XPerf today launched XPerf ClusterReady — a platform that validates GPU racks and clusters using real AI inference and training workloads, and reports tokens/sec output as part of every validation run. ClusterReady delivers a 10× speed-up over conventional rack and cluster validation, compressing what is typically a multi-week process into hours.
Working with MiTAC Computing, XPerf demonstrated strong cluster-level performance on AMD Instinct MI325X racks and validated AI workloads in hours rather than weeks — confirming that production AI infrastructure can be certified and shipped revenue-ready on Day 1.
Why this matters
As clusters scale to thousands of GPUs, the question is no longer "does the hardware work" — it is "does the cluster deliver the AI performance you are paying for?" ClusterReady answers that question before a cluster goes live, surfacing hidden hardware and interconnect issues that traditional manufacturing tests routinely miss, and reporting a verified tokens/sec baseline that customers can rely on at handover.
Built for the AI infrastructure ecosystem
XPerf is excited to work with OEMs, ODMs, rack-level integrators, and data center operators to accelerate deployment of high-performance AI infrastructure. The same validation engine that powers ClusterReady is also integrated into XPerf OpsPilot as the on-premise calibration source for autonomous cluster operations.
Meet us at NVIDIA GTC
XPerf will be at NVIDIA GTC. To set up a conversation, message Alex Carter, PhD at alex.carter@xperf.ai.