Chiplet based, high-performance, low-power AI chips

July 12, 2021 // By Jean-Pierre Joosting
Chiplet based, high-performance, low-power AI chips
Flexible and scalable Chiplet architecture significantly improves system capability and efficiency while reducing costs for AI training, inferencing, high-precision compute, and large-scale image processing.

A leading Chiplet based technology developer, Blue Ocean Smart System became a founding member of the Artificial Intelligence Computing Power Alliance (ICPA) at the World Artificial Intelligence Conference (WAIC), aiming to accelerate the commercialization of advanced artificial intelligence technology.

Founded in 2019, Blue Ocean Smart System was established with a deeply experienced technical team in broad fields such as GPU, NPU, multimedia, and high-performance computing. The company's core team has an average of more than 20 years of experience with a team that has shipped an accumulated 10 billion semiconductor SoCs with a peak annual volume of more than 2 billion units shipped in a single year. Blue Ocean Smart System is dedicated to developing original technology and products using an innovative Chiplet architecture. With this unique approach it claims to be the first company in the world to deliver two complementary Chiplets in Q4 of 2021 using a 12nm process. Blue Ocean has created a single design to support from edge to cloud to optimize and to conquer the challenges for each of these markets.

Using Blue Ocean's proprietary Chiplet architecture and technology, a GPGPU chiplet called BlueFin will support end-to-end high-precision compute (FP32) applications, scaling from 2 TFLOPS to 128 TFLOPS with the flexibility required by our customers. A second chiplet from Blue Ocean, an NPU chip called BlueDanio, supports a wide range of highly computational and energy-efficient AI training and inference applications with a flexible solution from 50 TOPS to 1,200 TOPS with a single design. At the same time, an integrated hardware and software development environment will greatly enhance the competitiveness of customer products, reduce product development costs, and shorten development cycles for the markets focused on AI training, inferencing, high-precision compute, and large-scale image processing.


Vous êtes certain ?

Si vous désactivez les cookies, vous ne pouvez plus naviguer sur le site.

Vous allez être rediriger vers Google.