India will have 'super-supercomputer' by 2017

[damiendehorn]

The Times of India is reporting that the country will begin building a new series of domestically produced supercomputers, the first of which will be installed in August of 2017. The Rs 4,500-crore (about $670 million) project is being managed by the Centre for Development of Advanced Computing (C-DAC).

In 2016, the government came up with a plan to build as many as 80 supercomputers over a period seven years. Not all will be developed in India, however. According to Ashutosh Sharma, Secretary in the Ministry of Science and Technology, some will be imported, while others will be built indigenously.

This isn’t the first time India has developed homegrown supercomputers. Starting in 1991, the country built and deployed a series of domestically produced systems, known as PARAM, which is Sanskrit for supreme. The effort was initiated due to a US arms embargo in the late 1980s, which prevented India from importing Cray systems at the time. Nine PARAM systems have been built, culminating with the PARAM Yuva II, a 524-teraflop machines that is powered by a combination of Intel Xeon and Xeon Phi processors.

The PARAM Yuva II is the only remaining domestically-produced supercomputer from India on the TOP500, ranking at number 165. The other 10 Indian systems on the list were built by Hewlett Packard Enterprise (6), IBM (2), and Cray (2). India’s top system is SahasraT, a Cray XC40, whose LINPACK performance is officially just shy of 1 petaflop, although at the time of delivery in 2015 it was supposed to provide more than 1.4 peak petaflops.

Presumably the machines under development at C-DAC will put India squarely into the petaflop club. No word on what the new systems will be called.

PARAM Yuva - II - R2208GZ Cluster, Xeon E5-2670 8C 2.600GHz, Infiniband FDR, Intel Xeon Phi 5110P
Site: Center for Development of Advanced Computing (C-DAC)
System URL: https://yuva.cdac.in/YuvaII/
Manufacturer: Netweb Technologies
Cores: 30,056
Linpack Performance (Rmax) 388.442 TFlop/s
Theoretical Peak (Rpeak) 520.358 TFlop/s
Nmax 1,225,088
Power: 209.95 kW (Submitted)
Memory: 14,144 GB
Processor: Xeon E5-2670 8C 2.6GHz
Interconnect: Infiniband FDR
Operating System: CentOS
Compiler: Intel Cluster Studio
Math Library: Intel MKL 11.0
MPI: Intel MPI

Current ranking on Top500 is 447

 

[cirr]

There are more than one ways to skin a cat. Ever heard of FPGAs? Unlike CPU-GPU kind of nodes which China's Ding-Dong super computers are based on, an FPGA based technology is much more flexible because the data paths are not fixed. Meaning, you can change the parallel processing as per the problem you are trying to solve. And FPGA scales really well with semiconductor denisity growth, after all at its core FPGAs are SRAM based lookup-tables. With such kind of technology it is very much possible for India to achieve 132.8 exaflops. And this is just one possible way for India to achieve this. There are many many possible architectures to deliver this performance.

FPGA?

You lot can't make a bloody lithium battery let alone chips of any description.

You are well advised to keep your mouth shut till such a time when your great great scientists and engineers come up with a computer that's 1/10th as fast as China's Taihulight. And that with any foreign help and input you may possibly solicit.

Exascale？Not in your wet dreams.

132 exascale? You need to have your head examined.

 

[cirr]

The Times of India is reporting that the country will begin building a new series of domestically produced supercomputers, the first of which will be installed in August of 2017. The Rs 4,500-crore (about $670 million) project is being managed by the Centre for Development of Advanced Computing (C-DAC).

In 2016, the government came up with a plan to build as many as 80 supercomputers over a period seven years. Not all will be developed in India, however. According to Ashutosh Sharma, Secretary in the Ministry of Science and Technology, some will be imported, while others will be built indigenously.

This isn’t the first time India has developed homegrown supercomputers. Starting in 1991, the country built and deployed a series of domestically produced systems, known as PARAM, which is Sanskrit for supreme. The effort was initiated due to a US arms embargo in the late 1980s, which prevented India from importing Cray systems at the time. Nine PARAM systems have been built, culminating with the PARAM Yuva II, a 524-teraflop machines that is powered by a combination of Intel Xeon and Xeon Phi processors.

