Dynamic Power and Thermal Management in Multicore Systems

Modern multicore processors deliver exceptional computational throughput by integrating dozens or even hundreds of cores on a single chip. As core counts and operating frequencies have risen, power density and heat dissipation have become critical constraints on performance, reliability and energy efficiency. Dynamic power and thermal management seeks to balance workload distribution, voltage and frequency scaling, task migration and active cooling to maintain safe operating temperatures while maximising performance per watt. Techniques such as dynamic voltage and frequency scaling (DVFS) adjust supply voltage and clock rate in real time to reduce power draw under lighter workloads, whereas dynamic thermal management (DTM) relies on thermal sensors and redistribution of tasks to cooler cores to avoid hotspots. Advanced schemes may combine predictive models, machine-learning inference and hardware feedback loops to anticipate temperature trends and manage “dark silicon” regions where portions of the chip are deliberately powered down to contain thermal budget. The global significance of this research spans from energy-aware exascale datacentres to battery-powered mobile devices, delivering improved system longevity, lower operational cost and reduced environmental impact.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Dynamic Power and Thermal Management in Multicore Systems publication trend

The graph below shows the total number of articles in dynamic power and thermal management in multicore systems across all publications each year (not limited to Nature Index journals).

Technical terms

Dynamic voltage and frequency scaling (DVFS): A technique that adjusts processor voltage and clock rate at run time to reduce power consumption under varying workload demands.

Dynamic thermal management (DTM): Methods that monitor on-chip temperature and redistribute or throttle workloads to prevent thermal violations and hotspots.

Dark silicon: Portions of a multicore chip that are intentionally powered down or left idle to stay within thermal and power limits.

Task migration: The process of moving computational threads or processes from one core to another to balance temperature, workload and energy consumption.

Many-core systems: Processors integrating a large number of cores (beyond traditional multi-core counts) on a single die, offering massive parallelism but posing significant thermal and power challenges.

References

1. 3D-DNaPE: Dynamic Neighbor-Aware Performance Enhancement for Thermally Constrained 3D Many-Core Systems. IEEE Access (2023).
2. Learning-Directed Dynamic Voltage and Frequency Scaling Scheme with Adjustable Performance for Single-Core and Multi-Core Embedded and Mobile Systems †. Sensors (2018).
3. DTaPO: Dynamic Thermal-Aware Performance Optimization for Dark Silicon Many-Core Systems. Electronics (2020).