Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
3
0
4
Achieving 51.7% area reduction and 81.5% power savings, SPARX redefines edge AI efficiency without compromising security or privacy.
Squeezing more performance out of low-precision AI hardware doesn't have to mean sacrificing efficiency: a novel bit-partitioning technique allows a single multiplier to handle both FP8 and FP4 formats with near-perfect hardware utilization.
Forget bulky GPUs at the edge: CORVET achieves 4.83 TOPS/mm2 and 11.67 TOPS/W with a CORDIC-powered, mixed-precision vector engine.