The PARAM Yuva II is the only remaining domestically-produced supercomputer from India on the TOP500, ranking at number 165. The other 10 Indian systems on the list were built by Hewlett Packard Enterprise (6), IBM (2), and Cray (2). India’s top system is SahasraT, a Cray XC40, whose LINPACK performance is officially just shy of 1 petaflop, although at the time of delivery in 2015 it was supposed to provide more than 1.4 peak petaflops.

Presumably the machines under development at C-DAC will put India squarely into the petaflop club. No word on what the new systems will be called.

PARAM Yuva - II - R2208GZ Cluster, Xeon E5-2670 8C 2.600GHz, Infiniband FDR, Intel Xeon Phi 5110P
Site: Center for Development of Advanced Computing (C-DAC)
System URL: https://yuva.cdac.in/YuvaII/
Manufacturer: Netweb Technologies
Cores: 30,056
Linpack Performance (Rmax) 388.442 TFlop/s
Theoretical Peak (Rpeak) 520.358 TFlop/s
Nmax 1,225,088
Power: 209.95 kW (Submitted)
Memory: 14,144 GB
Processor: Xeon E5-2670 8C 2.6GHz
Interconnect: Infiniband FDR
Operating System: CentOS
Compiler: Intel Cluster Studio
Math Library: Intel MKL 11.0
MPI: Intel MPI

Current ranking on Top500 is 447

670 million US dollars?

China spent just over 260 million USD for its Taihulight HPC, currently the world's fastest.

India sure has a way with spending its hard earned and limited cash.

 

[Beast]

FPGA?

You lot can't make a bloody lithium battery let alone chips of any description.

You are well advised to keep your mouth shut till such a time when your great great scientists and engineers come up with a computer that's 1/10th as fast as China's Taihulight. And that with any foreign help and input you may possibly solicit.

Exascale？Not in your wet dreams.

132 exascale? You need to have your head examined.

There's a limit to bragging. But Indian has cross the threshold.

 

[Kyusuibu Honbu]

@Deepika Maheshwari
I request you to edit racial slurs in your post 33, no point getting down to their level

 

[Hassan Guy]

About time they updated those call centre's.

 

[damiendehorn]

For clarity:

CHINA...Rmax 93,014.6 TFlop/s (93 PFlops/s), Rpeak 125,436 TFlop/s (125 PFlops/s)

Sunway TaihuLight - Sunway MPP, Sunway SW26010 260C 1.45GHz, Sunway
Site: National Supercomputing Center in Wuxi
Manufacturer: NRCPC
Cores: 10,649,600
Linpack Performance (Rmax) 93,014.6 TFlop/s
Theoretical Peak (Rpeak) 125,436 TFlop/s
Nmax 12,288,000
Power: 15,371.00 kW (Submitted)
Memory: 1,310,720 GB
Processor: Sunway SW26010 260C 1.45GHz
Interconnect: Sunway
Operating System: Sunway RaiseOS 2.0.5

INDIA....Rmax 388.442 TFlop/s (0.388 PFlops/s), Rpeak 520.358 TFlop/s (0.52 PFlops/s)

PARAM Yuva - II - R2208GZ Cluster, Xeon E5-2670 8C 2.600GHz, Infiniband FDR, Intel Xeon Phi 5110P
Site: Center for Development of Advanced Computing (C-DAC)
System URL: https://yuva.cdac.in/YuvaII/
Manufacturer: Netweb Technologies
Cores: 30,056
Linpack Performance (Rmax) 388.442 TFlop/s
Theoretical Peak (Rpeak) 520.358 TFlop/s
Nmax 1,225,088
Power: 209.95 kW (Submitted)
Memory: 14,144 GB
Processor: Xeon E5-2670 8C 2.6GHz
Interconnect: Infiniband FDR
Operating System: CentOS
Compiler: Intel Cluster Studio
Math Library: Intel MKL 11.0
MPI: Intel MPI

Unless C-DAC of india has a major computing breakthrough, I doubt they will come anywhere near building a exascale computing centre. To achieve this they would need to build a computer thats over 2000 times the speed of their fastest current C-DAC system. The latest Cray XC-50 can achieve about 1 PFlop/s (Rpeak) per cabinet, so in theory it would need 1,000 + of these to get anywhere near that number. Imagine the power that would need!!!

If C-DAC manages to build a afordable system that manages to achieve 20 PFlop/s (Rmax) and 35 PFlop/s (Rpeak) would be realistic.

 

[Dungeness]

Question! Does USA fabricates all of its own silicon chips? FUCKKEN-NO you idiot chink!

