FOR IMMEDIATE RELEASE
Nuvoton
and Skymizer Achieve MLPerf Tiny Benchmark Leadership Using NuMaker
M467HJ
Hsinchu, Taiwan, June 27,
2023 - Nuvoton Technology Corporation and Skymizer have
achieved MLPerf Tiny Benchmark leadership in the Cortex-M4 MCU category
through a combination of the NuMaker M467HJ evaluation board, and
Skymizer's ONNC ML optimization. The Nuvoton M467-series MCU uses an
Arm Cortex-M4F microcontroller operating at 200MHz, 67% faster than
typical Cortex-M4F speeds, which in collaboration with ML software
optimization using Skymizer's neural network technologies, enabled it
to achieve class-leading inference performance.
A
Highly Capable Platform for Embedded Solutions Using Edge AI
MLCommons is an independent machine learning performance benchmarking
association, which has become the trusted standard for evaluating ML
performance on a wide range of systems. The MLPerf Tiny Benchmark
focuses on practical ML use cases that run on embedded systems,
including visual wake word, keyword spotting, image classification, and
audio anomaly detection.
The Nuvoton M467 series boasts a large 512KB of SRAM, 1024KB of Flash
memory, and a rich set of integrated system features and peripherals,
including DSP, FPU, DMA, CAN-FD, I2S, USB, camera interface, crypto
accelerators, and 10/100 Ethernet MAC. These features make it an ideal
choice for systems designers to build devices that integrate tinyML
into various emerging applications, such as smart home automation,
smart city and infrastructure, light edge AI in IoT, smart
manufacturing, and many more emerging use cases.
Skymizer's ONNC compiler played a crucial role in optimizing the
machine learning software stack for the M467-series Cortex-M4F,
resulting in remarkable gains in inference speed and enhancing the
overall performance of machine learning applications.
The collaboration between Nuvoton Technology Corporation and Skymizer
showcases their commitment to providing cutting-edge solutions for ML
in power-efficient embedded systems. By combining hardware excellence
with software optimization, they have unlocked new possibilities for
ML-based applications on MCU, enhancing capabilities across industries.
The outstanding results achieved in the MLPerf Tiny Benchmark's
Cortex-M4 MCU segment highlight Nuvoton's and Skymizer's dedication to
pushing the boundaries of machine learning performance in
resource-constrained environments.
Learn more about the Nuvoton NuMaker M467HJ evaluation board:
https://www.nuvoton.com/board/numaker-m467hj/
See the MLPerf Tiny Benchmark result here:
https://mlcommons.org/en/inference-tiny-11/
About
Nuvoton Technology Corporation
Nuvoton Technology Corporation (Nuvoton) was founded to bring
innovative semiconductor solutions to the market. Nuvoton was spun-off
as a Winbond Electronics affiliate in July 2008 and went public in
September 2010 on the Taiwan Stock Exchange (TWSE). Nuvoton focuses on
the developments of microcontroller, microprocessor, smart home, cloud
security, battery monitoring, component, visual sensing and IoT with
security ICs and has strong market share in Industrial, Automotive,
Communication, Consumer and Computer markets. Nuvoton owns 6-inch wafer
fabs equipped with diversified processing technologies to provide
professional wafer foundry services. Nuvoton provides products with a
high performance/cost ratio for its customers by leveraging flexible
technology, advanced design capability, and integration of digital and
analog technologies. Nuvoton values long term relationships with its
partners and customers and is dedicated to continuous innovation of its
products, processes, and services. Nuvoton has established subsidiaries
in the USA, China, Israel, India, Singapore, Korea and Japan to
strengthen regional customer support and global management. For more
information, please visit http://www.nuvoton.com.
About
Skymizer
Skymizer is an AI system software company, established in 2013. Our
main focus is on providing a complete software development kit (SDK)
for Neural Networks. Based in Taipei and Hsinchu Science Park,
Skymizer's AI model compiler, called ONNC, is versatile and can be
utilized across various hardware structures, ranging from single IoT
core systems to complex multicore heterogeneous systems. In response to
the growing demand for AIoT, Skymizer offers comprehensive optimization
and hardware/software co-design solutions for MCUs. Moreover, Skymizer
is committed to broadening our horizons beyond MCUs and venturing into
new AIoT research fields, including micro NPUs and On-Device Learning.
For more information, please visit https://www.skymizer.com.