throughput compared to llama.cpp, peaking at 220 tokens/s and averaging in the 170s.
Specific Configuration:
- Model: qwen3827bnvfp4.ninfer
- Context/KV: 240k capacity, fp8 KV dtype, 16GiB host KV.
- Speculative Decoding: Enabled mtp and lm-head-draft with 3 draft tokens.
- Vision: Enabled vision and media-live (2GiB).
The author praised the project's performance optimization.
More from Infra
- Next-gen GraalVM benchmarks: 4x faster startup, 2x less memory, no build step needed — RealGeneKim · 2026-10-15
- 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