Ranger NETSHOW Forum Register Login
China's first petaflop supercomputer system - High-performance computers in China, 2009 TOP100 Organization announced the top 100 rankings, the Milky Way One topped the list. It is also the birth of the world, after China became the second American to develop a petaflop supercomputer country.
Tianhe One is not only China's first petaflop supercomputer, and the use of the innovative CPU + GPU design of heterogeneous computing,mac makeup, not only significantly enhance the performance of theoretical calculations, and reached a very high energy efficiency. Efficiency of 431.7 MFlops / W,mac brushes, the current arrangement in the Green500 fifth.
Milky Way One of the theoretical peak computing performance of up to 1,206,000,000,000,000 times double-precision floating-point operations, LINPACK performance of 560.3 trillion measured times. The measured performance data computer system has been submitted to the International TOP500 ranking, in accordance with the June 2009 performance data, performance measurement arrangement TOP500 Tianhe One fourth, third place at peak performance.
Milky Way One common use of multi-core Intel 6144 processor and 5120 graphics acceleration AMD processors, memory total capacity of 98TB, point to point communication bandwidth of 40Gbps, while the total capacity to achieve the shared disk 1PB . The supercomputer system will be deployed in Tianjin Binhai New Area of Tianjin National Supercomputing Center as a business host.
high performance computing in China under the 2009 TOP100 list published data, the number one computing power far more than second place, of course, the biggest bright spot is the first time calculated using the GPU acceleration, but also AMD's GPU:
Milky Way One of the more configuration follows:
computing array: 2560 computing nodes, each computing node integrates two Intel CPU, with 32GB of memory.
accelerate the array: 2560 accelerated nodes, each node with 2 speed AMD GPU, 2GB memory.
service array: 512 service nodes, each service node with two Intel CPU, 32GB memory.
using a total of 3072 Intel Quad Core Xeon E5540 2.53GHz and the 3072 Intel Quad Core Xeon E5450 3.0GHz, a total of 6144 CPU, 24576 cores.
using a total of 2560 AMD Radeon HD4870X2 graphics card to run in the 5120's 575MHz RV770 GPU, each GPU has 1GB GDDR5 memory pieces.
motherboard through each node point of view, is really two quad-core CPU with a HD4870X2 graphic card design, appearance and general graphics no difference, but even lower operating frequency than the HD4850, it should be to ensure stability, from the perspective of energy efficiency.
know the CPU and GPU configurations in detail, we again count the ways E5450 theoretical peak computing power as:
3072x (40.48 +48) = 271810.56GFlops ≈ 272TFlops (double precision)
● 2560 块 HD4870X2 theoretical peak computing power as:
; 920x2x2560 / 5 = 942080GFlops ≈ 942TFlops (double precision)
this together, then
That GPU in the contribution of super-computer performance over the CPU!
by estimating the point of view, the current GPU computing efficiency of only about 37%, the official said, after optimized, GPU computational efficiency can be increased from 20% to 70% or so, it seems that the have great potential to be tapped!
Last year, another look at Tesla GPU-based supercomputer, which is built by the Tokyo Institute of Technology's first CPU + GPU-based differences computing architecture supercomputer configuration, and the first to squeeze into the world top 500.
TSUBAME supercomputer by the addition of 170 units Tesla S1070 1U system, the theoretical peak performance for the 170TFLOP, 2008 年 11 月 measured Linpack performance to 77.48 TFLOP, ranked world No. 29, then through continuous commissioning and optimization, Linpack performance increased to 87.01TFlops, 2009 年 6 月 41 ranking.
Tokyo Institute of Technology's TSUBAME overall strength and the This shows that China's National Defense University's the development trend.
