Cortex and Ethos Explained

If you work with embedded systems, AI at the edge, or modern SoC architectures, you will eventually come across two important names from Arm: Cortex and Ethos. They often appear together in technical papers, product sheets, and chip specifications. But what exactly do they mean, and how are they different?

This guide answers the key question “what is Cortex and Ethos?” in the simplest and clearest way possible – while keeping everything technically correct and detailed enough for professionals. By the end, you will fully understand why both technologies are essential in today’s embedded and AI-driven devices.

What Is Cortex? (Simple Explanation)

Arm Cortex is a family of CPU cores that chip manufacturers can integrate into their processors. These cores run the operating system, control logic, program instructions, and almost everything that requires general computing.

Think of Cortex as the “brains” that handle everyday tasks.

Cortex processors are divided into three main groups:

Cortex-A: For powerful devices

These are the high-performance cores you find in:

• smartphones
• tablets
• smart TVs
• automotive infotainment systems
• industrial computers

Cortex-A cores run full operating systems such as Linux, Android, and Windows on Arm. They support advanced features like large caches, out-of-order execution, and SIMD (NEON) extensions.

Cortex-R: For real-time and safety

Cortex-R is built for fast and predictable real-time performance. These cores are used in:

• SSD and HDD controllers
• automotive systems
• robotics motors
• industrial control equipment

They are optimized for deterministic timing and reliability.

Cortex-M: For microcontrollers and IoT

Cortex-M cores power billions of devices worldwide, including:

• smart sensors
• wearables
• home appliances
• toys
• simple industrial devices

They are extremely low-power, small, and easy to program.

What Is Ethos? (Simple Explanation)

Arm Ethos is a family of NPUs – Neural Processing Units designed specifically for accelerating machine learning (ML) and deep learning workloads.

While Cortex cores can run AI models, they are not optimized for heavy neural network math. This is where Ethos comes in.

Ethos processors execute operations such as:

• convolution
• matrix multiplication
• activation functions
• feature extraction
• layer fusion

much faster and with far lower power consumption than a CPU.

Ethos comes in two main families:

Ethos-U: AI for microcontrollers

These “microNPUs” are designed to work next to Cortex-M cores. They make tiny devices capable of running:

• voice recognition
• gesture detection
• simple vision models
• anomaly detection
• TinyML applications

For example: a battery-powered sensor can run local AI without sending data to the cloud.

Ethos-N: AI for larger systems

This family is used in:

• smartphones
• consumer electronics
• embedded vision
• robotics
• smart cameras

Ethos-N NPUs provide high performance measured in TOPS (tera-operations-per-second), making them suitable for real-time computer vision and advanced AI tasks.

Why Arm Built Both Cortex and Ethos

Modern devices must run both general-purpose logic (apps, operating systems, control functions) and AI models (vision, speech, pattern detection).

One processor type cannot efficiently do both tasks.

• Cortex CPUs are flexible and can run anything, but they are not optimized for huge ML workloads.
• Ethos NPUs are extremely fast for AI, but they cannot run an operating system or general code.

This is why almost every modern SoC includes both:
1) Cortex to control the system
2) Ethos to accelerate AI

Together, they form a heterogeneous computing architecture: each processor does what it is best at.

Cortex vs Ethos – Clear Comparison Table

This table gives a simple and readable picture of the differences, helpful for engineers evaluating SoC architectures or students learning embedded design.

How Cortex and Ethos Work Together

Here’s a typical flow inside a modern AI-capable SoC:

1. The Cortex processor loads the AI model and prepares the data.
2. It sends the neural network layers to the Ethos NPU.
3. Ethos performs the heavy math: convolutions, pooling, matrix ops.
4. Ethos returns the inference results (for example, “object detected”).
5. Cortex handles the next actions-UI updates, control logic, networking, etc.

This teamwork is essential in:

• smart cameras
• autonomous robots
• drones
• automotive vision systems
• voice assistants
• IoT devices running TinyML

Without this combination, devices would be either too slow or too power-hungry to run modern AI locally.

Real-World Examples

Smartphones

A typical smartphone SoC includes:

• Cortex-A cores for apps, UI, OS
• Ethos-N NPU for camera AI, photo enhancement, face detection

This makes real-time AI features fast and efficient.

Tiny IoT sensors

A Cortex-M microcontroller paired with an Ethos-U55 enables:

• on-device keyword spotting
• gesture classification
• anomaly detection for predictive maintenance

And all with battery life measured in months or years.

Smart cameras and robotics

Vision algorithms run in real time thanks to Ethos accelerators while Cortex-A cores manage logic, navigation, or networking.

Why This Architecture Matters Today

AI is no longer a cloud-only feature. Devices now perform more processing on the edge, meaning:

• better privacy
• lower latency
• lower network usage
• better reliability

Without NPUs like Ethos, general CPUs like Cortex would struggle to keep up with modern AI demands.

Sources and References

To keep this explanation accurate and grounded, the following sources were used for architecture descriptions and technical details:

• Arm Developer Documentation — official architecture guides
• Arm Cortex Overview (Wikipedia)
• Arm Ethos Overview (Wikichip)
• TensorFlow Lite for Microcontrollers documentation
• ONNX Runtime AI Inference guides

These references describe the architectures and technologies that inspired the explanations above.

Conclusion: Understanding What Cortex and Ethos Really Are

To summarise everything in a clear and readable form:

• Cortex is Arm’s family of general-purpose CPU cores.
  They run your applications, the operating system, and all the control logic.
• Ethos is Arm’s family of neural processing units (NPUs).
  They accelerate deep learning tasks such as image classification, detection, and speech AI.
• Cortex + Ethos work together to deliver efficient, powerful, and flexible computing for modern edge and embedded systems.

If you are building or evaluating an SoC for AI-driven applications, understanding what is Cortex and Ethos will help you choose the right architecture, estimate performance, and design better systems.

Read more:

• Embedded System Design Services
• Robot SDK and More: Rockchip’s New Achievements
• Multimedia Boxes: What Is a Set-Top Box?
• RK3588 vs RK3588S: In-Depth Technical Comparison
• RK3688 vs RK3668: Rockchip’s Next-Generation Chips
• RK3566 vs RK3568: Comparison Guide
• Findchips: How to Search for Rockchip Products
• Lighting Power Density: Understanding LPD