Today, GSMA awarded Huawei Atlas 900 AI cluster with the Global Mobile Awards 2020 (GLOMO Awards) – Tech of the Future Award. Atlas 900 stood out with its world-leading AI computing power, ultimate heat dissipation system, and best-in-class cluster network. Atlas 900 accelerates global basic AI research and quickly brings AI applications to industries to advance the AI era with unparalleled AI computing power.
Innovative technology has propelled the mobile industry far beyond the wildest expectations of early tech pioneers. GSMA awards the GLOMO Award – Tech of the Future Award to recognize technology that is ahead of its time and reshapes the world. The award marks industry recognition of Huawei Atlas 900 AI cluster as a product that scales new heights in AI innovation.
Atlas 900 is the world’s fastest AI training cluster. It is a powerhouse that combines thousands of Huawei Ascend 910 AI processors for 256 to 1024 petaFLOPS (PFLOPS) FP16, providing the computing power of 500,000 PCs. During the ResNet-50 model training benchmark test, Atlas 900 shattered the world record by completing training in 59.8 seconds. It is the only product in the industry capable of completing such training under a minute.
Atlas 900 has diverse applications for scientific research and business innovation, allowing researchers to train AI models with images and videos more quickly. Not only does this benefit scientific research in more efficient and accurate astronomical, petrochemical, and weather predictions, but it also provides business benefits, such as faster time-to-market for autonomous driving.
The Atlas 900 AI cluster has three high-speed interconnect modes: Huawei Cache Coherent System (HCCS), PCIe 4.0, and 100GE. Atlas 900 uses data center switches from the Huawei CloudEngine series to support a full-mesh parameter synchronization network for AI model training boasting over 100 TB (10^12)/second. This reduces the synchronization latency by 10% to 70% for a huge leap in training efficiency.
Ultimate computing power can result in system overheating. The Atlas 900 AI cluster has an industry-leading heat dissipation system to cope with this issue. Its hybrid liquid cooling solution supports a liquid cooling ratio exceeding 95% with cabinet-level enclosed adiabatic technology. Ideal power usage effectiveness (PUE) for a data center is 1.0. A single cabinet with Atlas 900 provides heat dissipation of up to 50 kW for a PUE of below 1.1. In addition, Atlas 900 is far more space efficient than an 8 kW air-cooled cabinet, reducing the equipment footprint by 79%. The innovative liquid cooling system greatly reduces customer TCO with high power, high density, and low PUE.
Comments from the judges, “This entry created a platform for progress across a wide array of applications in the mobile industry and beyond, leveraging the latest technology to deliver incredible speed and performance with a minimal carbon footprint.”