The semiconductor industry has historically concentrated on microprocessors' (MPU) computational performance; however, recent years have shifted focus towards microcontrollers (MCU). This shift is essentially a result of the demands on edge computing. Designers are no longer singularly focused on maximizing performance but instead want the best balance between performance and power as possible. 

The MCU market is expected to see a 10.6% CAGR through 2027. Image used courtesy of Precedence Research

With this demand in mind, Texas Instruments has been hard at work trying to achieve the best of both worlds: MCU power efficiency with MPU performance. The result of this mashup is the release of its newest MCU portfolio, the Sitara AM2x. 

Last week, All About Circuits had the chance to hear directly from Mike Pienov and Sonia Ghelani, managers of the Sitara portfolio, to hear about the developments firsthand. 

Why MCUs? 

Developing technologies like the IoT demand edge computing for reasons including lower latencies, decreased system complexity, and increased security. However, the tradeoff here is power and battery life. 

How can a device meet the high computing demands of AI while meeting power specs that allow it to run for months off a coin cell battery? 

Where AI in standard computers and data centers rely on microprocessors and GPUs, edge devices can't afford such power-intensive computing infrastructure. 

MCU vs. MPU implementations. Image used courtesy of NWES

Instead of power-intensive infrastructures, many edge devices utilize more task-specific, low-power MCUs as their processing units. Of course, MCUs are low power mainly because of their decreased computing capacity, a tradeoff that has inspired a new direction in the AI and semiconductor industries. The ultimate goal would be MCU power consumption with MPU performance. 

Fields like TinyML have focused on achieving this goal through software optimization, while the semiconductor industry focuses on hardware improvements. The latter is the approach TI seems to be taking. 

TI’s Sitara AM2x Portfolio 

Today, TI announced its new Sitara AM2x portfolio of MCUs, explicitly aimed at edge computing. Pienov tells us the portfolio is trying "to address what we see as a kind of intermediate place or a gap between what traditional microcontrollers that are available in the industry are able to accomplish and then also what higher performance processors have traditionally been able to address."

AM243x block diagram. Image used courtesy of Texas Instruments

This new family of MCUs is said to achieve 10x the computing capability of traditional, flash-based MCUs. Much of the performance increase is accredited to the integration of non-traditional blocks into the MCU. 

These include on-chip high-performance RAM, signal processing accelerators, advanced analog control, and communication peripherals to allow for accurate and improved real-time sensing and control. 

Sitara AM2x's fundamental building blocks. Image used courtesy of Texas Instruments

On top of this, the MCU includes up to four Arm Cortex-R5F cores. Each core can run up to 800 MHz, with a total device consumption said to be less than 1 W at this level of performance. 

As Ghelani stated, "this translates to over 6000 DMIPS. If you look at traditional [MCUs]...typically they're in the hundreds of DMIPS range. So this kind of really gives customers a huge step function increase in terms of performance improvement." 

Further, the Sitara AM243x family includes support for multiple gigabit industrial Ethernet protocols and time-sensitive networking supporting protocol stacks including EtherNet/IP, EtherCAT, PROFINET, IO-Link Master. The inclusion of these protocols for high-performance networking shows TI's direct focus on industrial and factory-type applications. 

Blurring the Lines

As TI tells us, its goal with its new Sitara family is to blur the line between MCU and MPU, a crucial demand of on-edge computing applications. With its new portfolio, TI hopes to enable a new generation of industrial/factory applications with the balance between power and performance necessary for the future of automation.