Vasily G. Moshnyaga
ABSTRACT
Due to a large capacitance and enormous access rate, caches dissipate about a third of the total energy consumed by today's processors. In this paper we present a new architectural technique to reduce energy consumption in caches. Unlike previous approaches, which have focused on lowering cache capacitance and the number of accesses, our method exploits a new freedom in cache design, namely the voltage per access. Since in modern caches, the loading capacitance operated on cache-hit is much less than the capacitance operated on cache-miss, the given clock cycle time is inefficiently exploited during the hit. We propose to trade-off this unused time with the supply voltage, lowering the voltage level on the hit and increasing it during the miss. Experiments shows that the approach can save up to 60% of cache energy without large performance and area overhead.
- 1.J. Edmodson, et al, "Internal Organization of the Alpha 21164, a 300-MHz 64-bit Quad-issue CMOS RISC Microprocessor", Digital Technical Journal, Vol.7, no.1, pp.119-135, 1995. Google Scholar
Digital Library
- 2.J. Montanario, et al, "A 160 MHz 32b 0.5W CMOS RISC Microprocessor, IEEE ISSCC, Digest of Technical Papers, 1996.Google Scholar
- 3.K. Itoh, Low Power Memory Design, in Low Power Design Methodologies, Ed. by J. Rabaey and M. Pedram, Kluwer AP, pp. 201-251, 1996.Google Scholar
- 4.J. Bunda, W. Athas and D. Fussel, "Evaluating Power Implications of CMOS Microprocessor Design Decisions", Proc. 1994 ISLPED, pp.147-152, 1994.Google Scholar
- 5.J. Kin, M. Gupta, and W. Mangione-Smith, "The Filter Cache: An Energy-Efficient Memory Structure", Proc.MICRO-30, pp.184-193, 1997. Google ScholarDigital Library
- 6.U. Ko, P. Balsara, and A. Nanda, "Energy Optimization of Multi-level Processor Cache Architecture" Proc. 1995 ISLPED, pp.45-49, 1995. Google ScholarDigital Library
- 7.C. Su, and A. Despain, "Cache Design Trade-offs for Power and Performance Optimization: A Case Study", Proc. 1995 ISLPED, pp.63-68, 1995. Google ScholarDigital Library
- 8.Y. Shimasaki, et al, "An 8-mW, 8-kB Cache Memory Using an Automatic Power-Save Architecture for Lower Power RISC Microprocessors", IEICE Trans. Electron., Vol.E79-C, no.12, pp.1693-1697, Dec.1996.Google Scholar
- 9.K. Ghose and M. Kamble, "Reducing Power in Superscalar Processor Caches Using Subbanking, Multiple Line Buffers and Bit-Segmentation", Proc. 1999 ISLPED, pp.70-75, 1999. Google ScholarDigital Library
- 10.N. Bellas, et al, "Using Dynamic Cache Management Techniques to Reduce Energy in a High- Performance Processor", Proc. 1999 ISLPED, pp.70-75, 1999. Google ScholarDigital Library
- 11.K. Inoue, et al, "Way Predicting Set-Associative Cache for High Performance and Low Energy Consumption", Proc. 1999 ISLPED, pp.273-275, 1999. Google ScholarDigital Library
- 12.A. P. Chandrakasan, R. W. Brodersen, it "Low Power Digital CMOS Design", Kluwer Ac., 1996 Google ScholarDigital Library
- 13.A. Dancy, A.P. Chandrakasan, "Techniques for Aggressive Supply Voltage Scaling and Efficient Regulation", IEEE CICC, pp.579-587, 1997.Google Scholar
- 14.K. Suzuki, et al, "A 300MIPS/W RISC Core Processor with Variable Supply-Voltage Scheme in Variable Threshold- Voltage CMOS", IEEE CICC, pp.579-587, 1997.Google Scholar
- 15.D. A. Pederson and J. L. Hennesy, "Computer Architecture: a Quantitative Approach", Morgan Kaufmann Publishers, Inc., 1997 Google ScholarDigital Library
- 16.M. D. Hill, J. R. Larus, A. R. Lebeck, M. Talluri, and D. A. Wood, "WARTS: Wisconsin Architectural Research Tool Set," Computer Architecture News, Vol. 21 no.4, pp. 8-10, 1993. http://www.cs.wisc.edu/ larus/warts.html Google ScholarDigital Library
- 17.S. Wilton, N. Jouppi, "An enhanced Access and Cycle Time Model for On-Chip Caches",Research Report 93/5, Digital, June 1994.Google Scholar
Index Terms
- Reducing cache engery through dual voltage supply
Recommendations
Reducing cache misses through programmable decoders
Level-one caches normally reside on a processor's critical path, which determines clock frequency. Therefore, fast access to level-one cache is important. Direct-mapped caches exhibit faster access time, but poor hit rates, compared with same sized set-...
Comments