Skip to main content

Syntiant Core 2 Achieves Outstanding Results in Latest MLPerf Tiny v1.1 Benchmark Suite

IRVINE, Calif., July 05, 2023 (GLOBE NEWSWIRE) -- Syntiant Corp., a leader in edge AI deployment, today announced that its Syntiant Core 2™ programmable deep learning architecture delivered the lowest power energy performance across three categories in the most recent MLCommons’ MLPerf™ Tiny v1.1 benchmark suite, which measures how quickly a trained neural network can process new data for extremely low-power devices in the smallest form factors.

Syntiant Core 2 supports the execution of the most common network types, ranging from dense and convolution layers to architectures with depth-wise separable layers including MobileNet, to networks with skipped connections such as ResNet. This flexibility allowed the Syntiant architecture to be evaluated across three distinct categories, each tested with a different network architecture.

Syntiant’s results were performed at two operating points to demonstrate maximum throughput and minimum energy. In the keyword spotting category in the high-performance setting (1.1V core supply voltage and 98.7MHz clock frequency), Syntiant’s solution delivers 1.5 ms latency while consuming 43.8 uJ/inference, approximately 25x more efficient than any other submitted system. At the low-energy setting (0.9V / 30.7MHz), Core 2 requires 31.5 uJ and 4.4 ms per inference, 30x the energy efficiency than the next lowest submission.

“The results of the MLPerf Tiny v1.1 benchmarks demonstrate the Core 2’s compelling throughput and energy performance, as well as its versatility by executing the benchmark’s visual wake word and image classification workloads,” said Jeremy Holleman, chief scientist at Syntiant. “Just as in earlier versions of the benchmark for keyword spotting, the Syntiant Core 2 again achieved energy consumption orders of magnitude lower than other devices for the vision oriented tasks. The benchmarks also demonstrate the easy migration path from pre-trained models using Syntiant’s training development kit.”

The Core 2 delivers 4.1 ms latency while consuming 97.2 uJ/inference at 98.7MHz in the visual wake words category, while achieving 12.7 ms latency with 71.7 uJ at the low-energy setting. The architecture outperformed the nearest submitter in energy performance 20X and 25X, respectively.

For image classification workloads, the Core 2 performs at 5.1 ms latency and 139.4 uJ at 98.7 MHz clock frequency, approximately 25X the efficiency than the lowest DSP in the field. It also achieves 16 ms latency and 101.8 uJ at 30.7MHz, more than 35X the nearest submission. Full results of the MLPerf Tiny v1.1 benchmark suite can be downloaded here.

About the Syntiant Core 2

The Syntiant Core 2 is a highly flexible, ultra-low-power deep neural network inference engine with a highly configurable front-end interface embedded in the company’s second-generation Neural Decision Processors, such as the NDP200, NDP120 and NDP115. Built to run multiple applications simultaneously with minimal power consumption, including edge AI vision features such as person detection, object classification, motion tracking and occupancy monitoring, the Core 2 can also perform highly accurate ambient sensing and audio processing on-device. With its multimodal capabilities, the architecture also enables multi-sensor fusion, voice command recognition, acoustic event detection, infrared detection, anomaly and tamper detection, as well as other audio, motion and pressure sensing applications. Additionally, the Core 2 supports a variety of task-dependent under the hood optimizations for sparsity and time series that speed up inference and reduce power requirements without extensive post-training network optimization, shortening time-to-product by months or years as compared to more constrained and power-intensive solutions.

About MLCommons

MLCommons is an open engineering consortium with a mission to make machine learning better for everyone through benchmarks and data. The MLPerf benchmarks are full system tests that stress machine learning models, software and hardware and optionally measure power usage. The open-source and peer-reviewed benchmark suites provide a level playing field for competition that drives innovation, performance and energy-efficiency for the entire industry. For additional information on MLCommons and details on becoming a Member or Affiliate, visit MLCommons or contact

About Syntiant 
Founded in 2017 and headquartered in Irvine, Calif., Syntiant Corp. is a leader in delivering hardware and software solutions for edge AI deployment. The company’s purpose-built silicon and hardware-agnostic models are being deployed globally to power edge AI speech, audio, sensor and vision applications across a wide range of consumer and industrial use cases, from earbuds to automobiles. Syntiant’s advanced chip solutions merge deep learning with semiconductor design to produce ultra-low-power, high performance, deep neural network processors. Syntiant also provides compute-efficient software solutions with proprietary model architectures that enable world-leading inference speed and minimized memory footprint across a broad range of processors. The company is backed by several of the world’s leading strategic and financial investors including Intel Capital, Microsoft’s M12, Applied Ventures, Robert Bosch Venture Capital, the Amazon Alexa Fund and Atlantic Bridge Capital. More information on the company can be found by visiting or by following Syntiant on Twitter @Syntiantcorp or LinkedIn. 

Media Contact:

George Medici

Primary Logo

Data & News supplied by
Stock quotes supplied by Barchart
Quotes delayed at least 20 minutes.
By accessing this page, you agree to the following
Privacy Policy and Terms and Conditions.