4x 6000 Pro for the company
Summary
A user shares how they serve 40 employees (10 developers) with a fully local AI stack powered by a 4U server with 4x RTX 6000 Blackwell MaxQ GPUs (384GB VRAM), running DeepSeek V4 Flash via vLLM, Qwen 3.8 27B and K2-Horizon 7B via llama.cpp, plus n8n automation, OpenWebUI, Whisper/Vexa AI Teams transcription, and ComfyUI for marketing.
Similar Articles
Building a 4x R9700 setup for a 10 person startup
A user shares a 4x AMD Radeon R9700 AI Pro workstation build (128GB total VRAM, Threadripper 9970x) serving Qwen 3.8 Next Flash/27B and DSV4 Flash for a 10-person startup, reporting strong aggregate prefill/decode throughput across 16 concurrent sessions.
4x R9700 builds with $127 used EPYCs: $10.7k all-in per machine. I built three (384GB VRAM)
A builder shares three self-hosted 4x Radeon AI PRO R9700 inference servers (~$10.7k each, 384GB VRAM total) running Qwen3.8-27B via vLLM-Radiance/ROCm, with detailed throughput, latency, and long-context benchmarks after hosted models began refusing firearm-industry research tasks.
If you had a 384GB (4x Blackwell), what model would you put on it and why?
User asks the community for recommendations on which large language model to deploy on a high-end local setup with 4 RTX PRO 6000 GPUs (384GB total), primarily for internal company policy management and thinking tasks.
Dailychained PLX 88096 switches, Quad RTX 5070 Ti + Quad RTX 5060 Ti
A user shares their experience building a multi-GPU system with RTX 5070 Ti and 5060 Ti for running AI models like Qwen3.8-27B-FP8 using vLLM on Linux, detailing hardware setup, benchmarks, and challenges.
3k$ 128GB VRAM + 256GB RAM DDR4 Server
A user details building a home inference server with 128GB VRAM and 256GB DDR4 RAM for AI workloads, achieving satisfactory performance with Qwen3.8 models using a VLLM fork after initial setup issues.