All India needs to do is to work with Tiwanese fabricators and companies with FPGA design - likes of Xilinx or Altera. Rest is all software, which Indians are master in, TYVM. Now you chink, go on a Wong-March or we will drive a tank on you in Ting-Tong-Square.
Cya, Mr Ching Chong!

Your undisguised hatred towards China is beyond comprehension, the worst I have seen on PDF. The hateful people like you, along with your big mouthed politicians and your notoriously irresponsible media, help to make your otherwise great country laughing stock at times.

We will give you the benefit of doubt and congratulate you when the following tall claims actually Happen:

Maned Moon Landing by 2020;
132.8 Exaflops Super-supercomputer by 2017;
Maned Space Flight by 2015;
and of course, Superpower by 2012.

 

[Eminent Mainstream Media]

670 million US dollars?

China spent just over 260 million USD for its Taihulight HPC, currently the world's fastest.

India sure has a way with spending its hard earned and limited cash.

Seems like you have problem reading- the cost is for 80 supercomputers to be distributed to various Indian Universities and Institutions-

 

[lonelyman]

Yes. when they were hyping their petaflop computing capability and I posted the article claiming of exaflop capability, they collectively came down heavily on me and insulted me. One chinese came out with a calculation that some 7600 MW electricity shall be required to run the computer etc. As you rightly said, they stand tall in narrow nationalism and narrow thinking.

On the topic:

We shall be able to carry out designing simulation, fluid dynamic simulation and whether forecasting with a great accuracy along with many many complex calculations. Many time consuming studies shall became very easy. We shall be able to make very very compact design Hydrogen bomb.

Bullshit u guys only good at boasting PowerPoint, 2012 super power stuff

Oh it is nothing new. People even mocked at bicycle when they saw it for the first time. If any of such non conventional architecture arise out of India's pursuit of super-computer technology, they all will go quite.

Day dreaming at its best

Yes but who can not conceive such technology mock it out of inferiority complex.

Day dreaming paradise

For clarity:

CHINA...Rmax 93,014.6 TFlop/s (93 PFlops/s), Rpeak 125,436 TFlop/s (125 PFlops/s)

Sunway TaihuLight - Sunway MPP, Sunway SW26010 260C 1.45GHz, Sunway
Site: National Supercomputing Center in Wuxi
Manufacturer: NRCPC
Cores: 10,649,600
Linpack Performance (Rmax) 93,014.6 TFlop/s
Theoretical Peak (Rpeak) 125,436 TFlop/s
Nmax 12,288,000
Power: 15,371.00 kW (Submitted)
Memory: 1,310,720 GB
Processor: Sunway SW26010 260C 1.45GHz
Interconnect: Sunway
Operating System: Sunway RaiseOS 2.0.5

INDIA....Rmax 388.442 TFlop/s (0.388 PFlops/s), Rpeak 520.358 TFlop/s (0.52 PFlops/s)

PARAM Yuva - II - R2208GZ Cluster, Xeon E5-2670 8C 2.600GHz, Infiniband FDR, Intel Xeon Phi 5110P
Site: Center for Development of Advanced Computing (C-DAC)
System URL: https://yuva.cdac.in/YuvaII/
Manufacturer: Netweb Technologies
Cores: 30,056
Linpack Performance (Rmax) 388.442 TFlop/s
Theoretical Peak (Rpeak) 520.358 TFlop/s
Nmax 1,225,088
Power: 209.95 kW (Submitted)
Memory: 14,144 GB
Processor: Xeon E5-2670 8C 2.6GHz
Interconnect: Infiniband FDR
Operating System: CentOS
Compiler: Intel Cluster Studio
Math Library: Intel MKL 11.0
MPI: Intel MPI

Unless C-DAC of india has a major computing breakthrough, I doubt they will come anywhere near building a exascale computing centre. To achieve this they would need to build a computer thats over 2000 times the speed of their fastest current C-DAC system. The latest Cray XC-50 can achieve about 1 PFlop/s (Rpeak) per cabinet, so in theory it would need 1,000 + of these to get anywhere near that number. Imagine the power that would need!!!

If C-DAC manages to build a afordable system that manages to achieve 20 PFlop/s (Rmax) and 35 PFlop/s (Rpeak) would be realistic.

