ZDNet on Nostr: Making GenAI more efficient with a new kind of chip ========== EnCharge AI, in ...
Making GenAI more efficient with a new kind of chip
==========
EnCharge AI, in partnership with Princeton University, has received an $18.6 million grant from the US's Defense Advanced Research Projects Agency (DARPA) to advance low-power circuitry for generative artificial intelligence (GenAI) inference. EnCharge AI plans to sell its own accelerator chip and system boards for AI in edge computing. The company's approach combines digital and analog circuitry, with the analog portion performing the accumulate operation, resulting in improved energy efficiency. EnCharge AI's prototypes have shown a significant improvement in energy efficiency, processing 150 trillion operations per second per watt. The company aims to solve the data movement problem by reducing memory usage and performing calculations closer to memory using analog circuitry. The prototypes have been able to achieve this by using capacitors instead of measuring current, resulting in a less noisy and more economical approach. EnCharge AI plans to announce details of its first product later this year, with a focus on the inference opportunity. The company believes that energy efficiency will be a critical factor in the market for AI inference.
#EnchargeAi #PrincetonUniversity #Darpa #GenerativeArtificialIntelligence #Genai #Inference #EdgeComputing #AnalogCircuitry #EnergyEfficiency #MemoryUsage #Capacitors
https://www.zdnet.com/article/making-genai-more-efficient-with-a-new-kind-of-chip/
==========
EnCharge AI, in partnership with Princeton University, has received an $18.6 million grant from the US's Defense Advanced Research Projects Agency (DARPA) to advance low-power circuitry for generative artificial intelligence (GenAI) inference. EnCharge AI plans to sell its own accelerator chip and system boards for AI in edge computing. The company's approach combines digital and analog circuitry, with the analog portion performing the accumulate operation, resulting in improved energy efficiency. EnCharge AI's prototypes have shown a significant improvement in energy efficiency, processing 150 trillion operations per second per watt. The company aims to solve the data movement problem by reducing memory usage and performing calculations closer to memory using analog circuitry. The prototypes have been able to achieve this by using capacitors instead of measuring current, resulting in a less noisy and more economical approach. EnCharge AI plans to announce details of its first product later this year, with a focus on the inference opportunity. The company believes that energy efficiency will be a critical factor in the market for AI inference.
#EnchargeAi #PrincetonUniversity #Darpa #GenerativeArtificialIntelligence #Genai #Inference #EdgeComputing #AnalogCircuitry #EnergyEfficiency #MemoryUsage #Capacitors
https://www.zdnet.com/article/making-genai-more-efficient-with-a-new-kind-of-chip/