Gyrfalcon Technology Inc. emerged from relative stealth mode to become the first company with commercially available hardware that delivers high-performance, low-power AI processing at the edge, shipping to customers. GTI’s generally available patented Matrix Processing Engine is designed to enable high-performance AI processing for IoT and other edge equipment within the smallest energy envelope at an industry-leading ratio.
Unlike AI-enabling technologies offered by some of the industry’s cloud players, GTI has taken an “edge-first” approach to its IoT technology. By enabling core AI processing in the edge equipment – vs. relaying data to be processed by data center AI and back to the edge via the cloud – GTI’s solution avoids the latency in edge equipment receiving actionable insight and reduces the cost involved in transmitting data. In addition, the solutions increase privacy as in many cases data never leaves the edge equipment.
GTI provides unparalleled edge AI capabilities thanks to its patented Matrix Processing Engine, which runs AI solutions on an ASIC for domain-specific applications. Coupled with a two-dimensional matrix with AI processing in Memory, MPE enables the industry’s best performance-to-power usage meaning equipment can operate at high speeds to handle AI quickly, with low energy consumption.
The company’s first-generation product uses just 300mW of power to deliver a superior 9.3 TOPS/W that can process audio and video input. With its small size and the ability to combine between two and 32 chips on one board for heavy compute loads or separate task handling, the 2801S enables the broadest range of use cases and equipment designs, at a price point 10X lower than competitive hardware. Therefore, not only does the high performance differentiate AI products, but companies can also achieve significantly improved margins. This is a breakthrough for companies with an interest in manufacturing mass-market products at scale, where the technology needs to be made small and inexpensive without compromising performance, such as in a smart home lock or camera.
“Balancing the cost-performance-energy equation has been a challenge for developers looking to bring AI-enabled equipment to market at scale,” stated Dr. Lin Yang, chief scientist, GTI. “The GTI founding team has been watching the industry struggle with this challenge for decades, and believe that our AI Processing in Memory and Matrix Processing Engine provide an elegant solution to avoid having to make trade-offs. By deploying APiM and MPE on a standard, commoditized ASIC, GTI is enabling our customers to bring innovative, AI-enabled devices to the masses.”
“We are paving the way for the next wave of AI products to make it to market,” said Kimble Dong, CEO of GTI. “We recognized that device makers were compromising on essential design variables in AI-enabled equipment and have sought to solve this over the past few decades. Our offering marries our “edge-first” approach with ultra-fast AI data processing technology, low power consumption and a small chip design to enable the best AI experience and performance at a low cost, within any AI use case, physical fit and deployment.”
GTI is addressing the fast-growing IoT category, which is expected to grow on average 12 percent in the number of devices annually according to IHS Market. The company’s enablement of the smallest and most cost-effective IoT devices with edge AI is enhanced by self-learning capabilities necessary to maximize productivity. Data models with chip-adaptive training device setup is as simple as plugging in, powering up and getting devices started with a voiceprint, selfie or gesture. After initial setup, smart devices can take their cues from their surroundings, for example, a lock learning the residents of a home through arrival and departure patterns.