DDN®, data analytics platform provider, proudly announces a groundbreaking achievement in the field MLPerf™ Storage Benchmarksetting recent standards of performance and efficiency. DDN A3I™ (Accelerated Any-scale AI) systems have demonstrated unparalleled capabilities in multi-node configurations, cementing their role as fundamental drivers of high-demand machine learning (ML) workloads and revolutionary business outcomes.
Participating in a closed division of MLPerf, DDN focused on real-world, small-scale ML deployments, rigorously testing how many GPUs a single DDN device could efficiently handle across a variety of ML tasks.
Top game-changing MLPerf DDN results:
- Astonishing throughput: Single AI400X2T™ The NVMe device has achieved something extraordinaryRead speed 100 GB/sin Unet3D tests, saturating efficiently72 A100 GPUsAND36 H100 GPUs– maximizing your performance and exemplifying elite performance.
- Scalable innovations: In the Resnet50 task, the AI400X2T performed without any problems864 simulated GPU clientsin 32 computing nodes, which translates to36 GPUs per physical diskdemonstrating its ability to implement creative solutions on a huge scale.
- Optimized performance: During the Cosmoflow task, a single AI400X2T device was efficiently managed150 simulated GPU clientsin 30 compute nodes – achievement6 GPUs per disk—emphasizing the power of solutions thanks to optimized physical space.
These outstanding results underscore DDN’s commitment to increasing GPU performance while lowering power consumption, enabling organizations to accelerate project execution and dramatically raise productivity.
DDN enables industries around the world to unlock the full potential of AI, driving innovation across sectors such as autonomous vehicles, personalized medicine, financial forecastingAND climate change research. With DDN at the forefront, organizations aren’t just adapting to the future of AI – they’re defining it.