Samsung and Baidu to start production of an AI chip next year

If you’ve been keeping up with mobile chipset developments in recent years, then you are likely already familiar with the importance and benefits of manufacturing process and node minification. It is a broad and very complex topic with scientist and engineers currently racing towards the coveted 5nm mark and working on solutions like Extreme Ultraviolet lithography (EUV).

While traditionally Samsung has been acquiring and applying its in-house expertise in this regard to mobile ARM silicon, the skill set is very much transferable and this is where this new partnership with Baidu stems from. It marks an expansion of Samsung’s foundry services beyond mobile and into datacenter, HPC and AI applications. To quote Ryan Lee, vice president of Foundry Marketing at Samsung Electronics:

We are excited to start a new foundry service for Baidu using our 14nm process technology … Baidu KUNLUN is an important milestone for Samsung Foundry as we’re expanding our business area beyond mobile to datacenter applications by developing and mass-producing AI chips. Samsung will provide comprehensive foundry solutions from design support to cutting-edge manufacturing technologies, such as 5LPE, 4LPE, as well as 2.5D packaging.

Baidu might be the lesser known party in this collaboration, but mainly for geopolitical reasons. It is a leading Chinese-language Internet search provider with massive cloud and server infrastructures and services. Baidu KUNLUN is the name of the company’s new AI chip, built on its advanced XPU, a home-grown neural processor architecture for cloud, edge, and AI.

Samsung and Baidu are starting production of a cutting-edge AI chip next year

Thanks to Samsung’s 14nm process as well as I-Cube (Interposer-Cube) package solution, this new chip promises to be three times faster than the conventional GPU/FPGA-accelerated solutions for running Baidu’s neural models. Baidu KUNLUN will offer 512GBps memory bandwidth via Samsung’s I-Cube technology, connected to fast HBM 2 memory. KUNLUN will be capable of up to 260 Tera operations per second (TOPS) at 150 watts of power consumption. Which is impressive compared to existing GPU/FPGA solutions. OuYang Jian – architect and Baidu also commented on the joint project:

We are excited to lead the HPC industry together with Samsung Foundry … Baidu KUNLUN is a very challenging project since it requires not only a high level of reliability and performance at the same time, but is also a compilation of the most advanced technologies in the semiconductor industry. Thanks to Samsung’s state of the art process technologies and competent foundry services, we were able to meet and surpass our goal to offer superior AI user experience.

Samsung is already working on future advancements for the packaging technologies going into Baidu KUNLUN, like redistribution layers (RDL) interposer and 4x, 8x HBM integrated package.


Leave a Reply

Your email address will not be published. Required fields are marked *