What are the main challenges and trade-offs of using dynamic voltage and frequency scaling (DVFS) in SoCs?

2 Limitations of DVFS

However, DVFS also has some drawbacks and limitations that affect the SoC performance and functionality. One of the main challenges is to balance the trade-off between power saving and quality of service (QoS), which refers to the level of satisfaction and performance that the SoC delivers to the user or application. For example, reducing the voltage and frequency of a SoC may result in lower processing speed, higher latency, or lower accuracy, which can compromise the QoS. Therefore, DVFS needs to consider the QoS requirements and constraints of the SoC and its workload, and adjust the voltage and frequency accordingly.

3 Methods of DVFS

Different methods and techniques can be used to implement DVFS in SoCs, depending on the level of granularity, control, and complexity. Global DVFS applies the same voltage and frequency to the entire SoC based on overall workload and performance demand, making it simple and easy to implement. However, it may not be optimal or efficient for heterogeneous or dynamic SoCs. Local DVFS allows different components or regions of the SoC to have different voltage and frequency levels according to their individual workload and performance demand, but it requires more complex circuitry and control mechanisms. Adaptive DVFS dynamically adjusts the voltage and frequency of the SoC or its components or regions based on real-time feedback from sensors, monitors, or algorithms. This method is more responsive and intelligent, yet it requires more sophisticated hardware and software solutions such as power sensors, temperature sensors, performance counters, machine learning models, and optimization algorithms. It also involves sensing, processing, and communication costs.

4 Metrics of DVFS

To measure the power consumption, performance, and QoS of SoCs or their components or regions under different voltage and frequency levels, several metrics and indicators are needed to evaluate the effectiveness and efficiency of DVFS. These include power, which is the energy consumed per unit time, and can be divided into static (idle/standby mode) and dynamic (active/operation mode) power. Performance is another metric that measures the speed or throughput of the SoC or its components or regions; it can be calculated by dividing operations or instructions by the time taken to complete them, and expressed as frequency or clock rate. Lastly, QoS measures the level of satisfaction or performance that the SoC or its components or regions deliver to the user or application; parameters for this metric vary depending on the type and nature of the task or workload.

5 Future trends of DVFS

As SoCs become increasingly complex, heterogeneous, and dynamic, DVFS will continue to be an important factor in optimizing power consumption and performance. However, DVFS will also need to face new challenges and opportunities in the future, such as multicore and manycore SoCs which require more advanced coordination for parallelism and diversity management; near-threshold and sub-threshold SoCs which require more robust and adaptive techniques to cope with low-voltage operation; and neuromorphic and quantum SoCs which require new and innovative techniques to understand their unique characteristics. All of these emerging technologies require DVFS techniques that can ensure the QoS of the SoCs while enabling their scalability and feasibility.

6 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?