China's Baidu joins hands with Samsung to produce leading-edge AI chip
SEOUL, AJU - Through their first foundry cooperation, Baidu, a top internet search provider in China, and Samsung Electronics, the world's largest microchip producer, have developed a cloud-to-edge artificial intelligence chip called "Baidu KUNLUN" for mass production in early 2020. Samsung said the new chip would expand an AI ecosystem and transform user experiences
Samsung said in a statement on Wednesday that the new AI chip allows Ernie, a pre-training model for natural language processing, to infer three times faster than the conventional model. With natural language processing widely being adopted from search engines to mobile smart assistants, pre-training language models like Baidu's Earnie have received strong attention.
Leveraging the chip's limit-pushing computing power and power efficiency, Baidu can effectively support a wide variety of functions including large-scale AI workloads, such as search ranking, speech recognition, image processing, natural language processing, autonomous driving, and deep learning platforms. Samsung said it would expand its foundry business into high performance computing (HPC) chips designed for cloud and edge computing.
"We are excited to lead the HPC industry together with Samsung Foundry," Jian Ouyang, a Baidu architect, was quoted as saying. "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."
Samsung said the new AI chip would offer 512 gigabytes per second memory bandwidth and supplies up to 260 Tera operations per second (TOPS) at 150 watts. TOPS is a common performance metric used for a high-performance complex systems-on-a-chip (SoC). TOPS per watt extends that measurement to describe performance efficiency. The higher the TOPS per watt the better and more efficient a chip is.
Baidu KUNLUN is built on XPU, a neural processor architecture for cloud, edge and AI, as well as Samsung's 14-nanometer process technology with its I-Cube (Interposer-Cube) package solution. I-Cube technology, which connects a logic chip and high bandwidth memory with an interposer, provides higher density bandwidth on minimum size by utilizing differentiated solutions.
"We are excited to start a new foundry service for Baidu using our 14nm process technology," Samsung's foundry marketing vice president Ryan Lee said. "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 separated its foundry business in 2017. System LSI is only responsible for R&D and design of system semiconductors, while the foundry business is in charge of consignment production. In April, Samsung revealed a massive investment worth 133 trillion won ($114 billion) by 2030 to strengthen its competitiveness in System LSI and foundry businesses, saying it aims to become the world leader in not only memory semiconductors but also logic chips.