Huawei Atlas

2021

As the basis for intelligent video analytics solutions

On August 12, 2021, Huawei announced that together with ComBox Technology, they launched solutions based on the EDGE accelerators of the Huawei Atlas 500.

The companies will develop comprehensive solutions based on ComBox Technology's computer-assisted vision and Huawei Atlas 500 edge AI accelerators. The development will be used in companies from the services, retail, finance, transport, as well as production and construction organizations. Combox software, which is included in the register of Russian software, provides the highest accuracy of detection of various events for access control, for example, recognition of license plates and vehicle brands, availability of protective equipment, incident detection and penetration into hazardous areas, etc.

A key feature of solutions is data edge processing, which allows you to connect even those points of presence of companies where data transfer data centers to is physically impossible or economically not profitable. In addition, this approach improves the security of company presence points by processing data in real time and by providing a high degree of autonomy of the presence points, even when breaking communications from the center. All this was made possible thanks to the Huawei Atlas 500 accelerator, which provides the required performance and low power consumption, as well as is not sensitive to the operating environment.

The hardware platform for video analytics on the edge must be balanced - provide the required processing power at low power consumption, and also not be demanding for the operating environment. The Huawei Atlas 500 accelerators meet all the requirements, "said Ilya Dushin, executive director of ComBox Technology.

Data processing in close proximity to data sources is one of the main trends in the development of AI-based solutions. In particular, a large number of enterprises in China, such as banks, retail, gas stations, etc., are developing such solutions. We hope that our cooperation will help in the development of the AI sector in Russia, "commented Oleg Sirotyuk, director of research and development for the direction of intelligent computing systems Huawei.

At TAdviser SummIT Huawei, Russia will talk about the Huawei Atlas intelligent video processing platform

Oleg Sirotyuk, a company representative, Huawei May 26, 2021 at TAdviser SummIT in his speech will talk about the platform Huawei Atlas 500, which is widely used in the branches of the largest to banksChina implement various scenarios. intelligent video processing The presentation will present the Huawei Atlas 500 platform and an example of building an intelligent system based on it video surveillances.

2019

Atlas 900 Cluster Announcement

Atlas 900, a fast-acting artificial intelligence (AI) learning cluster, is a high-performance computing AI platform that will help adapt artificial intelligence to applications in a wide range of scientific research, as well as to develop innovative business solutions

The Atlas 900 cluster, which is based on advanced technologies developed by Huawei over the past decade, has combined the power of thousands of Ascend processors. The Atlas 900 cluster spends only 59.8 seconds training ResNet-50, the gold standard for measuring AI performance. This is 10 seconds faster than the previous world record.

Atlas 900 will open up new opportunities in a wide range of scientific research and in the field of creating innovative business solutions - from astronomy, weather forecasting and autonomous driving to oil exploration. Huawei also launched the Atlas 900 in September 2019 based on Huawei Cloud as a cluster service, making extraordinary processing power available to its customers from various industries. Huawei offered this service at a great discount to universities and research organizations around the world.

Huawei's Computer Computing Strategy

The industry's computational methods are changing, with rules-based models giving way to statistical models that underpin machine learning. According to Huawei, in the next five years the market will be won by statistical computing systems, and over 80% of the world's computing resources are involved in artificial intelligence.

To take a position in this market, Huawei will focus on four key initiatives.

• Development of innovative architecture. In 2018, Huawei introduced the innovative Da Vinci processor architecture, providing a stable and significant reserve of available computing power.

Huawei will continue to invest in basic research.

• Investment in processors for different applications. Huawei has a full line of processors: Kunpeng processors for general computing, Ascend processors for AI, Kirin processors for smart devices and Honghu processors for smart screens.
• Clear delineation of business. Huawei does not plan to directly sell its processors. Instead, the company will provide processors in the format of cloud services to its customers and in the form of components - partners, giving priority to supporting integrated solutions.
• Creating an open ecosystem. In the next five years, Huawei will invest another 1.5 billion dollars US dollars in its program for developers. The company aims to expand the program to support five million developers and enable Huawei partners around the world to develop intelligent applications and solutions of the next generation.

Atlas Platform Official Market Launch

Huawei announced on April 18, 2019 in Shenzhen the official release of the Atlas AI computing platform to the market. This event opens the next chapter in the history of commercial implementation of Atlas, confident in the company.

