vllm-nvidia-hardware

Size NVIDIA vLLM hardware by evaluating SKU matrices, power budgets, and cooling prerequisites for Hopper through Rubin platforms. Apply facility readiness checks and platform-matrix references to guide procurement and risk assessment for enterprise deployments.

Supported NVIDIA SKUs for vLLM hardware sizing include B200, B300, GB300 NVL72, and Vera Rubin roadmap platforms. SKU tables cover Hopper through Rubin architectures to inform deployment scale and context.

Yes, use this reference to size a GB300 NVL72 rack deployment for vLLM, including facility prerequisites and capital planning. It applies SKU matrices and power constraints to procurement decisions for Dell, Lenovo, and HGX racks.

Calculate power and cooling requirements for a 70B FP4 vLLM deployment by applying facility prerequisites and platform-matrix references. Check data-center readiness and power budgets against specific GB300 NVL72 hardware SKUs for 1M context windows.

Yes, it compares Dell, Lenovo, and HGX rack platforms for vLLM deployments using platform-matrix references. Assess hardware sizing, facility readiness, and risk across these platforms for Hopper through Rubin architectures.

Avoid using this skill for vLLM hardware sizing when evaluating non-NVIDIA accelerators or platforms outside the Hopper through Rubin roadmap. It strictly relies on NVIDIA SKU tables, power constraints, and facility prerequisites.