For safety-critical applications like advanced driver assistance systems (ADAS), autonomous vehicles, and medical devices as well as for equipment that relies on continuous uptime, performance monitoring and predictive maintenance of the silicon are hugely beneficial. For example, aging effects may degrade the silicon that operates a vehicle’s braking system, eventually leading to brake failure if not detected in advance. Similarly, aging effects or a lack of optimization can cause power utilization to go up, impacting battery life of portable IoT devices. Voltage can spike when there’s a malicious attack, while changes in data traffic on certain data buses can raise suspicions. With SLM technology integrated onto the chips, these scenarios can be flagged in time to prevent or mitigate negative outcomes.
Based on the more established product lifecycle management methodology, the emerging SLM applies the same holistic approach and types of capabilities to ICs. According to a 2020 research paper by Moor Insights & Strategy, “Analytics can improve design calibration, accelerate yield improvements, reduce testing time and time to market and, most importantly, predict failures or deterioration in the field.” The process involves embedding sensors and monitors into the silicon chips—this could encompass process/voltage/temperature (PVT) sensors, design-for-test (DFT) and built-in self-test (BIST) sensors, structural and functional monitors, embedded on-chip analysis tools, and data transport resources. The data collected is then funneled into a centralized database, where analytics engines extract useful insight to optimize the silicon throughout its lifecycle.
Consider, for example, the testing lifecycle phase. Chip data pulled off the test floor can be analyzed, with optimization insights fed back into the tester or other design tools to apply the recommended optimizations.
While solutions are well established for the early stages of the silicon lifecycle, there remains opportunity to enhance in-field optimization. To optimize hardware once it has been deployed in the field, whether that’s a car, a data center, or an IoT device, an SLM agent needs to be installed in the system to monitor it and access its data. This is where real-time, AI-based optimization comes into play. As it has positively impacted so many application areas, chip design included, AI provides the analytical prowess and speed to provide continuous system optimization.