TY - GEN
T1 - Migrating single FPGA chip multiprocessor with network on chip to 65nm and 45nm ASIC
AU - Hammami, O.
AU - Wang, Z.
AU - Houzet, Dominique
PY - 2011/1/1
Y1 - 2011/1/1
N2 - Multiprocessor on chip (MPSoC) with network on chip (NoC) are strongly emerging as prime candidates for complex embedded applications. In a general ESL design methodology and for significant size designs the use of prototyping and emulation through FPGA is necessary for intensive validation and test as well as careful design space exploration. Moving a design from FPGA to ASIC questions the gains and benefits which can be achieved both at an architectural level but also at the parallel programming level. In his paper we analyze the migration of an implemented, validated and tested single FPGA chip multiprocessor with network on chip towards 65nm and 45nm ASIC technologies. Our results show that although we can naturally expect an area gain, the working frequency is not significantly augmented in 45nm. This suggests that performance improvement can not be achieved by technology alone and area advantage should be exploited by selecting network on chip components with more aggressive features. This in turn affects parallel programming.
AB - Multiprocessor on chip (MPSoC) with network on chip (NoC) are strongly emerging as prime candidates for complex embedded applications. In a general ESL design methodology and for significant size designs the use of prototyping and emulation through FPGA is necessary for intensive validation and test as well as careful design space exploration. Moving a design from FPGA to ASIC questions the gains and benefits which can be achieved both at an architectural level but also at the parallel programming level. In his paper we analyze the migration of an implemented, validated and tested single FPGA chip multiprocessor with network on chip towards 65nm and 45nm ASIC technologies. Our results show that although we can naturally expect an area gain, the working frequency is not significantly augmented in 45nm. This suggests that performance improvement can not be achieved by technology alone and area advantage should be exploited by selecting network on chip components with more aggressive features. This in turn affects parallel programming.
UR - https://www.scopus.com/pages/publications/84860668548
U2 - 10.1109/ICM.2011.6177399
DO - 10.1109/ICM.2011.6177399
M3 - Conference contribution
AN - SCOPUS:84860668548
SN - 9781457722073
T3 - Proceedings of the International Conference on Microelectronics, ICM
BT - 2011 International Conference on Microelectronics, ICM 2011
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2011 23rd International Conference on Microelectronics, ICM 2011
Y2 - 19 December 2011 through 22 December 2011
ER -