TeslaK20 Eurora Supercomputer achieves world record for energy efficiency with Nvidia Tesla GPUs

Italy’s newest and fastest supercomputer, nicknamed Eurora, set the world record this week for energy efficiency, maintaining 3150 megaflops per watt of performance.

Eurora, a Supercomputer based at the Cineca facility in Bologna, Italy has recently broken the record for highest performance-to-energy-efficiency by 26 percent over the previous highest system on the Green500 list of the world’s most efficient supercomputers, the National Institutes for Computational Sciences’ “Beacon”. Powered by 128 of NVIDIA’s Tesla K20 GPUs and Intel Xeon 'Sandy-Bridge EP' E5-2687W, the Eurotech-built system was able to sustain 3150 megaflops per watt of performance, and a total of 350 teraflops in one rack.

EuroraSystemCinecaFacilityItaly Eurora Supercomputer achieves world record for energy efficiency with Nvidia Tesla GPUs
Eurora, deployed at the Cineca Facility in Bologna, Italy

The system is cooled using Aurora Hot Water Cooling technology, a process originally developed by IBM that passes warm water up to 45 degrees Celsius (113° F) directly to processors and memory. This cooling system is significantly more compact and energy efficient than a comparable air-cooling system. 

AuroraHWCT Eurora Supercomputer achieves world record for energy efficiency with Nvidia Tesla GPUs
Eurotech's Aurora Hot Water Cooling

The Aurora Cooling System, coupled with Tesla GPU accelerators based on NVIDIA Kepler, is proving to be an effective team of power saving, reducing total cost of ownership by between 30-50 percent. "GPU accelerators are inherently more energy efficient than CPUs, and Tesla K20 accelerators widen this gap considerably," said Sumit Gupta, general manger of the Tesla accelerated computing business at NVIDIA. "Energy efficiency has become the defining element of computing performance."

As the trend of using GPUs over traditional CPUs continues and cooling methods become more innovative, we should see a huge boost in both performance and energy efficiency in future super-systems and smaller devices alike.

Source: NVIDIA