Hardcore science: Da Vinci architecture.
British Code Technology. Hello, everyone, I'm Chen Jiakun, a fae engineer at Imma Technology. Today, I will take you to understand what is the da Vinci architecture. First of all, we all know that the world's mainstream chip architectures include: M architecture, X86 architecture, ARM architecture and RISC-V architecture. In particular, the x86 architecture and ARM architecture have occupied the mainstream market for a long time.
Where did the Da Vinci architecture come from? And what are the unique advantages? Let me give you a brief introduction. With the rise of domestic computing power, Huawei proposed a full-stack, all-scenario AI strategy at CONNECT 2018. As an important technical foundation, AI chips play an important role in it. To achieve AI collaboration across multiple platforms and scenarios, Huawei innovated and designed the da Vinci computing architecture.
The da Vinci architecture has the advantages of high computing power, high energy efficiency, flexibility and tailoring. Specifically, the da Vinci architecture uses 3dCube to accelerate matrix operations, which greatly improves the AI computing power per unit of power consumption. Each AI Core can achieve 4096 MAC operations in a single clock cycle, an order of magnitude improvement over traditional CPU and GPU computing.
What do you mean? For example, if you complete 4096 operations, a 2D structure needs 64 rows x 64 columns to calculate, and 3Dcube only needs a 16*16*16 structure to calculate. Therefore, the da Vinci architecture can greatly improve computing power, shorten the computing cycle, and achieve faster and stronger AI computing.
At the same time, in order to improve the completeness of AI computing and the computing efficiency of different scenarios, the da Vinci architecture also integrates a variety of computing units such as vectors, scalars, and hardware accelerators to support data accuracy requirements in multiple scenarios of calculation, training, and inference, so as to achieve full-scenario coverage of AI requirements.
Not only that, based on the flexible and scalable characteristics, the da Vinci architecture can meet the application scenarios of the device, edge, and cloud, and can be used in training scenarios ranging from tens of milliwatts to hundreds of watts, spanning all scenarios and providing optimal computing power.
In addition, another important advantage is that the development support of the da Vinci architecture is very complete, which can provide more support for operator open source frameworks, as well as provide a more complete tool chain, which greatly improves the work efficiency of developers and saves more time. Labor costs, accelerate the implementation of more AI applications.
Imma has been working closely with Huawei HiSilicon on AI+** scenario-based applications, and has launched a variety of products based on"Da Vinci architecture"The edge computing box gives full play to the advantages of the da Vinci architecture to the edge of AI applications, empowering smart construction sites, smart transportation, smart emergency and other fields, and realizing AI applications more efficiently and at a lower cost.
That's the end of this issue of popular science**, we'll see you next time, bye-bye.