Balanced inference efficiency and performance

Getting started

Optimized for the most cost- and power-sensitive designs, Ethos-N57 delivers premium AI experiences in mainstream phones and DTVs. With the highest performance, an open-source software framework and the largest AI ecosystem, the Arm AI platform makes it easy to develop on Arm.

Download the Ethos-N57 datasheet:

Download now

Key features

Balanced Performance

Delivers up to 2 TOP/s of performance using 1024 8-Bit MACs. 

Optimized Design

Drives up to 225% convolution performance uplift using Winograd on 3x3 kernels, delivering up to 90% MAC utilization.

High Efficiency

Internally distributed SRAM stores data close to the compute elements to save power and reduce DRAM access.

Futureproof

Supports a wide range of existing Machine Learning (ML) operations and future innovations through firmware updates and compiler technology.

Arm Ethos-N Block Diagram

Ethos-N57 mainstream ML inference processor contains eight compute engines

Key benefits

  • Supports a variety of popular neural networks, including CNNs and RNNs, for classification, object detection, image enhancements, speech recognition and natural language understanding
  • Reduces system memory bandwidth by 1.5-3x through clustering sparsity and workload tiling, with lossless compression for weights and activations on select networks
  • Maximizes the number of parameters stored on-chip by storing compressed weights and activations in local SRAM and decompressing them on the fly
  • Leverages sparse power gating techniques to reduce power by up to 50%
  • Improves performance and extends battery life through intelligent data management techniques to minimize memory movement with up to 90% of accesses on chip
  • Supports TrustZone system security to safeguard sensitive data with support for secure and non-secure modes

Specifications

Key features Performance (at 1GHz)
 2 TOP/s
MACs (8x8) 1024
Data types Int-8 and Int-16
Network support CNN and RNN
Efficient convolution Winograd support
Sparsity Yes
Secure mode TEE or SEE
Multicore capability 8 NPUs in a cluster
64 NPUs in a mesh
Memory system Embedded SRAM 512 KB
Bandwidth reduction
 Extended compression technology, layer/operator fusion
Main interface
 1xAXI4 (128-bit), ACE-5 Lite
Development platform Neural frameworks TensorFlow, TensorFlow Lite, Caffe2, PyTorch, MXNet, ONNX
Neural operator API Arm NN, AndroidNN
Software components Arm NN, neural compiler, driver and support library
Debug and profile Layer-by-layer visibility
Evaluation and early prototyping Arm Juno FPGA systems and cycle models

Ethos-N comparison table

    Ethos-N77
 Ethos-N57
 Ethos-N37
Key features Performance (at 1GHz)
4 TOP/s 2 TOP/s 1 TOP/s
MAC/Cycle (8x8) 2048 1024
512
Data types
 Int-8 and Int-16
Network support CNN and RNN
Efficient convolution
 Winograd support
Sparsity Yes
Secure mode
 TEE or SEE
Multicore capability 8 NPUs in a cluster
64 NPUs in a mesh
Memory system Embedded SRAM 1-4 MB 512 KB 512 KB
Bandwidth reduction 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
  Neural operator API Arm NN, AndroidNN
  Software components Arm NN, neural compiler, driver and support library
  Debug and profile Layer-by-layer visibility
  Evaluation and early prototyping Arm Juno FPGA systems and cycle models

Get support

Arm support

Arm training courses and on-site system-design advisory services enable licensees to efficiently integrate the Ethos-N57 processor into their design to realize maximum system performance with lowest risk and fastest time-to-market.

Arm training courses   Arm Design Reviews   Open a support case

Community Blogs

Blog post image
Using Arm v8 for Vision at the Edge

Read about the work to maximize performance using the Arm v8 architecture. Specifically, by taking advantage of the NEON (DSP instruction) and a Floating-Point Unit (FPU).

18 days ago by Mary Bennion
Blog post image
TensorFlow World PyBadge

TensorFlow is powering everything from data centers to edge devices, across industries. With TensorFlow 2.0 and the evolving ecosystem of tools and libraries, it is doing it all so much easier. At this year's TensorFlow World, Google and Arm are…

21 days ago by Matthew Du Puy
1
Blog post image
Optimizing AI workloads with Ethos NPUs

The new Ethos-N NPU series addresses the performance needs of the most demanding ML use cases to deliver a superlative UX without compromising battery life.

27 days ago by Dylan Zika
Blog post image
New Arm ML guide: Deploying a quantized TensorFlow Lite MobileNet V1 model

Announcing an ML how-to guide which gives an end-to-end solution on using the Arm NN SDK. It acts as a great starting point for mobile app developers who want to trial the Arm NN SDK.

2 months ago by Darren Doyle
Blog post image
AI Basics: Training vs Inference – What’s the Difference?

So you know your ML from your AI? Great. Want to know more about training and inference? Read on...

2 months ago by Hellen Norman
Blog post image
BFloat16 processing for Neural Networks on Armv8-A

The next revision of the Armv8-A architecture will introduce Neon and SVE vector instruction designed to accelerate Neural Networks using the BFloat16 format.

2 months ago by Nigel Stephens

Community Forums

Answered hardfault error
  • Cortex-M4
 0 votes 949 views 4 replies Latest 3 days ago by 42Bastian Schick Answer this
Answered MPS2+ Expansion ports possible frequency and usage
  • ANSI
  • FPGA
  • iOS
  • Cortex-M
  • Cortex-M Prototyping System (V2M-MPS2)
 0 votes 3529 views 6 replies Latest 4 days ago by kfzhang Answer this
Answered System Frequency for CortexA35
  • Cortex-A35
  • Armv8-A
 0 votes 3253 views 1 replies Latest 18 days ago by vstehle Answer this
Answered mips of cortex a53 0 votes 5026 views 6 replies Latest 18 days ago by vstehle Answer this
Answered Address of core register on M0+ core 0 votes 849 views 2 replies Latest 20 days ago by M_T Answer this
Answered How vRestoreContextOfFirstTask works 0 votes 1099 views 3 replies Latest 21 days ago by Shibasis Answer this
Answered hardfault error Latest 3 days ago by 42Bastian Schick 4 replies 949 views
Answered MPS2+ Expansion ports possible frequency and usage Latest 4 days ago by kfzhang 6 replies 3529 views
Answered System Frequency for CortexA35 Latest 18 days ago by vstehle 1 replies 3253 views
Answered mips of cortex a53 Latest 18 days ago by vstehle 6 replies 5026 views
Answered Address of core register on M0+ core Latest 20 days ago by M_T 2 replies 849 views
Answered How vRestoreContextOfFirstTask works Latest 21 days ago by Shibasis 3 replies 1099 views