GrAI One – ‘s werelds eerste AI Chip voor ultra-lage latentie at the Edge
SAN JOSE, Californië (BUSINESS WIRE) – GrAI Matter Labs (alias GML) kondigt vandaag GrAI One aan – ‘s werelds eerste AI-processor die geoptimaliseerd is voor ultralage latency en lage stroomverwerking at the Edge.
GrAI One, de eerste chip gebaseerd op de NeuronFlow-technologie van het bedrijf, vermindert drastisch de latentie van de toepassing, bijvoorbeeld, het vermindert de end-to-end latentie voor diepe leernetwerken zoals PilotNet tot de orde van microseconden. NeuronFlow combineert het dynamische dataflow-paradigma met neuromorfe computing om massaal parallelle in-netwerkverwerking te produceren. GrAI One is gericht op responskritische edge-toepassingen in autonome navigatie, mens-machine-interactie en slimme gezondheidszorgmarkten.
De volledig digitale chip meet 20mm2 in TSMC 28nm technologie en implementeert een netwerk van 196 neuron kernen met lokaal neuron/synaps geheugen voor een totaal van 200.000 neuronen. Het biedt een GPIO-interface voor het ontladen van AI-werklasten van een hostprocessor die kritisch zijn voor de latentie. Bij 100% neuron kerngebruik verbruikt GrAI One zo weinig als 35mW.
GrAI One – the World’s 1st AI Chip for Ultra-Low Latency at the Edge
SAN JOSE, Calif.–(BUSINESS WIRE)– GrAI Matter Labs (aka GML) today announces GrAI One – the world’s first AI processor optimized for ultra-low latency and low power processing at the edge.
GrAI One, the first chip based on the company’s NeuronFlow technology, drastically reduces application latency, for instance, it reduces the end-to-end latencies for deep learning networks such as PilotNet to the order of microseconds. NeuronFlow combines the dynamic dataflow paradigm with neuromorphic computing to produce massively parallel in-network processing. GrAI One is targeted at response-critical edge applications in autonomous navigation, human-machine interaction and smart healthcare markets.
The fully digital chip measures 20mm2 in TSMC 28nm technology and implements a mesh of 196 neuron cores with local neuron/synapse memories for a total of 200,000 neurons. It provides a GPIO interface to offload latency-critical AI workloads of a host processor. At 100% neuron core utilization, GrAI One consumes as little as 35mW.
GrAI One is supported by GML’s GrAIFlow software development kit. It is capable of both conventional program execution and machine learning computation via industry-standard languages like TensorFlow, Python and C++. The kit includes compiler, simulator, debugger, graphical editor and compute and network APIs. GrAIFlow is available now for early access customers for architecture evaluation and application programming. GrAI One is expected to begin sampling in Q4 2019.
The company’s CEO Ingolf Held said: “GrAI One processes edge AI applications orders of magnitude faster than traditional architectures while maintaining a power footprint suitable for battery powered devices. With GrAIFlow we offer our customers the opportunity to explore new ultra-low latency use cases, thereby bringing their innovation to every device on the edge.”
About GrAI Matter Labs
GML’s mission is to bring ultra-low latency, low power programmable AI processors for sensor analytics and machine learning to every device on the edge. GrAI Matter Labs has offices in Paris / France, in Eindhoven / The Netherlands and in San Jose / Silicon Valley. It is led by a team of visionary and seasoned engineers, and is backed by leading investors including iBionext, 360 Capital Partners and 3T Finance. For more information, please visit www.graimatterlabs.ai.
View source version on businesswire.com: https://www.businesswire.com/news/home/20191023005732/en/
Contacts
Europe
Ashoka Visweswara
US
Mahesh Makhijani
Check out our twitter: @NewsNovumpr