Nscale tests show more GPUs operating within the same power budget

Nscale demonstrated 37% more GPUs and nearly 50% higher aggregate inference throughput within the same provisioned power budget in tests using Nvidia GB300 NVL72 systems, Nvidia said. The tests used Nvidia DSX MaxLPS to coordinate power allocation with workload requirements. Nscale also plans to deploy Vera Rubin infrastructure for frontier-model training and production workloads as it expands across five European markets. The reported capacity gains come from the GB300 tests, rather than the planned Vera Rubin deployment.