High-Radix Scalable Modular Crossbar Switches

As process technologies have scaled, the increasing number of processor cores and memorieson a single die has also driven the need for more complex on-chip interconnection networks.Crossbar switches are primary building blocks in such networks-on-chip, as they can be usedas fast single-stage networks or as the core of the router switch in multi-stage networks.While crossbars offer non-blocking, single-hop, all-to-all communication, they tend to scalepoorly with the number of nodes due to the latency and energy of the long wires and highradixmultiplexor structures needed. In this work, we investigate how to improve crossbarperformance, energy-efficiency, and scalability.To better understand the design space and scaling limitations, we have developed an on chipswitch modeling tool calibrated using circuit-level simulations. The tool enables a designspace exploration showing how area, power, and performance vary across radix, data width,wire parameters, and circuit implementation. In addition to conventional design options,we examined capacitively coupled low-swing signaling to improve to energy consumption ofthe I/O wires. This exploration shows that the main bottlenecks are the long I/O wires andthe key to improving the performance and efficiency is to minimize the area. Using theseinsights, we present modular crossbar switches that can perform better at high radices thanthe monolithic designs. The modular sub-blocks are arranged in a controlled flow-through,pipelined scheme to eliminate global connections and maintain linear performance scalingand high throughput. Modularity also enables energy savings via deactivation of unusedI/O wires.To evaluate our design, we implemented a prototype radix-64 modular crossbar switchtestcip in 40nm CMOS bulk process. The testchip operates at 2.38GHz at 1V nominalsupply voltage and consumes 1.2W power. It offers 2.2X better throughput and 2.4X betterenergy-efficiency than published state of the art designs. We further evaluated modularcrossbar networks with the proposed crevaluation tool. The proposed design achieves more than 90% saturation throughput withan internal speed up of 1.5, supports high data line rates, and offers lower average networklatency compared to conventional crossbars. Evaluation results show that modular crossbarsare scalable to high-radices while still offering high-performance, energy-efficiency and onehopsimplicity.ossbar switches using BookSim2, a network on chip