The Ethos-U series of microNPUs brings innovative AI applications with high performance and efficiency to embedded systems.

Ethos-U series characteristics

Performance efficiency
Unlock new vision, voice, and vibration use cases with minimal impact to existing design across a range of ML performance requirements (GOP/s to TOP/s).

Flexible integrations
Build low-cost, highly efficient systems in high-performance Cortex-A and Neoverse systems or low-power Cortex-M-based systems.

Optimized design
Highly optimized with integrated DMA, MAC array, and element wise engines.

Unified software and tools
Develop, deploy, and debug AI applications with the Arm Endpoint AI solution using a common toolchain across Arm Cortex and Ethos-U processors.

Ethos-U series comparison

Key features Performance (at 1GHz)
64 to 512 GOP/s
512 GOP/s to 1 TOP/s
MACs (8x8) 32, 64, 128, 256
256, 512
Memory system Internal SRAM 18 to 50 KB
55 to 104 KB
System interfaces
Two 64-bit AXI
Two 128-bit AXI 
External memory
SRAM and flash
SRAM, DRAM, and/or flash
Development platform Recommended host CPU
Cortex-M55, Cortex-M7, Cortex-M4, Cortex-M33 Cortex-M55, Cortex-M7
Operating systems
Bare-metal or RTOS
Bare-metal, RTOS, or Linux