gary
21-07-2012, 06:14 PM
One of the engineering challenges of the SKA is that its computing requirements
exceed that which is available today.
Part of the jigsaw puzzle is about to fall in place with the Australian
government announcing yesterday the procurement of a 1.2 Petaflop Cray Cascade (http://www.cray.com/Programs/Cascade.aspx)
supercomputer which will be installed at the Pawsey Center in Perth.
A Peta is 1 x 10 to the power of 15.
Cray describe the machine as their "Next Generation Supercomputer (http://www.cray.com/Programs/Cascade.aspx)".
iVEC is an organization devoted to providing supercomuting capacity to
Western Australian researchers and is a joint venture between CSIRO and
WA's four public universities. iVEC and CSIRO manage the Pawsey
supercomputer center.
In a press release (http://www.ivec.org/node/532), iVEC announced -
Also purchased were components from SGI and Oracle and the system will be interfaced via dark fiber cables leased
from the National Broadband Network which run to the correlators at the Murchison antenna site.
The Cascade is Cray's first supercomputer based on next generation Intel Xenon processors.
Only last month, on 18th June, Intel announced its new Xenon Phi co-processor. Cray simultaneously
announced that the Cascade would include the Xenon Phi along with the Xeon CPU in its architecture.
The Xenon Phi took five years to design and is currently only in alpha-test pre-production but can provide teraflop performance on
a single chip. It employs 22nm fabrication technology and 3D Tri-gate transistors. Each chip has 50 computing
cores, what Intel refer to a their Many Integrated Core Architecture or MIC (http://en.wikipedia.org/wiki/Intel_MIC), and is designed for high degrees of parallelism.
At the International Supercomputer Conference in Hamburg, HPC.com interviewed James Reinders on the new Xenon Phi
and that video appears here - http://www.youtube.com/watch?v=eiiuvOxFUM0
You can also read more about the Xenon Phi here on a Jun 18th blog on Intel's web site -
http://blogs.intel.com/technology/2012/06/intel-xeon-phi-coprocessors-accelerate-discovery-and-innovation/
The Cray Cascade will use a Linux software environment.
One exciting upshot of technologies such as the Xenon Phi is that it will
make its way into home computing. Gamers out there would no doubt salivate
at the prospect of co-processors providing multi teraflop performance with
their machines networked to others in the country via a 100Mb/s optical fiber.
Over the course of the SKA's life, it will be fascinating to watch the continued
pace of high performance computing technology and whether it will keep up
with the project's thirst for computing cycles and data storage.
exceed that which is available today.
Part of the jigsaw puzzle is about to fall in place with the Australian
government announcing yesterday the procurement of a 1.2 Petaflop Cray Cascade (http://www.cray.com/Programs/Cascade.aspx)
supercomputer which will be installed at the Pawsey Center in Perth.
A Peta is 1 x 10 to the power of 15.
Cray describe the machine as their "Next Generation Supercomputer (http://www.cray.com/Programs/Cascade.aspx)".
iVEC is an organization devoted to providing supercomuting capacity to
Western Australian researchers and is a joint venture between CSIRO and
WA's four public universities. iVEC and CSIRO manage the Pawsey
supercomputer center.
In a press release (http://www.ivec.org/node/532), iVEC announced -
Also purchased were components from SGI and Oracle and the system will be interfaced via dark fiber cables leased
from the National Broadband Network which run to the correlators at the Murchison antenna site.
The Cascade is Cray's first supercomputer based on next generation Intel Xenon processors.
Only last month, on 18th June, Intel announced its new Xenon Phi co-processor. Cray simultaneously
announced that the Cascade would include the Xenon Phi along with the Xeon CPU in its architecture.
The Xenon Phi took five years to design and is currently only in alpha-test pre-production but can provide teraflop performance on
a single chip. It employs 22nm fabrication technology and 3D Tri-gate transistors. Each chip has 50 computing
cores, what Intel refer to a their Many Integrated Core Architecture or MIC (http://en.wikipedia.org/wiki/Intel_MIC), and is designed for high degrees of parallelism.
At the International Supercomputer Conference in Hamburg, HPC.com interviewed James Reinders on the new Xenon Phi
and that video appears here - http://www.youtube.com/watch?v=eiiuvOxFUM0
You can also read more about the Xenon Phi here on a Jun 18th blog on Intel's web site -
http://blogs.intel.com/technology/2012/06/intel-xeon-phi-coprocessors-accelerate-discovery-and-innovation/
The Cray Cascade will use a Linux software environment.
One exciting upshot of technologies such as the Xenon Phi is that it will
make its way into home computing. Gamers out there would no doubt salivate
at the prospect of co-processors providing multi teraflop performance with
their machines networked to others in the country via a 100Mb/s optical fiber.
Over the course of the SKA's life, it will be fascinating to watch the continued
pace of high performance computing technology and whether it will keep up
with the project's thirst for computing cycles and data storage.