GlobalFoundries, Witmem, and Microchip Announce Sampling of WTM-8 8K/120 FPS Video Processor for High-Performance, Low-Power AI Applications

The innovative AI video processing solution, built on GF’s 12LP+ and 28ESF3 platforms, features SuperFlash® technology to efficiently store and process millions of AI interference parameters 

MALTA, N.Y. and BEIJING, October, 24, 2024 — GlobalFoundries^® (GF), Witmem Technology, and Microchip Technology via its subsidiary Silicon Storage Technology (SST) today announced the sampling of Witmem’s WTM-8 8K/120FPS AI video processor, which is ideal for high performance, low power consumption applications such as AR/VR headsets and advanced 8K TVs and video screens. 

The WTM-8 achieves superior AI video processing performance by combining two chiplets, the digital logic chiplet built utilizing GF’s 12LP+ platform, and the analog compute-in-memory (aCIM) chiplet built utilizing GF’s 28ESF3 platform leveraging SST’s SuperFlash® technology. The WTM-8 efficiently stores and processes up to 64 million AI inference parameters within the SuperFlash^® memory array, eliminating unnecessary data movement and power. The WTM-8 is fully reprogrammable to future-proof the processor as AI video models rapidly evolve. 

The WTM-8 features include: 

• Supports AI-based video enhancement solutions including frame interpolation, super resolution, asynchronous Timewarp, object recognition and a multitude of complex edge computing scenarios. 
• Achieves 4K/8K @ 60/120FPS frame interpolation, super resolution, HDR wide dynamic range and superior noise reduction for video displays. 
• Best in class FPS/watt performance, achieving 8K/120FPS AI video processing consuming less than 2 watts.  
• Best in class FPS/$ cost effectiveness, outperforming GPU and DRAM based solutions that cost up to 10x more than the WTM-8. 
• SuperFlash memory array for optimal efficiency and processing.  

“Our partnership with Witmem to bring their WTM-8 AI video processor to the stage of sampling underscores GF’s commitment to creating future-ready solutions for edge AI devices,” said Michael Hogan, chief business unit officer at GF. “Witmem is redefining the edge computing landscape by combining chiplets built on GF 12LP+ with an industry-first analog compute-in-memory built on our 28ESF3 platform to achieve unparalleled performance.” 

“Witmem has discovered an ideal AI product collaboration partner in GF,” added Dr. Shaodi Wang, Witmem’s Founder and CEO.  “We look forward to further collaboration with GF to release the WTM-8 to mass production, and to pave the way for a new generation of low power, cost effective AI processor products for high-performance imaging, video display, and augmented reality applications.”  

Customers interested in GF’s 12LP+ and 28 ESF3 platform solutions should inquire at www.gf.com/about-us/contact-us for more information. 

Customers interested in Witmem’s WTM-8 and AI audio and video processing solutions should access the Witmem website located at www.witmem.com/WTM-8.php and contact sales@witmem.com for more information. 

Customers interested in SST’s ESF1, ESF3, and memBrain IP offerings should access the SST website located at www.sst.com.  

About GF 

GlobalFoundries (GF) is a leading manufacturer of essential semiconductors the world relies on to live, work and connect. We innovate and partner with customers to deliver more power-efficient, high-performance products for the automotive, smart mobile devices, internet of things, communications infrastructure and other high-growth markets. With our global manufacturing footprint spanning the U.S., Europe, and Asia, GF is a trusted and reliable source for customers around the world. Every day, our talented and diverse team delivers results with an unyielding focus on security, longevity, and sustainability. For more information, visit www.gf.com. 

About Witmem Technology 

Witmem Technology, founded in 2017, is dedicated to computing-in-memory technology development for edge AI and has received continuous financial funding from famous VCs. In 2022, Witmem released its first commercial computing-in-memory chip that has been used in several mass-production products showing 10-50X advantage in energy efficiency. Witmem will continuously enhance computing performance and build an energy-efficient infrastructure for the popularization of AI technology. www.witmem.com/index.php