As new generation of intelligent processors, Kalray MPPA processors have the capability to analyze on the fly, and in an intelligent manner, a very large amount of information to make decisions and interact in real time with the outside world.
Coolidge processor will support the company's ambitions on Data Centre and Automotive markets. A broad range of high-performance applications will take advantage of Coolidge, including embedded applications, intelligent cars and autonomous vehicles, advanced driver assistance systems (ADAS), aerospace & defense, medical equipment robots, data centre, networking, high performance computing, high performance storage, as well as, machine learning and artificial intelligence (AI) applications.
"I am proud to announce the tape-out of Coolidge, our third generation of intelligent processor", stated Eric Baissus, CEO of Kalray. "Since the IPO of Kalray in June 2018, it has been one of the main focus of the company to bring to life this new product. Autonomous vehicles, artificial intelligence and new intelligent data centres are driving a growing and demanding need for performance, lower power consumption, real time capabilities, combined with safety and security. Coolidge brings unique innovations and improvement to serve these needs. Numerous customers have identified the potential of our breakthrough solution and are eager to get the first samples of Coolidge."
Coolidge has been designed to address the challenges and constraints of the new generation of intelligent systems:
Coolidge overcomes many limitations of alternative solutions. Its standard programmability and flexibility makes it an appealing choice versus FPGA, as requirements for updates in the field is growing and cost pressure is increasing. The ability to run simultaneously several real-time and compute-intensive heterogeneous applications on Coolidge within a low-power budget, brings a compelling advantage over solutions such as GPUs. The patented parallel architecture of Coolidge enables massive acceleration of many algorithms, low-latency protocols or Artificial Intelligence applications. Associating Coolidge with other multi-core processors provides the opportunity for system makers to massively improve performance and optimize their Total Cost of Ownership (TCO).
Coolidge will be produced using a 16nm FinFET technology. It includes 80 cores - Kalray 64-bit time predictable and energy efficient VLIW cores, and supports the unique capability to combine multiple Coolidge together in order to bring larger amount of cores depending the need of the targeted application. In addition, 80 patented co-processors are tightly integrated within each core to bring a blazing boost of performance for Artificial Intelligence and compute intensive applications, to reach up to 4 Tera FLOPS and 25 TOPS - which is about 25x the performance of the previous generation - with the total chip consuming less than 20W. This superior ratio makes Coolidge best-in-class for high performance embedded applications. Many advanced high-speed interfaces such as 100 GbE or PCIe Gen4, allow for the integration in high-performance systems, while specialized interfaces such as CAN assure seamless compatibility with targeted applications such as autonomous cars.
Coolidge will be delivered to Kalray customers and partners with an upgraded version of Kalray Software Development Kit, AccessCore including Kalray Artificial Intelligence code generation tool (KaNN) as well as reference boards.