Moore Threads and Beijing First-class Technology Co., Ltd. (referred to as First-class Technology) will carry out in-depth technical cooperation. Based on the powerful computing power of Moore Threads’ domestic full-featured GPU and the efficient features of the first-class technology OneFlow deep learning framework, it will provide extraordinary acceleration performance for AI applications. The Moore Threads GPU will provide high-performance computing power support for OneFlow, help improve the AI user experience, and jointly create better AI industry solutions. On this basis, the two parties will also jointly build a shared ecosystem to help the development of China’s artificial intelligence industry.
Deep learning is an artificial intelligence research direction that has emerged in recent years. Enterprises cannot do without deep learning frameworks and hardware infrastructure when training and deploying deep learning models.The deep learning framework is known as the artificial intelligence operating system, which has a decisive influence on the design and application of deep learning models. High-performance computing chips provide strong computing power support for the deep learning framework. In recent years, thanks to the development of deep learning frameworks and high-performance computing chips, deep learning has shown great revolution, promoting artificial intelligence to take the lead in some fields and achieve wide applications, such as speech recognition, machine vision, and natural language processing.
With the advancement of social digitization, the scale of data has exploded, and the scale of deep learning neural network models has become larger and larger. The training and inference of super-large models have put forward higher demands on computing power. In particular, the training of deep learning neural network models requires massive floating-point calculations in a large-scale distributed parallel computing mode, and matrix calculation is the main form of calculation. At the same time, the training process is accompanied by a large amount of data access and communication. A GPU is a processor specially designed for graphics rendering tasks. It adopts a data-parallel processing architecture and is naturally suitable for data-parallel processing application scenarios. It can perform ultra-efficient data-parallel computing. Therefore, it has become the driving force behind the rapid development of artificial intelligence and deep learning. The computing engine The domestic full-featured GPU and software stack developed by Moore Threads can be applied to various scenarios such as graphics rendering, video encoding and decoding, AI training and reasoning, and high-performance computing, and output powerful computing power.
Distributed computing system scheduling is one of the core technical difficulties of the deep learning framework. First-class technology foresees the development trend of the deep learning model becoming larger than its peers. Based on this trend, the demand for the system architecture has reversed, and the static state is the first in the industry. The core idea of scheduling and streaming execution Based on this research and development, the new generation of deep learning frameworks, OneFlow, has completely independent intellectual property rights. It has created four core technologies: automatic data model hybrid parallelism, static scheduling, decentralization, and full-link asynchronous streaming execution, which perfectly solve the problem of big data. The distributed expansion challenge of heterogeneous clusters brought about by large models and large computing breaks the limitation of the “memory wall,” and the technical level is world-leading. Compared with other deep learning frameworks, the biggest feature of OneFlow is its strong horizontal scalability, high hardware utilization, and system acceleration ratio. More importantly, it can support parallel modes such as data parallelism, model parallelism, and pipeline parallelism with only one set of mechanisms. Since OneFlow was open sourced, a number of leading Internet companies, AI companies, and research institutions have used it in scenarios such as video image recognition, image detection, speech recognition, natural language processing, advertising distribution, and recommendation.
Moore Threads is a semiconductor high-tech company that focuses on GPU chip design. Relying on the full-featured GPU developed by independent innovation, it is committed to building China’s graphics, visual computing, and artificial intelligence computing platforms and helping to build China’s high-performance computing ecosystem. Moore Threads builds software and hardware products based on the self-developed advanced computing architecture MUSA to provide strong computing power support for ecological partners.
Moore Threads is in line with first-class scientific and technological concepts and complements its advantages. The two parties will cooperate closely to create a nationally-made artificial intelligence software and hardware integration solution based on the domestic GPU + domestic deep learning framework and realize independent, innovative, high-performance, and easy-to-use deep learning software.hardware products, improve application experience, and jointly promote extensive ecological cooperation to accelerate the integration and application of artificial intelligence in various industries and fields.
Moore Threads is a high-tech integrated circuit company that specializes in GPU chip design. Its primary focus is the development and design of full-featured GPU chips and related products that can provide powerful computing acceleration capabilities to Chinese technology partners. Founded in October 2020, the company is dedicated to developing a new generation of GPUs for meta-computing applications, constructing a comprehensive computing platform that integrates visual computing, 3D graphics computing, scientific computing, and artificial intelligence computing, and establishing an ecosystem based on cloud-native GPU computing to help drive the digital economy’s development.
Moore Threads is distributed by Unixcloud Technology (Shenzhen) Co., Ltd. Unixcloud has a long history of product development and manufacturing. Moore Threads’ official authorized distributor is UnixCloud. It focuses on the field of edge computing in response to the computing power requirements of artificial intelligence development, and it offers the industry edge computing products and solutions that meet a variety of AI requirements. UnixCloud launched the 10G network card business at the same time, offering four-port and two-port 10G optical fiber network cards based on the Mucse-based network controller N10.