10 AI Milestones of 2020
Over the last year, Inspur worked vigorously to advance AI hardware and deliver solutions to customers and a market facing an extraordinary year of unprecedented challenges. Integrating the latest technologies into state-of-the-art products, Inspur continued to expand its full-stack AI portfolio of computing platforms, AI resource platforms, and algorithm toolkits to meet customer needs — while setting performance records and raising industry standards along the way. Here are Inspur’s AI milestones of 2020.
Leading-Edge Tech at Time to Market
May: 5 New AI Servers Powered by NVIDIA A100 Tensor Core GPUs
Immediately following NVIDIA’s Ampere architecture release at GTC20, Inspur moved to upgrade its AI server suite to support 8 or 16 A100 GPUs, producing a comprehensive portfolio of AI products that can deliver performance of up to 40 PetaOPS.
June: AI Servers Supporting the Latest NVIDIA A100 PCIe Gen 4
Inspur released the NF5468M6 and NF5468A5 servers which can accommodate 8 pcs of double-width NVIDIA A100 PCIe GPUs in a 4U chassis. Both servers support the latest PCIe Gen 4 of 64GB/s bi-directional bandwidth, achieving superior AI computing performance.
July: Inference Platform for Resource Scheduling in Enterprise AI Environments
The AIStation inference platform increases resource utilization from 40% to 80% by enabling agile deployment of inference service resources. It also vastly reduces model deployment time from two to three days to just a few minutes by supporting unified scheduling of multi-source models. This all-in-one platform provides full support for training and inference scenarios, and enables model development to training, deployment, testing, release, and service.
July: New Server Optimized for Mobile Liquid Cooling Cluster
Inspur released a 2U4N high-density liquid-cooling server, the i24M5-LC, deployable in PUE<1.2 data centers. Meanwhile, it released a mobile rack-mounted Coolant Distribution Unit (CDU), which can be connected to the i24M5-LC with quick release connectors for easy deployment of a complete liquid cooling system, providing clients with an efficient and convenient mobile liquid cooling cluster solution.
October: Cloud SmartNIC Solution Based on NVIDIA DPU
Unveiled at GTC Fall 2020, the Inspur Cloud SmartNIC solution integrates the Inspur server with NVIDIA BlueField data processing unit (DPU). This pairing combines the capabilities of embedded processing, SmartNIC networking, and a high-performance PCIe 4.0 host interface, which enables oﬄoad functions such as traffic management, storage virtualization, and security isolation, significantly freeing up CPU computing resources.
November: Inspur’s AI Servers Upgraded to Support the Latest NVIDIA A100 80GB GPU
Further upgrades to NF5488A5 and NF5488M5-D to support the latest NVIDIA A100 80G GPU. The servers deliver a 15% performance increase in AI model training, with tens of billions of parameters.
Extreme Product Design Sets New Global Benchmarks
October: NF5488A5 Recognized for Record-Breaking Performance
The aforementioned NF5488A5, powered by 8 NVIDIA A100 GPUs with third generation NVLink interconnect and two AMD CPUs supporting PCIe 4.0, set 18 new performance records for the MLPerf v0.7 AI inference benchmark. It also topped the single server performance category for the MLPerf Training ResNet50 benchmark. It was shortlisted for CRN’s 10 Hottest Enterprise Servers of 2020.
Gaining Recognition in International Competitions and Reports
April: Inspur Takes Third Place in NeurIPS AutoDL 2019-2020
The Inspur team took third place in the AutoDL (Auto Deep Learning) 2019-2020 finals held by NeurIPS, a leading AI academic conference. The full-process AutoDL solution developed by Inspur achieved an average accuracy improvement of nearly 20% over the baseline, an average data reading efficiency improvement of 22%, and completed the model establishment, search and, generation processes in less than half an hour.
June: Inspur Tops VizWiz-VQA Challenge (Visual Question Answering Challenge) of CVPR 2020
July: Inspur Acknowledged by The Forrester Wave™ as a Leading Solutions Provider of Predictive Analytics and Machine Learning
The Forrester Wave™ Predictive Analytics and Machine Learning report reviewed 35 international PAML providers and recognized Inspur as a leading company among large established players the likes of Alibaba and Baidu. Inspur’s solutions in this field help to automate machine learning model development and boost AI productivity.