VeriSilicon has announced that Blue Ocean Smart System, an AI chiplet and SoC design company, has deployed high-performance AI chips with a scalable chiplet architecture powered by VeriSilicon’s processors.
These processors include the CC8400 General-Purpose Graphics Processor Unit (GPGPU), VIP9400 Neural Network Processor (NPU) and VC8000D Video Processor Unit (VPU). Blue Ocean’s chiplet-based products aim to address the requirements of automotive applications, data centers and high-performance computing applications.
The highly scalable CC8400 GPGPU from VeriSilicon provides excellent versatility and performance, and supports half-precision 16-bit floating-point and full-precision 32-bit floating-point, in addition to 8-bit, 16-bit and 32-bit fixed-point data precision. The smart solution delivers robust computing capabilities with optimized area and power consumption. Offering support for transformer models, the company’s VIP9400 NPU delivers AI capabilities for automotive applications and data centers.
Blue Ocean’s targeted HPC, AI and computing platforms are capable of meeting the requirements of a wide range of use cases with an expandable set of chiplet ICs that are universally programmable to support products for use from edge to cloud. By leveraging Blue Ocean’s architecture and patented BxLink technologies, VeriSilicon integrates its microarchitecture, hardware and software development environment to provide fully scalable solutions that enhance the competitiveness of customer products and reduce product development costs. It can also shorten development cycles for the markets focused on AI training, such as those within high-precision computing, large-scale image processing, autonomous driving and advanced robotics.
“The demand for high-performance computing in data center and automotive applications is increasing rapidly,” explained John Rowland, president of Blue Ocean. “VeriSilicon’s extensive portfolio of high-performance processors and their expertise in chip implementation makes them an excellent partner for our chiplet products’ deployments. The recent advancements in AI applications, such as ChatGPT, have further fueled the demand for more GPGPU and NPU computing.”
“VeriSilicon’s chiplet IP series is developed based on our high-performance processors, including GPGPU, NPU and VPU technologies that have been deployed across multiple generations of data center products,” commented Wei-Jin Dai, executive vice president and general manager of IP division at VeriSilicon. “Our VPU is already being used in 12 of the top 20 cloud platforms worldwide. Development of maturing UCIe and BOW die-to-die interfaces and decreasing chiplet packaging costs will accelerate the use of chiplets for complex functions.”