New $28M center led by U-M will develop computers of 2025
Designing the computers of 2025 is the focus of a new $28 million, five-year research center led by U-M.
The Center for Future Architectures Research (C-FAR) opened Thursday, and involves 14 other major research institutions. It aims to harness the power and boost the reliability of the tiniest transistors that will emerge over the next decade.
Transistors are the fundamental building blocks of modern electronics, and more than a billion of them make up each integrated circuit in today’s cell phones and personal computers. C-FAR will support the design of the next generation of computers that will enable applications such as computer vision, speech recognition, enhanced graphics, and “big data” analysis.
C-FAR is one of six new centers announced Thursday by the Semiconductor Research Corp., a non-profit collaboration between government and industry with support from the Defense Advanced Research Projects Agency and major firms. In addition to leadership of C-FAR, U-M researchers are involved in three of the other five new Semiconductor Technology Advanced Research network (STARnet) centers.
Together, the STARnet centers are charged with guiding the field through the approaching sunset of Moore's law, named for the Intel co-founder who in 1965 identified a trend that has fueled the semiconductor industry for the past 50 years.
Moore's law observes that the industry is able to shrink transistors enough to double the number that fit on an integrated circuit, or chip, every 18 months. With this increased density comes a doubling of computing power, which has transformed the room-sized computers of the '50s to the significantly more powerful smartphones of today. But the field is reaching the limit to miniaturization.
"It's a challenging time as we approach the end of Moore's law — not tomorrow, but soon," said Todd Austin, professor of electrical engineering and computer science and C-FAR director. "The dimensions of the transistors of today are in the tens of atoms. We can still make them smaller, but not without challenges that threaten the progress of the computing industry."
In addition to Austin, other U-M co-investigators in C-FAR — all in the Department of Electrical Engineering and Computer Science — are Scott Mahlke, professor; Valeria Bertacco, associate professor; and Kevin Fu, associate professor.
Researchers predict about another decade of squeezing silicon, and that will bring challenges along with opportunities. As nanoscale transistors shrink, they become less reliable. And in some cases, they offer more computational power than other parts of the system are capable of using.
"Industry is adding more and more cores. Last year, your cell phone had two, next year it will have four. The problem is your cell phone doesn't do much more than it did before because harnessing all those cores is a challenge," Bertacco said.
The researchers at C-FAR will work to overcome these challenges on three fronts. They will design specialized chips tailored for different applications. They'll rethink how to build computers so they're capable of analyzing massive data sets more efficiently, which will be a key application in the next decade. And they'll explore how to integrate tomorrow's technologies into conventional silicon processing techniques to ease the transition for industry. These advanced technologies include three-dimensional stacked chips and phase-change memory, which uses heat to store information in the molecules of a glass-like material.
"If we can build better systems with these smaller and less reliable transistors, we have an opportunity to make the next generation of faster, more energy-efficient computers. Imagine your tablet of tomorrow running high-end applications, like real-time medical imaging or computer vision, without needing a recharge for days," Bertacco said.
While the center is headquartered at U-M, it includes researchers from top universities across the nation.
"The center provides a unique environment to collaborate closely with other researchers and our sponsors to create rule-breaking technologies that are both effective and relevant," Austin said.
"If we can't continue to improve the speed and energy efficiency of tomorrow's computers, the computing industry will stagnate," Bertacco concluded. "It's important that we find ways to move the industry forward, even in the face of all these great challenges."
Other universities in C-FAR are Columbia; Duke University; Georgia Institute of Technology; Harvard University; Massachusetts Institute of Technology; Northeastern University; Princeton University; Stanford University; the University of California, Berkeley; the University of California, Los Angeles; the University of California, San Diego; University of Illinois at Urbana-Champaign; the University of Virginia; and the University of Washington, Seattle.
The other STARnet centers that U-M researchers are involved in are:
• The TerraSwarm Research Center, led by the University of California, Berkeley. TerraSwarm will address the potential and risks of pervasively integrating smart, networked sensors and actuators into our connected world. While such integration could have benefits such as better traffic control, energy efficiency and emergency response, those must be balanced with safety and privacy concerns. Those involved from the U-M Department of Electrical Engineering and Computer Science are Prabal Dutta, assistant professor; David Blaauw, professor; and Kevin Fu, associate professor.
• The Systems on Nanoscale Information fabriCs (SONIC) is led by the University of Illinois at Urbana-Champaign. SONIC aims to develop next-generation computing platforms with advanced circuit and device technology inspired by the robustness and energy efficiency of communication devices such as cell phones and biological systems such as the brain. Blaauw is a principal investigator.
• The Center for Spintronic Materials, Interfaces and Novel Architectures (C-SPIN), led by the University of Minnesota. C-SPIN will develop new material systems and transistor types that are based on the spin state of electrons, a quantum mechanical phenomenon. These could eventually complement or even replace traditional transistors. Dennis Sylvester, professor of electrical engineering and computer science, is an investigator.