Redditor benchmarks a DGX Station: GB300 throughput numbers revealed
CookEasy · reddit · 2026-11-14
A Redditor got a month alone with an ASUS ET900N G3 DGX Station (GB300, 250GB VRAM, 72 Grace cores, 750GB RAM, liquid-cooled) and shares day-one NVFP4 numbers: Qwen3.8-Flash-Next hits 2700 tok/s at 64 parallel on vLLM and 3002 on SGLang; DeepSeek V4.1 Flash manages 59 single / 246 at 16 parallel with a 19-minute load time. The box pulls 785W at 50°C, TensorRT-LLM won't install outside NGC containers, and Kimi K3.1 Pro doesn't fit. He's soliciting test requests for long context, finetuning, parallel agents, and more.
More from Infra
- Microsoft cloud outage passes 26 hours, engineer calls it a historic event on par with Heartbleed — esrtweet · 2026-11-14
- North Dakota town votes 3-to-1 against data center restrictions, expects $1.65M in taxes — kevinnbass · 2026-11-14
- InferenceForge claims 43% higher LLM inference throughput than best AgentX baseline — kaiwei_chang · 2026-11-14
- Hugging Face launches llama.cpp v1 engine to run AI models locally everywhere — LinkedInNews · 2026-11-14
- China Took 59% of 600,000+ Industrial Robots Installed in 2025; Humanoid Shipments Still Tiny — ingliguori · 2026-11-14
- SGLang joins NVIDIA AI Enterprise with NIM-certified inference microservices — ying11231 · 2026-11-14