UC Davis KiloCore CPU Packs 1,000 Cores, Sips 0.7 Watts Of Power And Runs On Single AA Battery
A bunch of researchers at the Department of Electrical and Computer Engineering at the University of California, Davis (UC Davis) have built a "many-core" processor unlike any other. Aptly named KiloCore, the new CPU contains a staggering 1,000 independent programmable processors, which is presumably the first of its kind.
"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.
Baas led the team that designed the chip architecture. While there have been other many-core CPUs, there haven't been any with more than 300 processors, the research team said. Most of those chips were created for research and a few of them ended up being sold commercially.
Image Source: UC Davis
IBM fabricated the KiloCore using its 32nm CMOS technology, a comparatively old process now that companies are pumping out 16nm and 14nm FinFET parts. It has 621 million transistors and a maximum computation rate of 1.78 trillion instructions per second. Each processor core is able to run its own small program independently of the others, giving it more flexibility than the Single-Instruction-Multiple-Data approach that's utilized by GPUs. Baas says the idea is to break up an application into tiny chunks and then run them in parallel on different processors. Doing so allows for high throughput with low energy consumption.The cores operate at a maximum clockspeed of 1.78GHz and transfer data directly to each other rather than using a pooled memory area that can create a bottleneck.
Since each processor is independently clocked, each can shut down when not needed to further save energy. According to Baas, it's the most energy-efficient many-core processor on the planet with the ability to execute 115 billion instructions per second while dissipating just 0.7 watts. That's low enough to be powered by a solo AA battery and is 100 times more efficient than a modern laptop.
There are already a handful of applications that have been developed for the chip, including wireless coding and decoding, video processing, encryption, and others that utilize large amounts of parallel data.
"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.
Baas led the team that designed the chip architecture. While there have been other many-core CPUs, there haven't been any with more than 300 processors, the research team said. Most of those chips were created for research and a few of them ended up being sold commercially.
Image Source: UC Davis
IBM fabricated the KiloCore using its 32nm CMOS technology, a comparatively old process now that companies are pumping out 16nm and 14nm FinFET parts. It has 621 million transistors and a maximum computation rate of 1.78 trillion instructions per second. Each processor core is able to run its own small program independently of the others, giving it more flexibility than the Single-Instruction-Multiple-Data approach that's utilized by GPUs. Baas says the idea is to break up an application into tiny chunks and then run them in parallel on different processors. Doing so allows for high throughput with low energy consumption.The cores operate at a maximum clockspeed of 1.78GHz and transfer data directly to each other rather than using a pooled memory area that can create a bottleneck.
Since each processor is independently clocked, each can shut down when not needed to further save energy. According to Baas, it's the most energy-efficient many-core processor on the planet with the ability to execute 115 billion instructions per second while dissipating just 0.7 watts. That's low enough to be powered by a solo AA battery and is 100 times more efficient than a modern laptop.
There are already a handful of applications that have been developed for the chip, including wireless coding and decoding, video processing, encryption, and others that utilize large amounts of parallel data.
