Kalray has introduced a DPU processor for AI and data processing, delivering up to 10x performance boosts, setting a new standard in neural network control and edge computing.
Kalray, a provider of hardware and software technologies, has launched Coolidge2, a new version of its 3rd generation Coolidge Data Processing Unit (DPU) processor. The company claims that the processor is designed for elevated performance in artificial intelligence (AI) and general data processing. The high-speed DPU enhances the performance of large language model solutions such as ChatGPT. The processor occupies a distinctive position in the data-intensive storage market and is optimised to complement Kalray’s current range of NVMe storage solutions. It offers unique capabilities and precise control over neural network processing, including memory transfers and utilisation of computing capabilities. The DPU delivers a significant performance boost, up to 10 times higher than the previous version. This remarkable improvement is particularly beneficial for applications involving deep learning algorithms in neural networks, especially image processing.
The DPU processor is based on Massively Parallel Processor Array (MPPA) architecture. The patent architecture boasts capabilities that grant precise control over neural network processing. From memory transfers to maximising the utilisation of the processor’s computing power, every element has been optimised to elevate the overall performance of the processor architecture. It is designed to deliver even higher performance levels in handling AI and general data processing tasks. The processor offers advanced edge computing and efficiently manages data-intensive storage workloads. The processor provides a blend of computing power, energy efficiency, programming simplicity, high-speed interfaces, parallel execution of multiple applications, deterministic operation, and real-time on-the-fly data processing.
“We are very excited to announce the tape-out of Coolidgeâ„¢2, in line with our roadmap. This is an essential milestone, marking the start of the manufacturing process and the culmination of a technological challenge. Coolidgeâ„¢2 will enable us to offer enterprises an extremely competitive alternative for data processing and AI in particular, with one of the best ratios for performance/energy consumption/price in the inference and edge computing markets”, says Eric Baissus, CEO of Kalray.
For more information, click here.