The Lightspeeur 5801 is Gyrfalcon's 4th generation best-of-breed AI solution. It delivers the best AI performance of any neural accelerator on the market. In addition to improved TOPS/Watt performance, it accepts 4x larger image input using more neural models.

The Lightspeeur® 5801S AI Accelerator delivers high performance combined with the efficiency of low power use that is optimal for edge AI devices. Combined with low cost, this chip supports designs intended for mass market device applications such as mobile, embedded, and loT.

This accelerator uses CNN (Convolutional Neural Networks) with a host processor to execute workloads for AI tasks such as Natural Language Processing (NLP), object detection, image classification, speech recognition, and other applications.

This processor uses GTl s patented APiM™ (Al Processing in Memory) technology, which supports on-chip parallelism that eliminates memory bottlenecks, for high performance devices with industry-leading power efficiency.

The Lightspeeur® family supports popular Al frameworks such as Caffe and TensorFlow. To speed system development, GTI provides Software Development kits (SDKs) and turn-key reference designs backed by system verification hardware, software, and tools.