Bringing AI to Hard Disk Drives

Disk drives are the main workhorse in today’s technology. Everything from facial recognition to the Alexa in your house rely on information stored in the cloud, which is supported by disk drives. To keep up with the exponential rate of data created every day, disk drives must evolve by reading and recovering data more efficiently.

Jimmy Zhu, ABB Professor of Electrical and Computer Engineering, and his students have developed the first machine learning-based data detection channel chip for hard disk drives. The team presented their work at the VLSI Symposium 2022 last June in Hawaii.

“We are integrating machine learning into hard disk drives to make them faster and more efficient,” says Zhu. “This is the first chip to use machine learning for data detection in disk drives. The performance of the machine learning-based chips surpasses the regular chips manufactured today.”

In present hard disk drives, data detection from data bits recorded in the magnetic layer on a disk surface uses conventional signal processing technology. By using machine learning, the process can be better optimized with significant performance improvement which will translate to an increased capacity for storing more data.

The Department of Electrical and Computer Engineering at Carnegie Mellon University is in a unique position to develop such an innovative chip. By offering courses like Advanced Digital Integrated Circuit Design, students can design computer chips and have them fabricated at the end of the semester.

“My students take this class to render the neural network chip,” says Zhu. “We design the chip throughout the year and can test a physical prototype at the end of the semester. This is a unique experience offered at Carnegie Mellon.”

Industry partners agree. Zhu states that disk drive companies are surprised that a university can handle this level of development and fabrication.

“We have taken the lead in machine learning technology,” says Zhu. “This research will impact any real time data detection which will change hard disk drives and the future of technology as we know it.”

This pioneering research work has been funded by the Data Storage Systems Center and its industrial sponsors. It is also part of the collaboration between Apple Computer and the Department of Electrical and Computer Engineering at Carnegie Mellon University.