News & Analysis
Comment
jeremybirch
I suppose he is not the other Chuck Moore, inventor of FORTH, who is still going ...
KB3001
I was implying ARM vs. x86, not ARM vs. GPUs.
GPUs, low-power pave path to exascale supercomputing
Sylvie Barak
11/3/2011 12:58 AM EDT
SAN FRANCISCO--The supercomputing race is accelerating, but the path to exascale computing faces serious challenges in terms of power efficiency, cost and data security.
Speaking at a High Powered Computing (HPC) roundtable with AMD, Appro, Cray, Penguin Computing and Supermicro on Wednesday (Nov. 2), Carl Claunch, a vice president and analyst at Gartner Research, called the supercomputing space a “continually moving target.”
Governments, corporations and the academic world continued to display an ever increasing appetite for system speed said Claunch, seeking to address increasingly large problem sets ranging from climate change to tactical nuclear defense and leaps forward in biomedicine.
Getting to exascale computing and quintillion calculations per second is therefore no longer science fiction, but a concrete goal—one the industry hopes to achieve by 2018, Claunch said.
Indeed, with supercomputers and HPC becoming ever more mainstream and relevant across a plethora of industries from nuclear physics to climate modeling to banking, the questions of efficiency, density and cost are becoming particularly pertinent, the panel agreed.
“Most supercomputer buyers are being constrained more by budget than by anything else,” said Claunch, noting that despite the great strides in computer modeling and the great elasticity in HPC, the drive for faster, smaller, cheaper systems remains strong.
Moore’s Law has paid off in terms of transistor density improvements, the panel agreed, but more work needs to be done in terms of power per watt, especially as more power translates into more FLOPS of compute speed.
Power budgets for supercomputers are rising, but ultimately a machine needs to be affordable enough to run in order to be effective.
The U.S. government, for instance, has plowed increasing amounts of money into the space, with a reported $126 million set aside for exascale computing for 2012 alone, to help ward off competition for supercomputer supremacy. China and Japan in particular have been making aggressive moves in the space.
“Supercomputing is an area we want to maintain U.S. leadership in, it’s important we stay in front,” maintained Margaret Williams, Cray’s senior vice president of HPC systems. Williams said Cray is frequently in contact with government officials and was lobbying hard for the U.S. department of energy to spend more money in the supercomputing space. “There’s a lot of investment needed,” she emphasized.
While some on the panel disagreed that the U.S. needed to invest simply to retain “bragging rights,” others noted that big investments in HPC eventually benefited the entire industry, with Don Clegg, vice president of marketing and business development at Supermicro, arguing that the financing eventually trickled down. “Today’s cutting edge is tomorrow’s mainstream system,” he said.
Speaking at a High Powered Computing (HPC) roundtable with AMD, Appro, Cray, Penguin Computing and Supermicro on Wednesday (Nov. 2), Carl Claunch, a vice president and analyst at Gartner Research, called the supercomputing space a “continually moving target.”
Governments, corporations and the academic world continued to display an ever increasing appetite for system speed said Claunch, seeking to address increasingly large problem sets ranging from climate change to tactical nuclear defense and leaps forward in biomedicine.
Getting to exascale computing and quintillion calculations per second is therefore no longer science fiction, but a concrete goal—one the industry hopes to achieve by 2018, Claunch said.
Indeed, with supercomputers and HPC becoming ever more mainstream and relevant across a plethora of industries from nuclear physics to climate modeling to banking, the questions of efficiency, density and cost are becoming particularly pertinent, the panel agreed.
“Most supercomputer buyers are being constrained more by budget than by anything else,” said Claunch, noting that despite the great strides in computer modeling and the great elasticity in HPC, the drive for faster, smaller, cheaper systems remains strong.
Moore’s Law has paid off in terms of transistor density improvements, the panel agreed, but more work needs to be done in terms of power per watt, especially as more power translates into more FLOPS of compute speed.
Power budgets for supercomputers are rising, but ultimately a machine needs to be affordable enough to run in order to be effective.
The U.S. government, for instance, has plowed increasing amounts of money into the space, with a reported $126 million set aside for exascale computing for 2012 alone, to help ward off competition for supercomputer supremacy. China and Japan in particular have been making aggressive moves in the space.
“Supercomputing is an area we want to maintain U.S. leadership in, it’s important we stay in front,” maintained Margaret Williams, Cray’s senior vice president of HPC systems. Williams said Cray is frequently in contact with government officials and was lobbying hard for the U.S. department of energy to spend more money in the supercomputing space. “There’s a lot of investment needed,” she emphasized.While some on the panel disagreed that the U.S. needed to invest simply to retain “bragging rights,” others noted that big investments in HPC eventually benefited the entire industry, with Don Clegg, vice president of marketing and business development at Supermicro, arguing that the financing eventually trickled down. “Today’s cutting edge is tomorrow’s mainstream system,” he said.
|
|
|
|
|
Navigate to related information
KB3001
11/3/2011 4:05 PM EDT
Performance per Watt is indeed the killer in HPC. I am not sure I agree ARM have no future in HPC. IMO, the window of opportunity has never been larger for them. It'll take time, yes, but I think it will happen.
Sign in to Reply
PJames
11/3/2011 6:15 PM EDT
GPUs are likely even more power efficient than ARM CPUs for many applications.
Sign in to Reply
KB3001
11/3/2011 6:29 PM EDT
I was implying ARM vs. x86, not ARM vs. GPUs.
Sign in to Reply
pixies
11/3/2011 4:06 PM EDT
Wonder if Chuck Moore is related to Gordon Moore.
Sign in to Reply
jeremybirch
11/8/2011 8:54 AM EST
I suppose he is not the other Chuck Moore, inventor of FORTH, who is still going strong into his 70's with a 144 compute core chip (see greenarraychips.com)
Sign in to Reply