Researchers at the University of Toronto will soon have at their disposal the most powerful supercomputer outside the United States—and apparently the most energy-efficient in the world.
According to the school, The IBM-built computer will be the 20th most powerful in the world, and it will be used to support research in climate-change prediction, aerospace, astrophysics, bioinformatics, chemical physics and medical imaging, among other areas.
Those behind the project say there is more to the supercomputer than its sheer power, however. They are also praising its energy efficiency. Chris Pratt, IBM Canada’s strategic initiatives executive, said the initiative maximizes both power and efficiency without sacrificing either.
“IBM started with the design of the data-centre efficiency and worked our way in. Whereas a lot of machines are designed from the machine out, this machine was designed to be energy efficient,” he said.
A 12,000-square-foot data centre that will house the computer will be built just north of Toronto, in the City of Vaughan. Its distance from the U of T is a product of the computer’s enormous thirst for energy. It will require an estimated two megawatts (or 2,000 kilowatts) of electricity in order to function. If it were to be built in downtown Toronto, the complex would have needed its own power substation. Its total energy consumption will roughly equal that of 1,500 homes.
Of course, it is impossible to determine exactly how much it will cost to run the supercomputer—or how much pollution that will generate. But an emissions calculator at the Toronto Hydro website could provide some insight.
If the machine runs for one hour and consumes 2,000 kilowatts, it expends 2,000 kilowatt-hours. According to the emissions calculator, 2,000 kilowatt-hours of consumption monthly translates into 5,688 kilograms of carbon-dioxide emissions every month—the equivalent of burning 2,275 litres of gas and driving over 18,500 kilometres (at a rate of 12 litres burned for every 100 km driven). It is also the equivalent of 20,000 lightbulbs.
And this is the world’s most energy efficient supercomputer.
The calculator also estimates the cost of 2,000 kilowatt-hours at $233.
Pratt focused not on the overall power consumption of the computer, but instead on the potential savings due to the company’s efforts to increase energy efficiency.
“What is significant is a machine of this capacity built out of prior-generation hardware would have required considerably more power to drive it,” he said. “The savings in using the new technology are in the order of 25-30 per cent.”
So using old technology, he said, the consumption could have reached 2.5 or even three megawatts.
Richard Peltier, the scientific director of SciNet consortium—a group of institutions and hospitals at the U of T that will provide a home for the computer—spoke about the importance of such a powerful computer to his research in climate science.
One of the major research initiatives undertaken by SciNet looks at climate-change prediction. The new supercomputer will make predictions more reliable because the increased power will allow researchers to test more complex climate-change models.
Peltier said less powerful computers can’t handle the load required of such complex models.
“The models that we use to do this are very complicated because you have to describe the entire atmosphere, the entire ocean, the sea ice which sits on top of the ocean, and the biosphere on the continents which interact with the overlying atmosphere,” he said, adding that predictions for years up to 2200 are based on conditions at the beginning of industrialization in 1860—a model spanning over three centuries.
Peltier said that high resolution models are necessary to make more precise predictions that apply to specific regions, such as the Arctic or the Great Lakes basin.
He said that the supercomputer will likely attract researchers from around the world.
“They are really poles of attraction in the research universe,” he said. “We expect that this is going to be a really useful recruiting tool.”
The $50-million project is funded by the Canadian Foundation for Innovation, Ontario’s provincial government, and the university. It is slated to be completed by next summer.