Hard facts, boasting Indians not good at

 

[Dungeness]

Ya but they are unable to achieve such thing and they have some 250 odd peta flop computing power and they do not have any plan in excess of 1 exa flop so they are unable to diget it. They think that that there is no other way to achieve such high computing speed. They are more butt hurt because they have hyped their computer a lot and they are saying that it is the fastes. Once ours will come in , they will not be able to match it in any foreseeable future. That worries them. It will be one more are where we shall outclass them by a huge margin. If they were humble , they will congratulate India but they won’t. They will try their best to miniaturize this achievement.

What achievement are you talking about? You call this 132 exaflpos super-supercomputer only exists on Toi as an achievement? No wonder people say your lot have a hard time distinguishing between delusion and reality.

We love to see India makes progress in every field, it is people like you keep embarrassing your own country and made it a laughing stock, with your unmatched ignorance and very loud mouth.

 

[AndrewJin]

Your undisguised hatred towards China is beyond comprehension, the worst I have seen on PDF. The hateful people like you, along with your big mouthed politicians and your notoriously irresponsible media, help to make your otherwise great country laughing stock at times.

We will give you the benefit of doubt and congratulate you when the following tall claims actually Happen:

Maned Moon Landing by 2020;
132.8 Exaflops Super-supercomputer by 2017;
Maned Space Flight by 2015;
and of course, Superpower by 2012.

Congrats ahead

What achievement are you talking about? You call this 132 exaflpos super-supercomputer only exists on Toi as an achievement? No wonder people say your lot have a hard time distinguishing between delusion and reality.

We love to see India makes progress in every field, it is people like you keep embarrassing your own country and made it a laughing stock, with your unmatched ignorance and very loud mouth.

Delusion is reality for them.
That's the mentality of their entire country.

670 million US dollars?

China spent just over 260 million USD for its Taihulight HPC, currently the world's fastest.

India sure has a way with spending its hard earned and limited cash.

lol, 99% money will be in their private bank accounts, btw, they won't be in 2000 rupees notes

 

[AndrewJin]

2017 is quickly approaching. Is there any chance India will deliver on this? Or is this yet another example of the famous Indian empty talk?


http://timesofindia.indiatimes.com/...-by-2017/articleshow/19450713.cms?referral=PM

India will have 'super-supercomputer' by 2017
NAGPUR: Indian scientists have always responded well to challenges and crises. One such challenge that lies ahead is the development of a 'super-supercomputer' along with USA and China who are already rushing for it. The Indian supercomputer man Vijay Bhatkar , who developed the Param supercomputer when USA denied technology to India, expressed confidence that India would be able to meet this challenge too by 2020. He believes India will become the world's largest economy by 2047 through developments in science and technology.

Bhatkar was speaking at the 56th foundation day programme of National Environmental Engineering Research Institute (Neeri). The idea of Exascale computer (computer with a capability of at least one exaflops -which is thousandfold increase over the first peta-scale computer developed in 2008. One exaflop has a computing ability of a quintillion, 1018, point operations per second.)

The idea of such a computer was first floated in 2009 at the supercomputing conference in 2009. Intel has announced it would develop the computer by 2018. USA has put aside a substantial sum for it while India too has committed $ 2 billion. Indian Space Research Organization (ISRO) and the Indian Institute of Science Bangalore are expected to develop a supercomputer with 132.8 exaflops by 2017.

Even granting it is developed by 2020, Bhatkar called upon the environmental scientists and mathematical modelling experts to make extremely precise weather predictions and effects of ongoing climate change in future at regional and national level. Environment science, he said, would be at the centre of all sciences and hence Neeri had a big role to play in future.

Calling upon younger generation of scientists to focus their minds for innovations in all fields, Bhatkar cited examples of innovations like the supercomputer, Apple's gadgets that changed the lives of millions. He said innovations should be capable of being converted into commercial products. Scientists should also innovate for solving people's problems and this, he said, would be possible with the recently announced science technology and innovation policy of the country.

Earlier, in his foundation day address, Neeri director Satish Wate listed the achievements of the institute in past one year. Neeri, after 1992-93, has for the first time generated an earning of over Rs 23 crore from sponsored research projects, an indication of the faith of industry and the people in the institution, he informed. Neeri this year won the prestigious world level award for development of electrolytic deflouridation technology. The institution also started its journey towards upgradation of infrastructure in next five year plan with a financial support of Rs 60 crore, he added.

