A chip with 1,000 processors; topological plexcitons; efficient thermoelectrics.
A chip with 1,000 processors
A microchip containing 1,000 independent programmable processors has been designed by a team at the University of California, Davis. Called the KiloCore chip, it contains 621 million transistors and was fabricated by IBM using its 32nm CMOS technology. Cores operate at an average maximum clock frequency of 1.78 GHz, and they transfer data directly to each other rather than using a pooled memory area.
“To the best of our knowledge, it is the world’s first 1,000-processor chip and it is the highest clock-rate processor ever designed in a university,” said Bevan Baas, professor of electrical and computer engineering at UC Davis.
Each processor core can run its own small program independently of the others in a multiple instruction, multiple data (MIMD) approach which the team says is more flexible than single instruction, multiple data (SMID) approaches utilized by processors such as GPUs; the idea is to break an application up into many small pieces, each of which can run in parallel on different processors, enabling high throughput with lower energy use, Baas said.
Because each processor is independently clocked, it can shut itself down to further save energy when not needed, said graduate student Brent Bohnenstiehl, who developed the principal architecture. Cores operate at an average maximum clock frequency of 1.78 GHz, and they transfer data directly to each other rather than using a pooled memory area.
The chip is the most energy-efficient “many-core” processor ever reported, according to the team, with the 1,000 processors able to execute 115 billion instructions per second while dissipating only 0.7 Watts, low enough to be powered by a single AA battery.
Applications include wireless coding/decoding, video processing, encryption, and others involving large amounts of parallel data. The team has completed a compiler and automatic program mapping tools for use in programming the chip.
Scientists at UC San Diego, MIT and Harvard University engineered “topological plexcitons,” energy-carrying particles that could help make possible the design of new kinds of solar cells and miniaturized optical circuitry.
“When light and matter interact, they exchange energy,” explained Joel Yuen-Zhou, an assistant professor of chemistry and biochemistry at UC San Diego. “Energy can flow back and forth between light in a metal (so called plasmon) and light in a molecule (so called exciton). When this exchange is much faster than their respective decay rates, their individual identities are lost, and it is more accurate to think about them as hybrid particles; excitons and plasmons marry to form plexcitons.”
Materials scientists have been looking for ways to enhance a process known as exciton energy transfer, or EET, to create better solar cells as well as miniaturized photonic circuits which are dozens of times smaller than their silicon counterparts.
The drawback with EET, however, is that this form of energy transfer is extremely short-ranged, on the scale of only 10 nanometers, and quickly dissipates as the excitons interact with different molecules.
One solution to avoid those shortcomings is to hybridize excitons in a molecular crystal with the collective excitations within metals to produce plexcitons, which are capable of travelling much farther. But the main problem with plexcitons is that their movement along all directions, which makes it hard to properly harness in a material or device.
The team’s solution to that problem by engineering particles called “topological plexcitons,” based on the concepts of topological insulators. According to Yuen-Zhou, “the exciting feature of topological insulators is that even when the material is imperfect and has impurities, there is a large threshold of operation where electrons that start travelling along one direction cannot bounce back, making electron transport robust. In other words, one may think about the electrons being blind to impurities.”
Adding this “topological” feature to plexcitons gives rise to directionality of EET, opening the door to plexcitonic switches which distribute energy selectively across different components of a new kind of solar cell or light-harvesting device.
By doping a thermoelectric material with minute amounts of sulfur, a team of researchers from Rensselaer Polytechnic Institute, University of Missouri, and the Max Planck Institute for Solid State Research found a new path to large improvements in the efficiency of materials for solid-state heating and cooling and waste energy recapture. This approach profoundly alters the electronic band structure of the material – bismuth telluride selenide — improving the figure of merit, a ranking of a material’s performance that determines efficiency in applications and opening the door to advanced applications of thermoelectric materials to harvest waste heat from computer chips all the way up to power plants.
Thermoelectric materials can convert a voltage to a thermal gradient – causing one side of a material to become hot or cold — and vice-versa. State-of-the-art thermoelectric materials are not very efficient, limiting their use to niche applications such as picnic refrigerators, domestic water heaters, car-seat climate control and night vision goggles.
The big challenge in generating power with thermoelectrics is how to get high voltage and low resistance at the same time.
“This is an exciting breakthrough because this allows us to untangle two unfavorably coupled properties that limit thermoelectric performance,” said Ganpati Ramanath, professor of materials science and engineering at Rensselaer. “Moreover, our approach works for both nanocrystals as well as bulk materials.”
“Seventy percent of all energy loss is heat. If we can generate even 5% more electricity from that waste heat, we’ll be on our way to making a big impact on power production and carbon dioxide emissions reduction,” said Theo Borca-Tasciuc, Rensselaer professor of mechanical engineering.