The Lightspeeur 5801 offers up to 2.8 TOPS of performance while using 224 mW of power, and thus delivers up to 12.6 TOPS/W.
Gyrfalcon did not announce which foundry or manufacturing process it is using to make the 5801. The 2801 was manufactured by TSMC in a 28nm CMOS process with MRAM (see TSMC embedded MRAM is key to Gyrfalcon AI chip). The Lightspeeur 2802 was reported implemented in 22nm CMOS.
The Lightspeeur 5801 has four times the image input size when compared to the Lightspeeur 2801, the chip Gyrfalcon introduced in 2017.This allows higher resolution images to be processed easily, in-line with increasing image sensor resolution and the clock speed has been enhanced to 50MHz to 200MHz.
Targeted include image recognition, object detection and tracking, natural language processing, natural language understanding, business intelligence, facial recognition and visual analysis for consumer electronics, smart home & office, smart city, industrial, enterprise and data center solutions.
LG uses the Lightspeeur 5801 in the Q70 mobile phone which it launched in October. LG uses the chip for imaging and video features
The Lightspeeur 5801 is offered with a Model Development Kit, and development of applications are possible with a Software Development Kit (SDK) which is available for Linux X86_64, Microsoft Windows, Android platforms and also ARMv7.l and ARMv8 instruction sets. It supports convolutional neural networks such as ResNet, MobileNet and VGG16, and TensorFlow, PyTorch & Caffe frameworks.
Related links and articles: