New Computer Chip to Use less Power and Enhance Computing Speed

New Computer Chip to Use less Power and Enhance Computing Speed


Researchers at Princeton University have designed a new form of computer chip that improves the performance and reduces power consumption of artificial-intelligence systems. This computer chip, coupled with some programming languages, could be majorly beneficial for gadgets such as smartphones, watches and other devices that consume more battery power. A technique known as in-memory computing is developed to overcome an analytical issue that compels processors to spend more energy and time in drawing data from stored memory. This technique enhances the speed and efficiency of the processors by processing data directly in the storage.

The University in collaboration with the U.S.-based semiconductor company Analog Devices had already successfully constructed the circuitry for in-memory computing. After performing lab tests, it was found that this new chip would function hundred times better than the other chips. However, at the same time, it was noted that the performance was limited due to the absence of a few components included in the new chip. The researchers also revealed that now that the new chip can work with standard programming languages such as C.

Performance Boost and Energy Savings by Chip to support Deep Learning Systems

The Princeton researchers designed the chip to support algorithms that enables a computer to perform complicated tasks and eventually help in making decisions by analyzing different data sets, especially in artificial intelligence (AI) systems. Furthermore, one of the researchers opined that it is important to save chip’s energy and improve performance as many applications in AI are likely to function on battery-powered devices, such as medical sensors or smartphones. He further said that power consumption and performance improvements can only be useful if they open to wide range of applications. This is where the need for programmability comes into picture.

In-memory computing has been able to solve the issues related to speed and energy, explained the Princeton researcher. However, the question that lies unanswered is if this in-memory computing system could be used in all I applications and thus will make programmability a prerequisite for system designers.