Scaling towards kilo-core processors with asymmetric high-radix topologies

N. Abeyratne, R. Das, Qingkun Li, K. Sewell, B. Giridhar, R. G. Dreslinski, D. Blaauw, T. Mudge
2013 2013 IEEE 19th International Symposium on High Performance Computer Architecture (HPCA)  
In this paper, we explore the challenges in scaling on-chip networks towards kilo-core processors. Current low-radix topologies optimize for fast local communication, but do not scale well to kilo-core systems because of the large number of routers required. These increase both power and hop count. In contrast, symmetric high-radix topologies optimize for global communication with fewer hop counts, but degrade local communication with their large, slow routers. To address both local and global
more » ... ommunication optimizations independently, we decouple the interconnect design using asymmetric high-radix topologies. By setting a design goal of matching router speed with wire speed, our proposed topologies use fast medium-radix routers to optimize for local communication and a few slow high-radix routers that reduce hop count to optimize for global communication. Our asymmetric high-radix designs are enabled by recently proposed Swizzle-Switches, which allow us to achieve performance scalability within realistic power budgets. We propose and evaluate two asymmetric high-radix topologies: Super-Star (asymmetric folded Clos) and Super-StarX (asymmetric folded Clos with superimposed mesh). Our evaluations show that the best performing asymmetric high-radix topology improves average network latency over a mesh by 45% while reducing the power consumption by 40%. When compared to symmetric high-radix topologies network throughput is improved by 2.9× while still providing similar latency benefits and power efficiency. 978-1-4673-5587-2/13/$31.00 ©2013 IEEE
doi:10.1109/hpca.2013.6522344 dblp:conf/hpca/AbeyratneDLSGDBM13 fatcat:azgwgm33qffgnekwnmqcgk264m