Indian advancements in Supercomputing

[JAISWAL]

it is sure an ambitious project, but if
.
we dream for moon then surely will land among stars.

 

[Virendra]

yeah baby !!!!!!
thats more like it
way to go

 

[sukhish]

132800 petaflops as compared to China's 2.7 petaflops. 6 years is indeed ambitious.

This is jaw dropping news, I mean exaflops is something I have never even heard of. slowly but surely
india is getting confident in every aspect of technology. I bet even US hasn't even done anything this ambitious.

 

[p2prada]

This is jaw dropping news, I mean exaflops is something I have never even heard of. slowly but surely
india is getting confident in every aspect of technology. I bet even US hasn't even done anything this ambitious.

The US stands at 2 Petaflops, behind China as of today.

US is finding our UID project a little too much to grasp. This will blow their tops off.

 

[chex3009]

It would be a remarkable feat to achieve in the time frame they are expecting it to be.

This will justify our mastery and no one will be able to question India in this field in the future.

 

[JAISWAL]

here is news pics.

.

 

[A.V.]

Please keep this thread updated over the months , we want to follow the development and see if this can be realized at the end of 6 years . if yes we have a winner on hand.

 

[tarunraju]

I really wish its architects look at parallel-processing technologies such as NVIDIA Tesla or AMD Stream, and not spend eight times more on just serial-load multi-core processors.

 

[JAISWAL]

sir,
the very easy and efficient way to achive this speed is to use cloud computing.
As it utilise processer speed more efficently and very cheaply.

 

[mattster]

Isnt the latest supercomputer design always going to be the fastest ??

Given that the most recent guys will have access to the fastest, most number of cores, and lowest-power CPUs as a starting base.

Plus if they can leverage the latest software advances in parallel-processing then it would seems like the latest guy should have the maximum advantage in the speed race.

 

[tarunraju]

sir,
the very easy and efficient way to achive this speed is to use cloud computing.
As it utilise processer speed more efficently and very cheaply.

I think you have no idea how cloud computing works, and how it cannot be used with supercomputing. Cloud computing is outsourcing your processing to data centers, which hold normal web-servers. At the client-side, the application is web-based. High internet bandwidth creates a sort of illusion that the application is locally-run.

You cannot use the cloud model for supercomputing for many reasons. First, you need absolute precision. Most computing models are designed to work with double-precision floating-point to minimize errors. Even the smallest error can collapse an entire computing operation. Second, the latency is extremely high.

That, however, is different from distributed computing, where you break your operation into slices, and distribute it among thousands of client users across the internet. They crunch those slices and send back the results. Thousands are sent the same slice, for increased consistency in results. DisCom is extremely time-consuming, and India doesn't really have a thing for popular DisCom operations such as Folding@Home or WCG.

Hence, the supercomputing model should be traditional, where everything is under one roof, and interconnected using the fast Infiniband connections. The architects should explore technologies such as NVIDIA Tesla, AMD Stream, or even Creative Zii.

 

[no smoking]

Can someone please explain to me how much is about "Rs 10,000-crore "? Thanks!

As to the prospect of this project, we have to wait and see.

 

[Armand2REP]

Can someone please explain to me how much is about "Rs 10,000-crore "? Thanks!

It is around $2 billion.

 

[debasree]

advance kudos to isro ,just wanna say go..........................for it.

 

[Yogesh]

Isnt the latest supercomputer design always going to be the fastest ??

Given that the most recent guys will have access to the fastest, most number of cores, and lowest-power CPUs as a starting base.

Not necessarily, every supercomputer that comes out may not be fastest in the world. For example recently ISRO made a supercomputer which is fastest in India but not in the world. Its just like there are core i7 processors but every computer that is sold doesnt necessarily has those processors, depends on requirements and budget as to how much an organization can invest into making those systems.

As the system will be designed soon , they wont be using processors that will come in 2015 or so. Its design requires a tradeoff between whats available now and how much they can depend on future technologies.

 

[SuperCommandoDhruv]

sir,
the very easy and efficient way to achive this speed is to use cloud computing.
As it utilise processer speed more efficently and very cheaply.

LOL!

It seems you heard that overhyped marketing video by Microsoft over its revolutionary "cloud computing"(just like claims of Moon landing). Its funny how people get "awed" by these MNC seminars. I met this guy(topper in his class) and he was blabbering in fluent english how pathbreaking cloud computing is and blah blah. He had no clue about what basically that thing is.

 

[The Messiah]

Any money that goes into increasing indigenous capabilities is fine by me.

 

[aramsogo]

Congrats to India. Truly impressive.

 

[Tianshan]

It is around $2 billion.

Quite expensive I think? Tianhe-1 cost maybe 90 million.

If India can achieve 132 exaflops, may be good value for money.

 

[Yogesh]

Quite expensive I think? Tianhe-1 cost maybe 90 million.

If India can achieve 132 exaflops, may be good value for money.

May not look expensive in 2017 for which it is planned.