Although the Cooperative operation of complex systems, whether large or maintenance are required to pay a high price. Some people may ask: a quadrillion times the ZHOU Xing Chinese Academy of Sciences, said the performance of the system computational performance of each unit is by no means a simple superposition.
low and can not stability. On the surface 'work' of the device more, the result is that 'the three boys no boy'. a
the seven technologies include: multi-array parallel architecture can be configured collaborative, high-speed Internet communications technology expansion, and efficient heterogeneous collaborative computing technology, based on isolation of safety control technology, virtualization support network computing technology, multi-level large-scale power system fault-tolerant integrated control technologies and systems technology.
ZHOU Xing said, the use of a series of innovative technology, this computer has four characteristics:
- high performance. Whether it is 1206000000000000 times of peak performance, or 563.1 trillion times measured Linpack performance, are located in the domestic top, world-class.
- low power consumption. Energy is calculated to create per watt performance, Reference published in June 2009 Green500 (global top 500 supercomputers green) position,
- high security. For users, equivalent to the rent a bank safe, the key to master their own hands.
- easy to use. Meanwhile, as a general standardization of the international super-computer,vibram 5 fingers,
carrying three functions of public service facilities in accordance with the project
, provide supercomputing services.
First, as to provide supercomputing services to the global public research facilities, No matter where the user can through the Internet using the
Second, as an industry driven platform,
Thirdly, as a platform to attract talent,
Milky Way One: The numerical simulation of typhoon forecasting, aided design aircraft aerodynamic shape
giant computer users in China is mainly concentrated in oil exploration, weather forecasting,mac makeup wholesale, basic research and other fields. The end of the last century, Western countries automotive, construction, animation and other industries have begun to spread a giant computer.
and satellite remote sensing data processing, data analysis, financial engineering, power and pollution of the Bohai Sea ecosystem simulation, theoretical calculation, and basic science.
with the National Center for Supercomputing and began construction in Tianjin, as well as the center in Shenzhen. One north of two national supercomputer center, will become an important engine for high-tech development in China. Is it just science supercomputer
's Nan said that the application of super computers with people's daily lives, the most common weather is the result of supercomputing. With the supercomputer capability of services to further expand in the future, is likely to give birth to a large supercomputer services industry, providing services for individual users. ■
----- Net bubble
China's first petaflop supercomputer system - High-performance computers in China, 2009 TOP100 Organization announced the top 100 rankings, the Milky Way One topped the list. It is also the birth of the world, after China became the second American to develop a petaflop supercomputer country.
Tianhe One is not only China's first petaflop supercomputer, and the use of the innovative CPU + GPU design of heterogeneous computing,mac makeup, not only significantly enhance the performance of theoretical calculations, and reached a very high energy efficiency. Efficiency of 431.7 MFlops / W,mac brushes, the current arrangement in the Green500 fifth.
Milky Way One of the theoretical peak computing performance of up to 1,206,000,000,000,000 times double-precision floating-point operations, LINPACK performance of 560.3 trillion measured times. The measured performance data computer system has been submitted to the International TOP500 ranking, in accordance with the June 2009 performance data, performance measurement arrangement TOP500 Tianhe One fourth, third place at peak performance.
Milky Way One common use of multi-core Intel 6144 processor and 5120 graphics acceleration AMD processors, memory total capacity of 98TB, point to point communication bandwidth of 40Gbps, while the total capacity to achieve the shared disk 1PB . The supercomputer system will be deployed in Tianjin Binhai New Area of Tianjin National Supercomputing Center as a business host.
high performance computing in China under the 2009 TOP100 list published data, the number one computing power far more than second place, of course, the biggest bright spot is the first time calculated using the GPU acceleration, but also AMD's GPU:
Milky Way One of the more configuration follows:
computing array: 2560 computing nodes, each computing node integrates two Intel CPU, with 32GB of memory.
accelerate the array: 2560 accelerated nodes, each node with 2 speed AMD GPU, 2GB memory.
service array: 512 service nodes, each service node with two Intel CPU, 32GB memory.
using a total of 3072 Intel Quad Core Xeon E5540 2.53GHz and the 3072 Intel Quad Core Xeon E5450 3.0GHz, a total of 6144 CPU, 24576 cores.
