EdgeCortix’s Sakura-I chip selected by BittWare for AI inference solutions
BittWare, an embedded hardware manufacturer, has incorporated the Sakura-I chip into its AI inference-optimized ASIC-based acceleration solutions. Sakura-I is a processor created by EdgeCortix, a company specializing in edge artificial intelligence solutions.
EdgeCortix claims its Sakura-I chipset for edge artificial intelligence is highly efficient, offering low power consumption below 10 Watts in addition to low inference latency and high data throughput. The company plans to implement this solution in various industries and applications, including surveillance and security, as well as quality assurance in manufacturing facilities.
“Together with BittWare, we are delivering a complete edge AI acceleration solution for small form factor computer boxes to edge inference servers that can be used as ‘drop-in,’ highly efficient solutions in lieu of common power-hungry GPUs,” says Sakyasingha Dasgupta, the founder and CEO of EdgeCortix.
According to EdgeCortix, the Sakura-I AI processor is an excellent option for power-efficient inference at the edge. It can be used in various applications, including computer vision and generative AI. Benchmark tests show that this edge AI chip outperforms current edge GPUs in terms of inferences per second and uses less power.
The Sakura-I processor utilizes 12 nm TSMC FinFET process technology to manage data streaming from high-resolution cameras for real-time edge applications. The company says it can process generative AI models, including the transformer architecture typically employed in large language models.
“We are delighted to be partnering with EdgeCortix to bring their edge-focused Sakura-I acceleration solution to market. Extending our offerings into ASIC-based acceleration solutions optimized for AI inference is a natural step for both our business and our brand,” says Craig Petrie, the vice president of BittWare.
In 2022, EdgeCortix and BittWare joined forces to create an AI acceleration solution for Intel Agilex FPGAs. This solution is designed for deep learning inference tasks and prioritizes high performance and low latency. It offers dedicated AI compute capability of up to 20 tera-operations per second.
Article Topics
AI/ML | BittWare | chip | edge AI | EdgeCortix | energy efficiency | GPU | inference
Comments