Andrea Booker, Dell Technologies & Steen Graham, Scalers AI | Dell Technologies World 2023
summary
Early-access tests run on PowerEdge XE9680 reveal a production rig with eight H100 GPUs and Broadcom Ethernet that runs LLM workloads about 2x faster than previous generations. Demonstrations include fine-tuned on‑prem foundation models using embeddings + vector databases, Stable Diffusion image generation and emerging multimodal/video workloads, and secured deployments for proprietary data. Main bottleneck is GPU supply; expected time to customer results is measured in months, with human‑in‑the‑loop validation.
AI-written summary · about
