Advanced On-chip InterconnectsPrincipal Investigators: Abderazek Ben Abdallah (PI), Khanh N. Dang (PI)
Future System-on-Chip (SoC) will contain hundreds of components made of processor cores, DSPs, memory, accelerators, and I/O all integrated into a single die area of just a few square millimeters. Such complex systems will be interconnected via a novel on-chip interconnect closer to a sophisticated network than to current bus-based solutions. This network must provide high throughput and low latency while keeping area and power consumption low. Our research solves several design challenges to enable massively parallel many-core systems, specifically investigating fault-tolerance, 3D-TSV integration, photonic communication, low-power mapping, and adaptive routing.
HotCluster: Thermal-Aware Defect Recovery for 3D-ICs
TSV is essential for low-power 3D-ICs, but lifetime reliability is threatened by high operating temperatures that accelerate fault rates. HotCluster is a hotspot-aware self-correction platform for clustering defects in 3D-NoCs. It strategically places redundant TSV groups based on regional heat profiles. By redistributing redundancies, HotCluster reduces hardware overhead by 60% compared to uniform distribution while maintaining system stability under 50% defect rates.
IaSiG: Online Detection and Correction of TSV Failures
IA-SiG addresses interconnect reliability in 3D-ICs through an online detection and correction method. By analyzing output syndromes, the system localizes and fixes defective TSVs in real-time. Integrating IA-SiG onto 3D-NoCs using a grid-search empirical method allows for temperature-aware redundancy insertion, maintaining the required Mean Time to Failure while minimizing hardware costs.
Scalable Photonic Networks-on-Chip & Wavelength Shifting
Photonic links provide massive bandwidth but face scalability issues. We propose a novel Wavelength-Shifting mechanism that combines the benefits of photonic speed with scalable routing. This approach reduces the number of required photonic devices by 60% and achieves latency and power performance an order of magnitude better than electro-assisted architectures, regardless of traffic patterns or communication distance.
TSV-OCT: Scalable Online Multiple-TSV Defects Localization
Unlike traditional methods that require system interruptions, TSV-OCT (On-Communication Test) detects open and short defects in parallel with active data transactions. Utilizing a statistical detector and isolation-and-check algorithms, it localizes up to 5 times more defects than standard ECC techniques without halting real-time applications.
3D-OASIS-FT: Adaptive Fault-Tolerant Architecture
3D-FTO ensures message delivery in many-core systems even in the presence of transient and permanent faults. The architecture uses a Hybrid-Look-Ahead (HLAFT) routing algorithm and Random-Access-Buffers (RAB) to eliminate deadlocks and relieve congestion. This system achieves a 12.5% reduction in latency and 11.8% throughput enhancement with integrated power management.
Registered Patents / 特許
Collaborations & ContactWe collaborate with the following companies: Contact: Abderazek Ben Abdallah (E-mail: benab@u-aizu.ac.jp) |