To mark the occasion, Neeri had put up an exhibition of 70 breakthrough technologies from different laboratories of Council of Scientific and Industrial Research (CSIR) in fields of health and medicine, agriculture, water, leather etc for the public and it received good response. The institute also organized a model making competition for schools and college students.

Jaya Sabjiwale compered the programme and Prakash Kumbhare summed up.

Scientists get awards

Continuing its tradition of encouraging scientists for better research, Neeri director Satish Wate gave away prizes to scientist and technical staff for their contributions during the past year. The list of awardees included Nitin Labhsetwar-outstanding senior scientist, Pravin Naoghare-- outstanding junior scientist, C Padmakar-- outstanding technical officer, V M Kulkarni, outstanding technician, Vandana Cinthray--outstanding office personnel, Sharvari Deshmukh- outstanding young researcher, and Piyush Mahore-outstanding project personnel.

Awards were also given to best performance by various departments. These included Environmental Materials Division for best publications, Environmental Biotechnology Division for an international patent (which went to D B Satpute and D M Dharmadhikari), and Environmental Impact Risk Assessment division for maximum earning from extra budgetary resource and Water Technology and Management Division for electro-deflouridation technology.

Let's celebrate their launch ahead of time!

 

[AZADPAKISTAN2009]

Me too

 

[cirr]

There are more than one ways to skin a cat. Ever heard of FPGAs? Unlike CPU-GPU kind of nodes which China's Ding-Dong super computers are based on, an FPGA based technology is much more flexible because the data paths are not fixed. Meaning, you can change the parallel processing as per the problem you are trying to solve. And FPGA scales really well with semiconductor denisity growth, after all at its core FPGAs are SRAM based lookup-tables. With such kind of technology it is very much possible for India to achieve 132.8 exaflops. And this is just one possible way for India to achieve this. There are many many possible architectures to deliver this performance.

Listen dunderhead, stop sounding as if the likes of Altera, Xilink and Lattice are Indian companies.

Does India make chips of any sort? I mean ANY sort? The answer is a resounding NO.

In tne meanwhile, China has started using FPGAs onboard its micro-satellite and satellite constellation:

中新网哈尔滨11月18日电 (记者 王舒 史轶夫)18日晚，哈尔滨工业大学对外发布消息称，作为“长征十一号”运载火箭搭载的脉冲星试验卫星“XPNAV-1”载荷之一，由该校研制的“可编程SOC计算载荷”打破国外高性能宇航级计算器件的禁运限制，将完成一系列科学验证与空间探测实验。

本月10日7时42分，“长征十一号”运载火箭发射成功，火箭搭载的脉冲星试验卫星“XPNAV-1”顺利升空并成功进入预定轨道。作为星上载荷之一，哈工大电气学院自动化测试与控制研究所研制的“可编程SOC计算载荷”开始了空间探索之旅，将完成一系列科学验证与空间探测实验。此次搭载实验是研究所在嵌入式高性能计算、空间辐射防护技术、系统健康管理等领域研究成果在空间的首次工程化验证。

“可编程SOC计算载荷”旨在应对未来微小卫星、卫星星群的高性能低成本计算需求。研究针对空间辐射环境的可靠性设计加固技术，采用工业级高性能异构器件实现故障可容忍的空间计算系统，从而可将星载计算机性能提升10倍以上，突破每秒20亿条指令，并大幅降低成本，打破国外高性能宇航级计算器件的禁运限制。

目前，该计算载荷已经在轨开机自检，数据传输及各项指标均正常。后续该计算载荷将完成工业级器件空间辐射防护设计技术验证，完成工业级静态随机存储器(SRAM)工艺现场可编程门阵列器件(FPGA)的可重构技术验证、单粒子翻转及防护方法技术验证、在轨总剂量性能退化实时监测技术验证、在轨闩锁检测防护技术验证，并进行空间高光谱图像分类处理方法原型验证、空间锂离子电池荷电状态评估方法技术验证、基于异构加速方法的飞轮动力学方程解算技术验证等一系列实验。

通过此次搭载，实验成果将为卫星尤其是微小卫星、深空探测、空间站等航天任务提供具备自主知识产权的在轨计算能力支持，解决装备在轨自主运行管理、在轨系统健康管理、装备人工智能等先进技术理念的空间应用计算瓶颈问题，降低微小卫星成本，助力中国航天事业发展。(完)

Have no idea what the above says? Time to learn Chinese dude.