The Huawei Atlas computing platform is based on processors Ascend Huawei AI, which takes advantage of the Da Vinci 3D Cube architecture, which is characterized by ultra-high computational density and high energy efficiency.

According to April data, Atlas offers products of various form factors: accelerator modules, boards, peripheral AI stations, etc. Such an extensive range provides a versatile solution for AI infrastructure for the device-periphery-cloud model. The solution focuses on applications such as smart city, telecommunications, finance, the Internet and the electricity sector.

• Atlas 200 Accelerator AI Module: A device in a form factor twice the size of a standard credit card supports the analytics functions of 16-channel HD-quality video broadcast in real time, and can be used in various terminal systems: cameras, robots, drones, etc.
• Atlas 200 DK Developer AI Toolkit: Tools allow you to create a development environment with integrated graphics programming capabilities in less than 30 minutes. Thanks to its comprehensive Ascend processor capabilities, the Atlas 200 DK AI toolkit allows you to create applications and implement them in device-periphery-cloud scenarios without having to make any changes to program codes.
• Atlas 300 AI Accelerator: The accelerator board is implemented in half-size form factor and is equipped with standard PCIe interfaces. The device provides 64 TOPS (operations per second) performance for INT8 computing operations and supports HD video analysis for 64 channels in real time, consuming only 67 watts of power. The Atlas 300 accelerator also boasts 32 GB of memory, which increases the efficiency of video analysis processes and active generation of logical outputs.
• Peripheral AI station Atlas 500: characterized by high performance, the device is able to perform data processing in real time at the boundaries of the computer system. The performance of INT8 for one station is 16 TOPS, and the daily power consumption does not reach 1 kWh. Designed for complex environmental conditions on the periphery, the Atlas 500 station is able to function stably at temperatures from -40 to + 70 degrees Celsius.

Huawei Intelligent Computing has already developed a full range of specialized products, including servers FusionServer Pro based on intelligent control and acceleration modules, the Atlas AI computing platform and the Mobile Data Center (MDC) computing platform using Ascend AI processors, as well as servers processor-based TaiShan. Kunpeng ARM

2018: Announcement of the Atlas AI computing platform

In October 2018 Huawei , she introduced the Atlas intelligent computing platform based on - AIprocessors a series Huawei Ascend and mass components of heterogeneous computing. The Atlas platform includes modules, maps, panels, peripheral stations and devices for creating infrastructures artificial intelligence technology that is compatible with any scenarios per end user, periphery and. cloud The solution has powerful computing potential and helps accelerate the process of intelligent transformation.

In the context of the rapid development of artificial intelligence technologies, industry solutions based on these technologies are also on the rise. Their introduction in all spheres of life plays a key role in the process of creating an intellectual world. At the same time, computing resources create additional capabilities. With an artificial intelligence (AI) -based infrastructure solution compatible with any scenario and focused on end-users, peripherals, and the cloud, Huawei helps its customers achieve commercial success in the age of intelligent technology.

The intelligent computing platform was another step in the evolution of the Atlas intelligent cloud hardware platform, released in 2017. The development includes the Atlas 200 AI accelerator module, the Atlas 300 AI accelerator card, the Atlas 500 AI peripheral station and the Atlas 800 AI appliance universal platform, aimed at enterprise customers.

• The Atlas 200 AI accelerator module has miniature dimensions - it is half the size of a standard credit card, but at the same time supports 16-channel HD video analytics in real time. When installed on cameras and drones, the Atlas 200 consumes only 100 watts.
• The Atlas 300 AI accelerator card is several times smaller than the standard PCIe card, it is designed for data centers and peripheral servers. The Atlas 600 accelerator card supports maximum data accuracy and provides up to 64 TOPS INT8s based on a single card.
• The Atlas 500 AI Peripheral Station is an AI-based system that supports 16-channel HD video processing. In size, it is no larger than a conventional set top box. The station is well suited for various scenarios and can be used in transport, healthcare, automated retail sales, in intelligent production, etc.
• The Atlas 800 AI software and hardware complex creates an optimized AI environment and a pre-installed basic software library and can be started in two hours. In addition, the device includes software cluster management and job scheduling, and system-level performance monitoring capabilities, which significantly reduces the requirements for enterprise use.