GBT is implementing machine learning driven, pattern matching technology for its Epsilon - The EE

GBT is implementing machine learning driven, pattern matching technology for its Epsilon

San Diego, United States – GBT Technologies Inc., is implementing a machine learning driven, pattern matching technology within its Epsilon, microchip’s reliability verification and correction electronic design automation (EDA) tool. Design rules are getting increasingly complex with each new process node and design firms are facing new challenges in the physical verification domain.

One of the major areas that are affected by the process physics, is reliability Verification (RV). Microchips are major components nearly in every major electronics application. Civil, military and space exploration industries require reliable operations for many years, and in severe environments. High performance computing systems require advanced processing with high reliability to ensure the consistency and accuracy of the processed data. Complex integrated circuits are in the heart of these systems and need to function with high level of dependability.

Particularly in the fields of medicine, aviation, transportation, data storage and industrial instrumentation, microchip’s reliability factor is crucial. GBT is implementing new machine learning driven, pattern matching techniques within its Epsilon system with the goal of addressing the advanced semiconductor’s physics, ensuring high level of reliability, optimal power consumption and high performance. As Epsilon analyses the layout of an integrated circuit (IC), it identifies reliability weak spots, which are specific regions of an IC’s layout, and learns their patterns. As the tool continues analysing the layout it records problematic zones taking into account the pattern’s orientations and placements.

In addition, it is designed to understand small variations in dimensions of the pattern, as specified by the designer or an automatic synthesis tool. As the weak spots are identified, the tool will take appropriate action to modify and correct them. A deep learning mechanism will be performing the data analysis, identification, categorisation, and reasoning while executing an automatic correction.

The Machine Learning will understand the patterns and record them in an internal library for future use. Epsilon’s pattern matching technology will be analysing the chip’s data according to a set of predefined and learned-from-experience rules. Its cognitive capabilities will make it self-adjust to newest nodes with new constraints and challenges, with the goal of providing quick and reliable verification and correction of an IC layout.

The company released a video which explain the potential functions of the Epsilon tool here.

“The ability to analyse and address advanced IC’s reliability parameters is necessary to mitigate risk of system degradation, overheating, and possible malfunction. It can affect microchip’s performance, power consumption, data storage and retrieval, heat and an early failure which may be critical in vital electronic systems. Epsilon analyses a microchip data for reliability, power and electrothermal characteristics, and performs auto-correction in case violations found.

We are now implementing an intelligent technology for Epsilon with the goal of utilising pattern matching algorithms to formulate a smart detection of reliability issues within integrated circuits layout. The new techniques will analyse and learn weak spots within microchip’s data, predicting failure models that are based on the process’ physics and electrical constraints knowledge. It will take into consideration each device’s function, connectivity attributes, electrical currents information, electrothermal factors and more to determine problematic spots and perform auto-correction.

Particularly for FinFet and GAA FET (Gate All Around FET) technologies, a device’s functionality is developed with major reliability considerations ensuring power management efficiency, optimal thermal analysis aiming for long, reliable life span. Using smart pattern matching methods, we plan to improve reliability analysis, achieving consistency and accuracy across designs within advanced manufacturing processes.

As dimensions of processes shrink, IC’s layout features become much more complex to analyse for electrical phenomenon. To provide an intelligent answer for these complexities, we are implementing deep learning-based pattern matching technology with the goal of ensuring efficient, ‘green’ microchip’s power consumption, higher performance, optimised thermal distribution, and ultimately superior reliability” states Danny Rittman, the company’s CTO.

There is no guarantee that the company will be successful in researching, developing or implementing this system. In order to implement this concept, the company will need to raise adequate capital to support its research and, if researched and fully developed, the company would need to enter into a relationship with a third party that has experience in manufacturing, selling and distributing this product. There is no guarantee that the company will be successful in any or all of these critical steps.

Follow us and Comment on Twitter @TheEE_io

By continuing to use the site, you agree to the use of cookies. more information

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.