Wave Computing Technology Chris Nicol explained that the company's next-generation DPU combines its own data stream architecture and 64-bit multi-threaded core technology with TSMC's 7nm process technology to enhance its edge-applied AI solutions. Handling performance and reducing the risk and cost of the production process. In addition, Broadcom's packaging and testing experience and 7nm silicon intellectual property (IP) platform will also benefit the company, further enhance the efficiency and capacity of the AI system, in response to the huge Machine learning data set.
Frank Ostojic, senior vice president and manager of Broadcom's ASIC product division, also pointed out that the complexity of AI workloads and neural networks is exploding, and he believes that the integration of 7nm advanced process nodes with data flow technology will help AI. The industry is innovating faster from the cloud to the Internet of Things.
Overall, Wave Computing's current version of the DPU uses a 16nm process node, while the next-generation DPU will be produced using TSMC's 7nm process, leveraging Broadcom's design platform, mass production technology, and its certified 7nm 56Gbps and 112Gbps Serializer/Deserializer (SerDes) for high performance deep learning applications, complete product development and production.
In response to this cooperation, Karl Freund, senior analyst at Moor Insights&Strategy, a market research organization, said that the combination of 7nm process and data stream technology will help to create a processing platform that can meet the stringent requirements of AI applications. He is also optimistic about Broadcom's advanced processes. Knowledge of node design can assist Wave Computing in the development of AI technology.