TY - GEN
T1 - Direct-modulated optical networks for interposer systems
AU - Jokar, Mohammad Reza
AU - Zhang, Lunkai
AU - Dallesasse, John M.
AU - Chong, Frederic T.
AU - Li, Yanjing
N1 - Publisher Copyright:
© 2019 Association for Computing Machinery.
PY - 2019/10/17
Y1 - 2019/10/17
N2 - We present a new interposer-level optical network based on direct-modulated lasers such as vertical-cavity surfaceemitting lasers (VCSELs) or transistor lasers (TLs). Our key observation is that, the physics of these lasers is such that they must transmit significantly more power (21) than is needed by the receiver. We take advantage of this excess optical power to create a new network architecture called Rome, which splits optical signals using passive splitters to allow flexible bandwidth allocation among different transmitter and receiver pairs while imposing minimal power and design costs. Using multi-chip module GPUs (MCM-GPUs) as a case study, we thoroughly evaluate network power and performance, and show that (1) Rome is capable of efficiently scaling up MCM-GPUs with up to 1024 streaming multiprocessors, and (2) Rome outperforms various competing designs in terms of energy efficiency (by up to 4) and performance (by up to 143%).
AB - We present a new interposer-level optical network based on direct-modulated lasers such as vertical-cavity surfaceemitting lasers (VCSELs) or transistor lasers (TLs). Our key observation is that, the physics of these lasers is such that they must transmit significantly more power (21) than is needed by the receiver. We take advantage of this excess optical power to create a new network architecture called Rome, which splits optical signals using passive splitters to allow flexible bandwidth allocation among different transmitter and receiver pairs while imposing minimal power and design costs. Using multi-chip module GPUs (MCM-GPUs) as a case study, we thoroughly evaluate network power and performance, and show that (1) Rome is capable of efficiently scaling up MCM-GPUs with up to 1024 streaming multiprocessors, and (2) Rome outperforms various competing designs in terms of energy efficiency (by up to 4) and performance (by up to 143%).
UR - http://www.scopus.com/inward/record.url?scp=85076603821&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85076603821&partnerID=8YFLogxK
U2 - 10.1145/3313231.3352368
DO - 10.1145/3313231.3352368
M3 - Conference contribution
AN - SCOPUS:85076603821
T3 - Proceedings of the 13th IEEE/ACM International Symposium on Networks-on-Chip, NOCS 2019
BT - Proceedings of the 13th IEEE/ACM International Symposium on Networks-on-Chip, NOCS 2019
PB - Association for Computing Machinery
T2 - 13th IEEE/ACM International Symposium on Networks-on-Chip, NOCS 2019
Y2 - 17 October 2019 through 18 October 2019
ER -