Arm Machine Learning Processor 

Industry-leading performance and efficiency for inference at the edge.

Machine Learning Processor Block Diagram.

The Arm Machine Learning processor is an optimized, ground-up design for machine learning acceleration, targeting mobile and adjacent markets. The solution consists of state-of-the-art optimized fixed-function engines that provide best-in-class performance within a constrained power envelope.

Additional programmable layer engines support the execution of non-convolution layers, and the implementation of selected primitives and operators, along with future innovation and algorithm generation. The network control unit manages the overall execution and traversal of the network and the DMA moves data in and out of the main memory.

Onboard memory allows central storage for weights and feature maps, thus reducing traffic to the external memory and therefore, power.


Key Features

  • Specially designed to provide outstanding performance for mobile with up to 4 TOPs; additional optimizations provide a further increase in real-world use cases.
  • Best-in-class efficiency at >4 TOPs/W.
  • Programmable layer engines for future-proofing.
  • Highly tuned for advanced geometry implementations.
  • Scalable onboard memory reduces external memory traffic.
  • Arm NN works with Android NNAPI to provide a translation layer between major neural network frameworks, such as TensorFlow and Caffe, and the Arm Machine Learning processor, as well as other Arm IP.

Performance

  • Greater than 4 TOPs in mobile environments.
  • Propriety optimizations provide further increase in real-world use cases.
  • Efficiency of >4 TOPs/W.

Find out more

Find out more about Arm Machine Learning processor

Contact us

To learn more about Machine Learning on Arm, visit our ML Developer community.

Learn more

 


Key Benefits

  • Most efficient solution to run neural networks.
  • Designed for the mobile and adjacent markets.
  • Optimized, ground-up design for machine learning acceleration.
  • Best-in-class performance with state-of-the-art, fixed-function engines.
  • Programmable engines for future innovation and algorithms.
  • Massive efficiency uplift from CPUs, GPUs, DSPs and accelerators.
  • Completes Arm’s heterogeneous Machine Learning platform solution.
  • Enabled by open-source software.
  • Industry-leading performance in thermally- and cost-constrained environments.

Applications

Mobile

AR/VR

IoT

Smart camera

Healthcare

Medical

Logistics

Small area

Robotics

Home

Consumer 

Drones

Wearables

Webinar - Project Trillium: Optimizing ML Performance for any Application

Project Trillium is a suite of Arm IP designed to deliver scalable ML and neural network functionality at any point on the performance curve, from sensors, to mobile, and beyond. 

 

Find out more

Community Forums

Suggested answer Axi4 Write Transaction 0 votes 32 views 1 replies Latest 10 hours ago by vstehle Answer this
Not answered The Peripherals memory map of FVP_BaseR_Cortex-R52x1
  • cortex-r52
  • Fixed Virtual Platforms (FVPs)
  • PrimeCell UART (PL011)
0 votes 48 views 0 replies Started yesterday by Jex1x Answer this
Suggested answer Basic difference between "Generic User Guide" and "Technical Reference Manual" 0 votes 109 views 1 replies Latest 2 days ago by 42Bastian Schick Answer this
Suggested answer VMSAv8-64 and spinlock 0 votes 825 views 3 replies Latest 4 days ago by Ciro Donnarumma Answer this
Not answered Event Recorder with STM32F0
  • Cortex-M0
  • SRAM
  • STM32 F0
  • event
0 votes 74 views 0 replies Started 5 days ago by NSharp Answer this
Answered A question aboout Monitor Vector Base Address Register(MVBAR)
  • Cortex-A9
  • TrustZone
0 votes 383 views 6 replies Latest 5 days ago by scribnote5 Answer this
Suggested answer Axi4 Write Transaction Latest 10 hours ago by vstehle 1 replies 32 views
Not answered The Peripherals memory map of FVP_BaseR_Cortex-R52x1 Started yesterday by Jex1x 0 replies 48 views
Suggested answer Basic difference between "Generic User Guide" and "Technical Reference Manual" Latest 2 days ago by 42Bastian Schick 1 replies 109 views
Suggested answer VMSAv8-64 and spinlock Latest 4 days ago by Ciro Donnarumma 3 replies 825 views
Not answered Event Recorder with STM32F0 Started 5 days ago by NSharp 0 replies 74 views
Answered A question aboout Monitor Vector Base Address Register(MVBAR) Latest 5 days ago by scribnote5 6 replies 383 views