Hardware for Artificial Intelligence
Spanning Intelligent Computations across Server to Edge Devices
Although pervasive, almost all computing platforms today are synonymous with the ‘von-Neumann’ computer model. A key characteristic of the von-Neumann model is its clear demarcation of a ‘computing unit,’ from the “storage unit”. which is physically separated from a ‘storage unit.’ This inevitably leads to frequent energy and throughput of intensive data movement between the two units. The resulting ‘Memory-Wall Bottleneck’ renders state-of-the-art computing platforms inefficient for Artificial Intelligence (AI) applications. Therefore, it is not surprising that almost all the payloads for AI computations, such as learning and inference, are performed on a remote server. In turn, this creates a ‘Cognitive-Wall Bottleneck,’ wherein edge devices are deprived of intelligent computations and rely on remote server-like compute resources for critical decision making.
At ASIC Lab, we aim to develop alternate high-throughput and energy-efficient computing paradigms using emerging and existing Silicon technologies, thereby making AI computations accessible to both server and edge devices. In addition to charge-based computing, we’re exploring alternate state variables, including electron spin, photonics, and phononics, for unconventional data-intensive processing. Our approach enables a cross-layer optimization across materials, devices, circuits, architecture, and algorithms for future truly pervasive AI applications.
Hardware for Smart Sensors
Investigating Next Generation Sensors with Alternate State Variables
Defense, health care, space, automobiles, and many other areas find extensive usage of sensors for data collection. Sensors also form the backbone that provides data to Machine Learning (ML) and AI algorithms for developing intelligent systems. Two key recent developments that have made sensor research more relevant than ever before are developments in materials and associated growth techniques, and the availability of smart ML/AI algorithms, which has brought in more data interpretation and analysis ability.
At ASIC Lab, we’re exploring a two-pronged approach for the next generation of sensors. This includes (1) exploring novel sensing methods and devices and (2) bringing computation closer to sensors for efficient localized data interpretation. Energy, area, and cost-efficient sensors can fuel multi-modal data interpretation, improving the overall robustness.
Smart and Secure Manufacturing
Expanding Security of Integrated Circuits
Perhaps the most important aspect that led to the unprecedented dimensional scaling of transistors following Moore’s Law is constant innovation and advancement in semiconductor manufacturing. With the chip complexity efficiently scaling to billions of transistors, the globalization of the semiconductor business has exposed chip design and fabrication to security threats. Hardware trojans, IP theft, etc. pose serious security threats to the entire computing eco-system.
In response to such an imminent need to enable secure, trusted manufacturing of Integrated Circuits (ICs), ASIC Lab is focused on developing both design-time and fabrication-time techniques to protect ICs from any unwanted tampering.
Novel Device Integration
Toward Fine-Grained Heterogeneous Computing
The semiconductor process is a technology that has been well-established over the years. However, the increase in computing demand has necessitated exploration of various integration approaches such as 2.5D and 3D.
At ASIC Lab, we focus on heterogenous and monolithic integration and backend design of emerging logic/memory technologies for enabling novel circuit and system architectures. Our expertise spans across circuit design, floor planning, chip layout, tapeout and post-silicon verification. We collaborate with advanced semiconductor foundries as well as the USC cleanroom.