DDN Achieves Performance in MLPerf Benchmarking, Empowering Transformative AI Business Outcomes
Single AI400X2T NVMe appliance with 100GB/s read speed in Unet3D tests, efficiently saturating 72 A100 GPUs and 36 H100 GPUs
This is a Press Release edited by StorageNewsletter.com on October 10, 2024 at 2:03 pmDDN (DataDirect Networks Inc.) announces an achievement in the MLPerf Storage Benchmark, setting new standards for performance and efficiency.
Company’s A³I (Accelerated Any-scale AI) systems demonstrated unmatched capabilities in multi-node configurations, solidifying its role as essential drivers for high-demand ML workloads and transformative business outcomes.
“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, SVP, 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.”
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.
AI400X2T NVMe appliances
Game-changing highlights from DDN’s MLPerf results:
- Astounding Throughput: A single AI400X2T NVMe appliance achieved a 100GB/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 supported 864 simulated GPU clients across 32 compute nodes, translating to 36 GPUs/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/drive – highlighting the solutions’ power with an optimized physical footprint.
These results emphasize the firm’s commitment to enhancing GPU performance while slashing energy consumption, enabling organizations to accelerate project completion and boost productivity.
The company empowers industries WW to unlock the full potential of AI, fueling innovation across sectors such as autonomous vehicles, personalized medicine, financial forecasting, and climate change research. With the firm at the helm, organizations are not just adapting to the future of AI – they are defining it.
Resource:
MLPerf Storage Benchmark Suite Results