Supercomputing in India

mokoman

Senior Member
Joined
May 31, 2020
Messages
6,237
Likes
33,936
Country flag
:hmm: arn't these manufactured in fabs in taiwan ?

Nvidia and even Chinese have specialized hardware for machine learning

You can find them even in consumer level graphics cards .

Chinese been claiming their indigenous designed chips (for AI machine learning) have higher efficiency than even Nvidia for a while now.
 

shade

Senior Member
Joined
Feb 28, 2016
Messages
14,301
Likes
87,022
Country flag
I don't think any think any company in India manufacture such high end hardware ..
It's too high capex and has a small "market" which is why nobody makes this type of HW in India, like Defense tech industry basically.
In US IBM and Cray are private corp that makes Supercomputer level hardware but these are the only two private corps even globally afaik.

In Chongistan they have a "fully made in China" Supercomputer, no phoren COTS hardware involved, that is because they spent the biggest bucks imaginable on designing/manufacturing/testing their Supercomputer systems.

As with everything else, GoI past and present isn't willing to put forth the big bucks, and continuous supply of it for fully indigenous Supercomputer.
It takes a lot of money, designing IC/Motherboards, manufacturing the ICs, getting big brain people to write custom software, etc.
As of now this isn't a priority of GoI's tax funds for obvious reasons, perhaps if our economy does well in next 15-20 years we can expect a "program" to build indigenous super computer.


:hmm: arn't these manufactured in fabs in taiwan ?

Nvidia and even Chinese have specialized hardware for machine learning

You can find them even in consumer level graphics cards .

Chinese having claiming higher efficiency hardware for AI machine learning for sometime now.
For ML first are purpose built ASICS
then Nvidia GPUs because they have CUDA cores, most ML packages use CUDA cores, which is why.

But for SC usage everything is custom made, so not an issue exactly.
 

mokoman

Senior Member
Joined
May 31, 2020
Messages
6,237
Likes
33,936
Country flag
It's too high capex and has a small "market" which is why nobody makes this type of HW in India, like Defense tech industry basically.
In US IBM and Cray are private corp that makes Supercomputer level hardware but these are the only two private corps even globally afaik.

In Chongistan they have a "fully made in China" Supercomputer, no phoren COTS hardware involved, that is because they spent the biggest bucks imaginable on designing/manufacturing/testing their Supercomputer systems.

As with everything else, GoI past and present isn't willing to put forth the big bucks, and continuous supply of it for fully indigenous Supercomputer.
It takes a lot of money, designing IC/Motherboards, manufacturing the ICs, getting big brain people to write custom software, etc.
As of now this isn't a priority of GoI's tax funds for obvious reasons, perhaps if our economy does well in next 15-20 years we can expect a "program" to build indigenous super computer.




For ML first are purpose built ASICS
then Nvidia GPUs because they have CUDA cores, most ML packages use CUDA cores, which is why.

But for SC usage everything is custom made, so not an issue exactly.
Nvidia has tensor cores for AI .

CUDA is more general purpose i think.
 

SKC

Senior Member
Joined
Aug 16, 2014
Messages
8,808
Likes
30,317
Country flag
It's too high capex and has a small "market" which is why nobody makes this type of HW in India, like Defense tech industry basically.
In US IBM and Cray are private corp that makes Supercomputer level hardware but these are the only two private corps even globally afaik.

In Chongistan they have a "fully made in China" Supercomputer, no phoren COTS hardware involved, that is because they spent the biggest bucks imaginable on designing/manufacturing/testing their Supercomputer systems.

As with everything else, GoI past and present isn't willing to put forth the big bucks, and continuous supply of it for fully indigenous Supercomputer.
It takes a lot of money, designing IC/Motherboards, manufacturing the ICs, getting big brain people to write custom software, etc.
As of now this isn't a priority of GoI's tax funds for obvious reasons, perhaps if our economy does well in next 15-20 years we can expect a "program" to build indigenous super computer.




For ML first are purpose built ASICS
then Nvidia GPUs because they have CUDA cores, most ML packages use CUDA cores, which is why.

