Huawei Atlas 900 AI Cluster Wins the Only Best of Show Award for IA at Interop Tokyo 2020

The Interop Tokyo 2020 panel of judges has awarded Huawei Atlas 900 AI cluster the Best of Show Award in recognition of its powerful AI computing and superior heat dissipation. Huawei Atlas 900 is the only AI product of 2020 to receive the Best of Show Award. Atlas 900’s exceptional performance is a catalyst for global AI research and drives AI application across industries.

The Best of Show Award is the highest honor of the Interop conference. Only products that meet the high quality standards demanded by the Japanese IT market are considered.

Recommended AI News: Opinion: Dealing With The Deadly Concoction Of Disease And Disaster

Tony Xu , CEO of Huawei Ascend Computing, said: “The Huawei Atlas AI computing solution offers powerful computing and energy efficiency for all AI scenarios across devices, edge and cloud. The Atlas 900 AI cluster offers hubs Data with effective computing, high linearity, and best energy efficiency to accelerate data-driven research, such as astronomical exploration, weather forecasting, oil exploration, and genetic sequencing. Research advances translate into practical benefits for people of all the world”.

Atlas 900 is the world’s fastest AI training cluster. It offers a total computing power of 256 to 1024 petaFLOPS at an average precision (FP16), equivalent to the computing power of 500,000 personal computers. Atlas 900 broke the world record with the ResNet-50 reference test for model training completing the training in 59.8 seconds. Atlas 900 is the only product capable of completing training in less than a minute. Atlas 900 has extensive applications in scientific research and business innovation for the fastest training of AI models with images and videos.

Atlas 900 integrates three interfaces for high-speed interconnection: Huawei Cache Coherence System (HCCS), PCIe 4.0, and 100G Ethernet. The Atlas 900 AI cluster uses switches from the Huawei CloudEngine data center to work on a dedicated 100 TB / s full mesh, non-blocking network for parameter synchronization. The network reduces parameter synchronization latency by 10 to 70 percent to streamline IA model training.

Recommended AI News: Pretected Reveals The Most Popular Cars In America For 2020

Heat dissipation is a critical problem for the IA training cluster with such computational capacity. That is why the Atlas 900 AI cluster adopts a revolutionary system for heat dissipation. Lead the industry innovation with a complete liquid cooling solution and a rack scale coated adiabatic design. The design offers tremendous heat dissipation even for single racks with power consumption of up to 50 kW. It achieves a less than 1.1 Energy Efficiency (PUE) usage for data centers, nearly reaching the ideal PUE of 1.0. Atlas 900 improves air-cooled 8-kW zippers by reducing equipment space by 79%.

Huawei drives cooperation to build an Ascend computing industry with open hardware, open source software, and partner enablement. Huawei offers full stack AI computing application and infrastructure solutions to power AI industries and create pervasive intelligence.

Recommended AI News: Accedian, New Context, And Savari, Provide Assured And Secured Communications Foundation For Next Generation