XPerf ™ Helps Customers Make Every AI Accelerator Count on AMD Instinct™ GPU Based Systems.

ClusterReady® Presents a User-Friendly Dashboards Presenting Key Performance Indicators

On a four-node, 32-GPU AMD Instinct™ MI325X rack, ClusterReady™ measured 95.8% scaling efficiency from one node to four, and 33,695 tokens/sec on Llama 2 70B LoRA.

XPerf pairs AMD Instinct™ performance with XPerf's workload testing to help customers assess the deployment they run.

ROUND ROCK, TX, UNITED STATES, November 12, 2026 /EINPresswire.com/ -- XPerf Inc™., an AI infrastructure software company building the control plane for heterogeneous data centers through AI-native cluster validation and optimization, today highlighted its ClusterReady™ software for assessing AI workload performance on AMD Instinct GPU-based systems. XPerf’s work combines AMD's high-performance AMD Instinct™ GPU platforms with XPerf's independent, reproducible validation for workload testing, so customers can assess how their AI workloads will perform. This work reflects XPerf’s commitment to openness, transparency, and verifiable AI infrastructure performance across increasingly heterogeneous environments.
AI adoption is pushing customers to evaluate a wider range of hardware, and AMD has built an open ecosystem to meet that demand. Measurements on real-world AI workloads can inform buying decisions. XPerf contributes to this process through workload testing on AMD Instinct-based systems.

Supporting the AMD Instinct Ecosystem
This work builds on GPU system performance from AMD and the open ecosystem customers increasingly want to build on. XPerf shows how that performance holds up under real training and inference workloads — measured independently, across multi-vendor, multi-generation hardware. Together, they give customers something neither vendor claims nor hand-rolled tests can: transparent, trustworthy proof that scales from a single research rack to enterprise data centers and cloud service providers, which makes adopting AMD solutions an easier, better-informed decision and helps the whole AI infrastructure market move faster.

In May 2026, XPerf tested ClusterReady™ on AMD Instinct™ MI325X- and MI350X-based systems on behalf of MiTAC Computing and Akash Systems as part of the MLPerf^® Training v6.0 round. In earlier published results on AMD Instinct™ MI325X-based systems, XPerf's software demonstrated 95.8% scaling efficiency from a single node to a four-node rack — an audited result on AMD hardware through XPerf’s validation and optimization software. That experience informs XPerf's continued work helping customers assess AMD Instinct GPU-based infrastructure.

Looking Ahead
XPerf intends to collaborate with AMD and its partners on the MLPerf^® Training rounds going forward, building on their MLPerf Training v6.0 work, and to deepen their participation in MLPerf benchmarks over time. MLPerf, maintained by MLCommons^®, is the peer-reviewed standard for measuring AI training and inference performance. Putting results through its transparent process gives AMD customers auditable, industry-comparable data they can trust.
XPerf also continues to explore opportunities to apply ClusterReady™ to future AMD Instinct-based infrastructure.

“Customers adopting AMD Instinct™ GPUs want proof, not promises. ClusterReady™ gives them an independent, reproducible measurement of AI workload performance on AMD-based systems, and turns them deployment-ready in hours with root cause analysis and remediation in minutes, so that customers know what to expect before large-scale deployment.” — Alex Carter, Founder & CEO, XPerf Inc.

“AMD Instinct™ GPUs and ROCm™ software provide an open platform for AI training and inference. XPerf is doing great work to help customers assess system readiness and workload performance on AMD Instinct-based systems.” — Marilyn Basanta, Senior Director, Datacenter GPU Product Management, AMD

The two halves reinforce each other. AMD Instinct™ GPU-based systems deliver the performance customers are adopting for demanding AI training and inference; XPerf's validation makes that performance verifiable, because it is measured independently across multi-vendor, multi-generation hardware — not by the accelerator vendor. For a customer standardizing on AMD, this helps customers assess performance consistency and deployment readiness.

XPerf's ClusterReady™ software helps customers validate AI racks and clusters in hours instead of weeks through AI workloads and stress testing. XPerf’s broader portfolio includes ClusterBeacon™, which autonomously predicts and resolves cluster failures, helping customers move from validation through to reliable, optimized operations. To learn more, visit XPerf.ai.

About XPerf Inc.
Founded by ex-Intel engineers with deep experience deploying and managing GPU and ASIC clusters at scale, XPerf Inc. is an AI infrastructure software company building the control plane for heterogeneous data centers. Its products cover the full AI cluster lifecycle: the ClusterReady™ platform validates AI racks and clusters 10× faster than L11/L12 workflows through AI workloads and stress testing; ClusterBeacon™ software predicts, prevents and resolves GPU cluster failures on-premises; and the ClusterMarshal™ scheduler helps teams choose where, when and how each GPU job runs. XPerf Inc. is a member of NVIDIA Inception, CNCF, the gBETA accelerator program, and the Endless Frontier Labs accelerator program.
Learn more: xperf.ai

Trademarks
XPerf, ClusterReady, ClusterBeacon and ClusterMarshal are trademarks of XPerf Inc. AMD, the AMD Arrow logo, AMD Instinct, ROCm, and combinations thereof are trademarks of Advanced Micro Devices, Inc. MLPerf and MLCommons are trademarks of MLCommons Association. NVIDIA is a registered trademark of NVIDIA Corporation. Intel is a trademark of Intel Corporation. Other names are for informational purposes only and may be trademarks of their respective owners.

Forward-Looking Statements
This press release contains forward-looking statements regarding a collaboration between XPerf Inc. and AMD and a shared roadmap toward future MLPerf^® benchmark participation. Such statements reflect current intentions and are subject to change; neither company guarantees future benchmark participation, results, or performance outcomes. These statements involve risks and uncertainties that could cause actual outcomes to differ materially. XPerf undertakes no obligation to update these statements after the date of this release.

Lucia Dominguez Cabanes
XPerf Inc.
+1 607-438-8136
press@xperf.ai
Visit us on social media:
LinkedIn
YouTube