Introduction
DDN, a leading provider of data intelligence platforms, has made a groundbreaking achievement in the MLPerf Storage Benchmark, setting new standards for performance and efficiency. The company’s A³I (Accelerated Any-scale AI) systems demonstrated unmatched capabilities in multi-node configurations, solidifying its role as essential drivers for high-demand machine learning (ML) workloads and transformative business outcomes.
DDN’s Commitment to AI and ML Workloads
"Our MLPerf results emphatically showcase DDN’s unrivaled ability to deliver maximum performance, scalability, and cost-effectiveness for AI and ML workloads," said James Coomer, senior vice president of Products, DDN. "These benchmarks are not just numbers; they represent our commitment to helping customers accelerate their operations, enhance productivity, and maintain fiscal responsibility in an increasingly competitive landscape."
Participation in the Closed Division of MLPerf
In DDN’s participation in the closed division of MLPerf, the company focused on real-world small-scale ML deployments, rigorously testing how many GPUs can be efficiently saturated by a single DDN appliance across diverse ML tasks.
Game-Changing Highlights from DDN’s MLPerf Results
- Astounding Throughput: A single AI400X2T NVMe appliance achieved an extraordinary 100 GB/s read speed in Unet3D tests, efficiently saturating 72 A100 GPUs and 36 H100 GPUs—maximizing its output and exemplifying elite performance.
- Scalable Innovation: In the Resnet50 task, the AI400X2T seamlessly supported 864 simulated GPU clients across 32 compute nodes, translating to 36 GPUs per physical drive, showcasing its ability to drive innovative solutions at scale.
- Optimized Efficiency: During the Cosmoflow task, a single AI400X2T appliance adeptly managed 150 simulated GPU clients across 30 compute nodes—achieving 6 GPUs per drive—highlighting the solution’s power with an optimized physical footprint.
Conclusion
These stellar results emphasize DDN’s commitment to enhancing GPU performance while slashing energy consumption, enabling organizations to accelerate project completion and dramatically boost productivity. DDN empowers industries worldwide to unlock the full potential of AI, fueling innovation across sectors such as autonomous vehicles, personalized medicine, financial forecasting, and climate change research. With DDN at the helm, organizations are not just adapting to the future of AI – they are defining it.
FAQs
Q: What is DDN’s MLPerf performance like?
A: DDN’s MLPerf performance is unmatched, with a single AI400X2T NVMe appliance achieving extraordinary 100 GB/s read speed in Unet3D tests and efficiently saturating 72 A100 GPUs and 36 H100 GPUs.
Q: How does DDN’s solution drive innovation at scale?
A: DDN’s solution seamlessly supports 864 simulated GPU clients across 32 compute nodes, translating to 36 GPUs per physical drive, showcasing its ability to drive innovative solutions at scale.
Q: What are the benefits of using DDN’s solution?
A: DDN’s solution enhances GPU performance, slashes energy consumption, and enables organizations to accelerate project completion and boost productivity, making it an attractive option for industries worldwide.
