Chen Huixin (Chart 3), Director of IP Sales, Cadence Asia Pacific, pointed out that artificial intelligence principles and algorithms have been developed in the academic world for decades and are not strictly new technologies. In recent years, The hot topic, the main processor computing performance across the threshold of demand, so that these algorithms can come in handy in practical use.
However, the core of AI is the algorithm, so the key to how the chip designer develops a SoC that can execute the algorithm most efficiently is the mastery of the algorithm, but algorithms are not the strengths of traditional SoC developers, In China, many chip developers adopt the strategy of finding a partner and make artificial intelligence layout by developing chips for partners familiar with the algorithm.
On the other hand, there are a number of start-ups in the Chinese market who are also involved in the development of related chips, some of which start out as business ventures and software / network companies into the field of chip design. The common characteristic is the high degree of mastery of artificial intelligence algorithms.
In fact, as Chen Huixin observed, in addition to the advantages of using artificial intelligence chips for mobile phones or traditional mobile phone chip factories, the development of artificial intelligence chips in China is relatively fast at present, and most of these new faces are developed. For example, in the Cambrian era, Chip developers seem to be still adapt to the changes brought by artificial intelligence.
Overall, the demand for performance is still the most important factor affecting the product development schedule of chip design companies.As mobile devices such as mobile phones are essentially lightweight AI applications, handset chipmakers need to rapidly develop corresponding products and technical challenges Is relatively low.However, if it is other applications that will use the heavyweight AI algorithm, the processor chip on which it is mounted naturally has a relatively high performance requirement and the product development speed will be slower.
At present, the leading artificial intelligence chip companies in China, such as the Cambrian and Huawei, are all Cadence IP authorized customers. In addition to the processor core, some customers also use the high-speed serial / deserializer (SERDES) ) IP Because artificial intelligence will create tremendous data throughput demands, the need for high-speed interfaces will benefit from artificial intelligence in addition to the processor itself.