Home |
Menu
Energy disssipation in Data buses due to Bus Capacitance
Click here for result summary
Click here for the source code
ABSTRACT:
The power dissipation associated with driving data buses depends on both Substrate and Inter-wire capacitances. Lot of bus encoding schemes have been proposed to reduce the power dissipation in these buses. Most of the works foucessed on minimizing the transitions. A previous work also considered the Inter-wire capacitance and proposed a coding scheme to reduce power dissipation. In this work I have considered both the Substrate, Inter-wire capacitances and calculated energy savings associated with each of the popular bus coding techniques. For this purpose run-time memory traces from Mediabench benchmark set have been extraced and used.
RESULTS:
To evaluate the energy savings in each of the coding schemes software was developed to calculate energy consumed during data transmission in various coding schemes for these benchmarks. In this calculation energy overhead of coding and decoding logic is not considered. Figures 1, 2, 3 and 4 show the energy consumption in data buses for these benchmarks for different capacitance ratios. This ratio plays an important part of energy consumption associated with the interaction of adjacent lines and this value is the ratio of Interwire to Substrate capacitance which is purely based on the technology parameters. Energy savings for each coding schemes relative to the standard twos complement coding are calculated and the results for Capacitance ratio = 0 (ignoring Interwire capacitance) are compared with the similar results produced in one of the previous works and the results concur with each other. To check Adaptive coding technique, a synthetic benchmark has been used and the energy savings are compared with the manual calculation.
CONCLUSION:
In this work some of the low power coding techniques have been considered and percentage energy savings of these schemes with respect to twos complement coding are calculated. It is based on an model for the data bus that explicitly considers the inter-wire capacitance. It was observed that percentage energy savings for these coding techniques decrease as the capacitance ratio increases. The coding schemes should be modified to incorporate effect of Interwire capacitance also to reduce overall power consumption so that decrease in energy savings due to increase in capacitance ratio is less. Adaptive technique shows better energy savings than all other coding techniques. Even for a LUT of size 8 it gives good energy savings comparing with traditional approaches. For a LUT of size 128 the increase in energy savings in all the benchmarks is very high. It can be concluded that either the coding scheme in [1](Refer report) or the Adaptive coding scheme in [2](Refer Report) can be used to decrease energy consumption where inter-wire capacitance is non-negligible comparing to Substrate capacitance.