Dual-Stripline Configuration for Efficient Routing in Chiplet Interconnects
By Shekar Geedimatla,Jayaprakash Balachandran,Midhun Vysakham,Srinivas Venkataraman,Shalabh Gupta
Routing density is becoming in big challenge in die-to-die interconnects. In this paper, we propose use of the dual-stripline configuration for routing signals in high-density interconnects. The scheme can improve the routing density by up to 33% when compared with the conventionally used stripline configuration. To address the challenges of crosstalk due to the proximity between vertically adjacent signal lines, halfpitch offset between lines on vertically adjacent layers has been proposed. The proposed routing scheme has been validated using 3D full-wave electromagnetic simulations. The simulations show that the scheme can be used for increasing the routing density in the Bunch-of-wires interface by 25%, while meeting all the Bunch-of-wires channel specifications, which include eye-opening value above 60% unit interval at a bit error rate of 10−15, with data rates of 16 Gbps per wire.
To read the full article, click here
Related Chiplet
- Optical Chiplet
- Reverse Gearbox Chiplet
- FlexGen Multi-Die Smart Network-on-Chip (NoC) IP
- Ncore Multi-Die Interconnect IP
- Integrated voltage regulator (IVR) chiplet
Related Technical Papers
- Chiplet-Escape: An Efficient Obstacle-Avoiding Escape Routing Method for Die-to-Die Interconnections in Chiplet-Based Designs
- Link Quality Aware Pathfinding for Chiplet Interconnects
- ChipLight: Cross-Layer Optimization of Chiplet Design with Optical Interconnects for LLM Training
- LASH-PAT: Deadlock-Free Adaptive Shortest-Path Routing for Inter-Chiplet Interconnects
Latest Technical Papers
- LASH-PAT: Deadlock-Free Adaptive Shortest-Path Routing for Inter-Chiplet Interconnects
- ThermLeT: Transformer-based Temperature Prediction for 2.5D Chiplet Architecture
- Technical Challenges for Ultra Fine Pitch Hybrid Bonding
- DT-MATCH: Defect- and Topology-Aware Chiplet Matching for Yield Optimization in Heterogeneous Integration
- Starscream: Reimagining LLM Sharding on AMD Instinct GPUs with GPU Partitioning