The Arm Ethos-N processor series delivers the highest throughput and efficiency in the lowest area for Machine Learning inference from cloud to edge to endpoint.


Ethos-N series characteristics

Key features of the Ethos-N family of devices:

  • Enables simultaneous, high-performance AI use cases in Full HD, real-time object detection, super resolution and segmentation
  • Tackles a broad spectrum of Machine Learning (ML) requirements, with performance spanning from 1 TOP/s to > 650TOP/
  • Scalable multicore technology allows up to eight NPUs to be configured in a tightly coupled cluster, up to 64 NPUs in a mesh configuration
  • Arm AI platform supports all popular frameworks and operators and provides flexible future capability through software updates
  • Offers online and offline inference deployment with heterogeneous support using opensource software such as Arm NN, Arm Compute Library, and TV
  • Provides layered security to protect both ML models and input data such as biometrics for financial payments
  • Allows for support and protection of memory and easy handling of multiple users via tight system integration through the ACE-Lite master port and optional SMMU integration
  • Leverage AI partners developing optimized algorithms ahead of hardware availability
  • Allows early performance feedback with Ethos-N NPU Static Performance Analyzer (SPA) and Arm DS-5 Streamline

Download the Ethos-N datasheet:

Download


Ethos-N comparison

    Ethos-N78
Ethos-N77
Ethos-N57
Ethos-N37
Key features Performance (at 1GHz)
10. 5. 2. 1 TOP/s 4 TOP/s 2 TOP/s 1 TOP/s
MAC/Cycle (8x8) 4096, 2048, 1024, 512 2048 1024
512
Efficient convolution
Winograd support delivers 2.25x peak performance over baseline
Configurability 90+ design options Single product offering
Network support CNN and RNN
Data types
Int-8 and Int-16
Sparsity Yes
Secure mode
TEE or SEE
  Multicore capability 8 NPUs in a cluster
64 NPUs in a mesh
Memory system Embedded SRAM 384KB – 4MB 1-4 MB 512 KB 512 KB
Bandwidth reduction Enhanced compression Extended compression technology, layer/operator fusion, clustering, and workload tilling
Main interface 1xAXI4 (128-bit), ACE-5 Lite
Development platform Neural frameworks TensorFlow, TensorFlow Lite, Caffe2, PyTorch, MXNet, ONNX
  Inference deployment Ahead of time compiled with TVM
Online interpreted with Arm NN
Android Neural Networks API (NNAPI)
  Software components Arm NN, compiler and support library, driver
  Debug and profile Heterogeneous layer-by-layer visibility in Arm Development Studio Streamline
  Evaluation and early prototyping Ethos-N Static Performance Analyzer (SPA), Arm Juno FPGA systems, Cycle Models