using a total of 2560 AMD Radeon HD4870X2 graphics card to run in the 5120's 575MHz RV770 GPU, each GPU has 1GB GDDR5 memory pieces.
motherboard through each node point of view, is really two quad-core CPU with a HD4870X2 graphic card design, appearance and general graphics no difference, but even lower operating frequency than the HD4850, it should be to ensure stability, from the perspective of energy efficiency.
know the CPU and GPU configurations in detail, we again count the ways E5450 theoretical peak computing power as:
3072x (40.48 +48) = 271810.56GFlops ≈ 272TFlops (double precision)
● 2560 块 HD4870X2 theoretical peak computing power as:
; 920x2x2560 / 5 = 942080GFlops ≈ 942TFlops (double precision)
this together, then
That GPU in the contribution of super-computer performance over the CPU!
by estimating the point of view, the current GPU computing efficiency of only about 37%, the official said, after optimized, GPU computational efficiency can be increased from 20% to 70% or so, it seems that the have great potential to be tapped!
Last year, another look at Tesla GPU-based supercomputer, which is built by the Tokyo Institute of Technology's first CPU + GPU-based differences computing architecture supercomputer configuration, and the first to squeeze into the world top 500.
TSUBAME supercomputer by the addition of 170 units Tesla S1070 1U system, the theoretical peak performance for the 170TFLOP, 2008 年 11 月 measured Linpack performance to 77.48 TFLOP, ranked world No. 29, then through continuous commissioning and optimization, Linpack performance increased to 87.01TFlops, 2009 年 6 月 41 ranking.
Tokyo Institute of Technology's TSUBAME overall strength and the This shows that China's National Defense University's the development trend.
Although the Cooperative operation of complex systems, whether large or maintenance are required to pay a high price. Some people may ask: a quadrillion times the ZHOU Xing Chinese Academy of Sciences, said the performance of the system computational performance of each unit is by no means a simple superposition.
low and can not stability. On the surface 'work' of the device more, the result is that 'the three boys no boy'. a
the seven technologies include: multi-array parallel architecture can be configured collaborative, high-speed Internet communications technology expansion, and efficient heterogeneous collaborative computing technology, based on isolation of safety control technology, virtualization support network computing technology, multi-level large-scale power system fault-tolerant integrated control technologies and systems technology.
ZHOU Xing said, the use of a series of innovative technology, this computer has four characteristics:
- high performance. Whether it is 1206000000000000 times of peak performance, or 563.1 trillion times measured Linpack performance, are located in the domestic top, world-class.
- low power consumption. Energy is calculated to create per watt performance, Reference published in June 2009 Green500 (global top 500 supercomputers green) position,
- high security. For users, equivalent to the rent a bank safe, the key to master their own hands.
- easy to use. Meanwhile, as a general standardization of the international super-computer,vibram 5 fingers,
carrying three functions of public service facilities in accordance with the project
, provide supercomputing services.
First, as to provide supercomputing services to the global public research facilities, No matter where the user can through the Internet using the
Second, as an industry driven platform,
Thirdly, as a platform to attract talent,
Milky Way One: The numerical simulation of typhoon forecasting, aided design aircraft aerodynamic shape
giant computer users in China is mainly concentrated in oil exploration, weather forecasting,mac makeup wholesale, basic research and other fields. The end of the last century, Western countries automotive, construction, animation and other industries have begun to spread a giant computer.
and satellite remote sensing data processing, data analysis, financial engineering, power and pollution of the Bohai Sea ecosystem simulation, theoretical calculation, and basic science.
with the National Center for Supercomputing and began construction in Tianjin, as well as the center in Shenzhen. One north of two national supercomputer center, will become an important engine for high-tech development in China. Is it just science supercomputer
's Nan said that the application of super computers with people's daily lives, the most common weather is the result of supercomputing. With the supercomputer capability of services to further expand in the future, is likely to give birth to a large supercomputer services industry, providing services for individual users. ■
----- Net bubble
