News & Analysis
Comment
bhhlophe
I wonder how many of those systems use FPGAs for algorithm acceleration as ...
Jack.L
We do have cost effective super computing .... it is likely sitting on your desk ...
Japan takes Sparc to 8 Petaflops
Rick Merritt
6/20/2011 4:00 AM EDT
Challengers on the horizon
Despite its huge lead, challengers are already on the horizon. The IBM Blue Waters and potentially the Cray XK6 will be its most likely challengers for the top spot, said Dongarra.
Blue Waters is a joint effort of the University of Illinois at Urbana-Champaign, its National Center for Supercomputing Applications, IBM, and the Great Lakes Consortium for Petascale Computation. It uses the IBM Power 7, a CPU with eight four-way multithreaded cores—up to 300,000 of them—but the design has faced some delays, Dongarra said.
Blue Waters also uses a new interconnect based on a mixture of ideas drawn from Infiniband and IBM's proprietary designs. A single IBM hub chip implementing the interconnect supplies an aggregate 1,128 GBytes/s peak bandwidth divided up between various nodes, super-nodes and general I/O tasks.
The Cray XK6 is a hybrid system using AMD x86 processors and Nvidia GPUs. It will also use a custom interconnect.
Intel, Ethernet still dominate
For the first time, all of the top 10 systems achieved petaflop-level performance. The U.S. leads with five systems performing at that level, Japan and China have two each and France has one.
Intel continues to dominate the list with its chips in 387 (77.4 percent) of Top 500 systems, slightly down from 398 systems (79.6 percent) six months ago. Intel’s Westmere processors are now used in 169 systems, up from 56 in the November rankings.
AMD follows with 65 systems (13.0 percent), up from 57. IBM trails with its Power processors in 45 systems (9.0 percent), up from 40.
CPUs with six or more cores are now in 212 systems (42.4 percent) with most of the rest (231 system, 46.2 percent) using quad-core chips.
Gigabit Ethernet is still the most-used internal system interconnect technology used in 233 systems, up from 227 systems. It is followed by Infiniband in 205 systems, down from 214 systems. However, Infiniband-based systems account for almost twice as much performance (23.0 petaflops) than Gigabit Ethernet ones (11.6 petaflops).
IBM and Hewlett-Packard continue to sell the bulk of the systems at all performance levels of the Top 500. IBM has 213 systems (42.6 percent) on the current list, compared to HP with 153 systems (30.4 percent). HP had 158 systems (31.6 percent) six months ago, compared to IBM with 200 systems (40 percent).
The U.S. is the leading user of supercomputers with 256 of the 500 systems (down from 274). Europe is home to 125 systems, and Asia has 103 systems--up from 84.
China keeps increasing its number of systems and is now up to 62, making it clearly the second largest user of supercomputers, ahead of Germany, the UK, Japan and France.
Despite its huge lead, challengers are already on the horizon. The IBM Blue Waters and potentially the Cray XK6 will be its most likely challengers for the top spot, said Dongarra.
Blue Waters is a joint effort of the University of Illinois at Urbana-Champaign, its National Center for Supercomputing Applications, IBM, and the Great Lakes Consortium for Petascale Computation. It uses the IBM Power 7, a CPU with eight four-way multithreaded cores—up to 300,000 of them—but the design has faced some delays, Dongarra said.
Blue Waters also uses a new interconnect based on a mixture of ideas drawn from Infiniband and IBM's proprietary designs. A single IBM hub chip implementing the interconnect supplies an aggregate 1,128 GBytes/s peak bandwidth divided up between various nodes, super-nodes and general I/O tasks.
The Cray XK6 is a hybrid system using AMD x86 processors and Nvidia GPUs. It will also use a custom interconnect.
Intel, Ethernet still dominate
For the first time, all of the top 10 systems achieved petaflop-level performance. The U.S. leads with five systems performing at that level, Japan and China have two each and France has one.
Intel continues to dominate the list with its chips in 387 (77.4 percent) of Top 500 systems, slightly down from 398 systems (79.6 percent) six months ago. Intel’s Westmere processors are now used in 169 systems, up from 56 in the November rankings.
AMD follows with 65 systems (13.0 percent), up from 57. IBM trails with its Power processors in 45 systems (9.0 percent), up from 40.
CPUs with six or more cores are now in 212 systems (42.4 percent) with most of the rest (231 system, 46.2 percent) using quad-core chips.
Gigabit Ethernet is still the most-used internal system interconnect technology used in 233 systems, up from 227 systems. It is followed by Infiniband in 205 systems, down from 214 systems. However, Infiniband-based systems account for almost twice as much performance (23.0 petaflops) than Gigabit Ethernet ones (11.6 petaflops).
IBM and Hewlett-Packard continue to sell the bulk of the systems at all performance levels of the Top 500. IBM has 213 systems (42.6 percent) on the current list, compared to HP with 153 systems (30.4 percent). HP had 158 systems (31.6 percent) six months ago, compared to IBM with 200 systems (40 percent).
The U.S. is the leading user of supercomputers with 256 of the 500 systems (down from 274). Europe is home to 125 systems, and Asia has 103 systems--up from 84.
China keeps increasing its number of systems and is now up to 62, making it clearly the second largest user of supercomputers, ahead of Germany, the UK, Japan and France.
|
|
|
|
|
Navigate to related information
sohonokos
6/20/2011 12:57 PM EDT
10MW with 68,000 processers?! D-Wave is targeting a couple of Kilowatt pwr with a SINGLE chip processor.
Sign in to Reply
kqalea
6/20/2011 11:42 PM EDT
hmm, I don't think D-wave Get any compiler to use :)
beside The D-wave never say they can stable calculate long time
8 Petaflops sound much real to me
Sign in to Reply
selinz
6/20/2011 3:56 PM EDT
I'd be curious to see how the users of these system compare. Entertainment, gaming, academic, industrial, etc.
Sign in to Reply
Neo1
6/21/2011 12:46 AM EDT
Quite impressive I would say, that too without using the GPUs for number crunching. That they used a Sparc makes it all the more interesting because of it's legacy architecture.
I guess nearly 30% of that power is used by the system itself apart of the cores. But still it is a huge power and could nearly dominate the expenditure for any institution using this level of processing.
of course this mega processing would only be deployed in govt R&D and weather stations in drug development by big pharma companies.
Though these numbers are pleasing, it would be more valuableto see some really affordable super computing within the budget of a medium sized company, say for ex: $100k-500K
Sign in to Reply
Jack.L
6/21/2011 1:45 PM EDT
We do have cost effective super computing .... it is likely sitting on your desk in front of you!
Supercomputing is all time relative.
A Cray-1 peaked at 160Mips, 250MFlops.... a small fraction of what the processor and GPU in your laptop and desktop are capable of.
So yes, this level of performance will arrive at $100K - $500K, then $5K and maybe even $500 ... it will just take time.
Sign in to Reply
bhhlophe
6/22/2011 2:40 PM EDT
I wonder how many of those systems use FPGAs for algorithm acceleration as Xilinx and Alteration claim massive performance gains.
Imagine a supercomputer with Xilinx Virtex-7-2000s that can be cool but also a programming challenge.
Sign in to Reply