To counter the pressure on global energy resources due to increasing computing needs – yes, AI, we’re looking at you – research institute imec is proposing a radical shift away from traditional computing methods.
The solution, detailed in IEEE spectrum engineering magazine, involves exploiting the fundamental properties of superconductors to significantly reduce power consumption, creating an innovative, superconducting processor.
This promising technology has been in development for a few years and uses standard CMOS manufacturing techniques that could potentially provide computing power that is a hundred times more energy efficient than today’s best chips, and could “lead to a computer that matches the amount of computing resources of a data center in a system the size of a shoebox.”
20 exaflops
Imec’s research involved designing a new kind of processor from the ground up, in close collaboration between CMOS engineers and full-stack development teams.
Imec switched from using niobium as a superconducting material to the related compound niobium-titanium nitride, because it can withstand temperatures used in CMOS fabrication without losing its superconducting properties.
The resulting superconducting chip, optimized for AI processors, resembles a typical 3D CMOS system-on-chip. One significant difference, however, is that most of the chip must be immersed in liquid helium to ensure optimal operating temperature, almost 4 Kelvin.
Compared to conventional CMOS chips, superconductors dissipate only a fraction of the energy in the form of heat. This feature makes it possible to stack computer chips directly on top of each other, reducing the physical footprint while maintaining the density gain due to Moore’s Law.
Imec estimates that a stack of 100 superconducting boards, each containing superconductor processing units (SPUs), superconducting SRAM and DRAM memory stacks and housed in a 20 x 20 x 12 centimeter cooling environment, can output as many as 20 exaflops in BF16 number format. This is twenty times as much as the most powerful supercomputer currently in existence (Frontier at Oak Ridge National Laboratory, Tennessee), which consumes a total of 500 kilowatts of power. It also offers 100x the level of energy efficiency.
Imec’s innovation is not intended to replace existing CMOS computing technologies, but rather to complement them. Researchers at the Institute believe the technology will boost AI and machine learning and integrate seamlessly with quantum computing.
The smaller data center footprint, enabled by superconducting digital technology, will allow these systems to be placed closer to their target applications, ultimately delivering huge advances in sectors such as agriculture, healthcare and energy.