The chip presents 512 gigabytes per second (GBps) reminiscence bandwidth and provides as much as 260 Tera operations per second (TOPS) at 150 watts. In addition, the brand new chip permits Ernie, a pre-training mannequin for pure language processing, to deduce 3 times quicker than the standard GPU/FPGA-accelerating mannequin. Leveraging the chip’s limit-pushing computing energy and energy effectivity, Baidu can successfully help all kinds of features together with large-scale AI workloads, akin to search rating, speech recognition, picture processing, pure language processing, autonomous driving, and deep studying platforms like PaddlePaddle.
Through the primary foundry cooperation between the 2 firms, Baidu will present superior AI platforms for maximizing AI efficiency, and Samsung will increase its foundry enterprise into excessive efficiency computing (HPC) chips which might be designed for cloud and edge computing.
“We are excited to lead the HPC industry together with Samsung Foundry,” stated OuYang Jian, Distinguished Architect of Baidu. “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. “
“We are excited to start a new foundry service for Baidu using our 14 nm process technology,” stated Ryan Lee, vice chairman of Foundry Marketing at Samsung Electronics. “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.”
As increased efficiency is required in numerous functions akin to AI and HPC, chip integration know-how is turning into increasingly more essential. Samsung’s I-Cube know-how, which connects a logic chip and excessive bandwidth reminiscence (HBM) 2 with an interposer, supplies increased density/ bandwidth on minimal measurement by using Samsung’s differentiated options.
Compared to earlier know-how, these options maximize product efficiency with greater than 50% improved energy/sign integrity. It is anticipated that I-Cube know-how will mark a brand new epoch within the heterogeneous computing market. Samsung can be creating extra superior packaging applied sciences, akin to redistribution layers (RDL) interposer and 4x, 8x HBM built-in bundle.
