Algorithm Predicts CPU Power Consumption Incredibly Fast
Computer engineers at Duke University have developed a new artificial intelligence (AI) method that can accurately predict the power consumption of any type of computer processor. The most impressive feature of this new method is its ability to do this over a trillion times per second, while using very little computing power itself.
The new technique is called APOLLO and has been validated on high performance real world microprocessors. It could be applied in different ways, including to improve efficiency and inform the development of microprocessors.
the to research detailing this new method was published at MICRO-54: 54th Annual IEEE / ACM International Symposium on Microarchitecture.
Zhiyao Xie is the first author of the article and a doctoral student in the lab of Yiran Chen, professor of electrical and computer engineering at Duke.
âThis is an intensively studied problem that has traditionally relied on additional circuitry to solve,â said Xie. âBut our approach works directly on the background microprocessor, which opens up a lot of new opportunities. I think that’s why people are so excited about it.
Modern computer processors
Modern computer processors rely on computational cycles performed 3 trillion times per second. In order to maintain the performance and efficiency of the chip, the power consumption of these rapid transitions must be tracked. When a processor consumes too much power, it can lead to overheating and damage. When the power changes rapidly, it can lead to internal electromagnetic complications which results in a slower processor.
Software can predict and stop these extremes, and computer engineers can use it to protect hardware and increase performance. However, this process requires additional hardware and computing power.
âAPOLLO approaches an ideal power estimation algorithm that is both precise and fast and can easily be integrated into a processing core at a low energy cost,â said Xie. “And because it can be used in any type of processing unit, it could become a common component in future chip design.”
Rely on AI
APOLLO relies on artificial intelligence for its power. The algorithm relies on AI to identify and select 100 of a processor’s millions of signals, which correlate with its power consumption. These 100 signals are then used to build a model of power consumption, and the algorithm monitors them to predict chip performance in real time.
This process is self-contained and data-driven, which means it can be implemented on almost any type of computer processor architecture.
âOnce the AI ââhas selected its 100 signals, you can look at the algorithm and see what they are,â Xie said. âA lot of the selections make intuitive sense, but even if they don’t, they can provide feedback to designers, informing them of the processes most strongly correlated with power consumption and performance. “
The researchers say the algorithm still needs further testing and comprehensive evaluations on different platforms. This is necessary before it can be adopted by commercial computer manufacturers.