Why power architectures constrain New Space AI missions
Satellite AI outpaces legacy power systems. Modern compute needs sub-volt, high-amp power that traditional architectures simply can't deliver until now
Current multiplier technology advances enable new AI processor power solutions
Presented by Siyu Chai, Product Manager Global AI/Computing Market, Vicor
Leveraging a Factorized Power Architecture (FPA) and current multiplication, Vicor provides solutions for lateral power delivery, vertical power delivery and a new solution, lateral-vertical power delivery. Siyu addressed the problem of impedances and their effects on PDN losses in relation to the processor core voltages. Advances in power module packaging, next generation current multiplication and current density was also be addressed. These technologies enable wafer scale and clustered computing AI systems with current requirements in the tens of thousands of amps and will unleash xPU performance. The presentation deck is available to download.
Originally presented at OCP China Day 2022 on August 22, 2022
Resources
Resources for the design engineer
AI for space applications
AI has the potential to change our world with applications delivering real impact. Learn how Spacechips is powering the next generation of AI in space
Spacechips high current transponder powers on-orbit AI-driven communication
Smaller satellites with sophisticated computational capabilities is in demand. Learn how low-noise, AI-enabled power is driving creative new applications
Current multipliers: The obvious choice for powering AI processors and other demanding applications
AI processors need to handle low-voltage, high-current demand, which can cause power system bottlenecks. Learn how current multiplication can change that.