But for SC usage everything is custom made, so not an issue exactly.
Exactly! Chinese spent so much on indigenizing supercomputing hardware because they had plan to deploy Supercomputer on massive scale which justifies the expenditure.
China has maximum no of super computer in top 500 ranks. probably the maximum no of super computers in use in any 1 single country currently.
We dont have plan to use them on such massive scale. no need to make each and every hardware inside country.
AMD, IBM, NVidia, ARM all are making super computer processing units. Just design around their products.
 

NAMICA

Tihar Jail
Banned
Joined
Oct 29, 2020
Messages
1,278
Likes
6,688
Country flag
:hmm: arn't these manufactured in fabs in taiwan ?

Nvidia and even Chinese have specialized hardware for machine learning

You can find them even in consumer level graphics cards .

Chinese been claiming their indigenous designed chips (for AI machine learning) have higher efficiency than even Nvidia for a while now.
Lol intel is having most powerful AI based chip.
 

shade

Senior Member
Joined
Feb 28, 2016
Messages
14,301
Likes
87,022
Country flag
Are you sure ??

Most i think are still running Nvidia or Intel .

They are still 2 generations behind intel.

The Atmanirbhar Cheen "Sunway TaihuLight" ranks Numbah 4 in this TOP500 list, above it are IBM POWER architecture based SCs with Nvidia GPUs and NUMBAH WAN is Japan with some Fujitsu made custom ARM processor SC.

The Sunway SC doesn't have any GPU/Accelerator cores.

below the ping pong SC all others are Intel Xeon, IBM POWER, or Cray custom architecture based
GPU is always of Nvidia, Tesla or Ampere Architecture.

 

mokoman

Senior Member
Joined
May 31, 2020
Messages
6,237
Likes
33,936
Country flag

The Atmanirbhar Cheen "Sunway TaihuLight" ranks Numbah 4 in this TOP500 list, above it are IBM POWER architecture based SCs with Nvidia GPUs and NUMBAH WAN is Japan with some Fujitsu made custom ARM processor SC.

The Sunway SC doesn't have any GPU/Accelerator cores.

below the ping pong SC all others are Intel Xeon, IBM POWER, or Cray custom architecture based
GPU is always of Nvidia, Tesla or Ampere Architecture.

Deleted my comment . o_O

They are trying to catch up in consumer level cpus too.

Reached intel i3 level performance last year , while intel released i11 this year.

Fuckers will soon catch up there too.
 

Indx TechStyle

Kitty mod
Mod
Joined
Apr 29, 2015
Messages
18,277
Likes
56,182
Country flag
:hmm: arn't these manufactured in fabs in taiwan ?

Nvidia and even Chinese have specialized hardware for machine learning

You can find them even in consumer level graphics cards .

Chinese been claiming their indigenous designed chips (for AI machine learning) have higher efficiency than even Nvidia for a while now.
Will take a while before general chip fabrication takes place and matures in India. Here while is to be expressed in many decades.
 

HariPrasad-1

Senior Member
Joined
Jan 7, 2016
Messages
9,602
Likes
21,068
Country flag
Don't know much Brother , CDAC made progress in designing an EXA scale Mechine . May be they are trying to develop own processor for super computers . These machines are energy hungry and eat electricity in 50-100 MW . Cooling requirements are far more than what an average super Computer want . Don't know the progress in these areas . Meanwhile China developed a prototype of ExaScale machine in 2018 .
Somebody told that it will use flexible bus to to increase the speed and reduce electricity consumption. I do not know what exactly it means though I could understand a bit.
 

sorcerer

Senior Member
Joined
Apr 13, 2013
Messages
26,919
Likes
98,471
Country flag
India’s first Quantum Computer Simulator toolkit launched | Dehradun News - Times of India
Tapan Susheel / TNN / Aug 28, 2021, 04:01 IST

2 minutes


Roorkee: Rajeev Chandrasekhar, Union minister of electronics and information technology, launched the country’s first ‘Quantum Computer Simulator (QSim) Toolkit’ on Friday. According to the ministry, the toolkit will enable researchers and students to research quantum computing cost-effectively.

 

sorcerer

Senior Member
Joined
Apr 13, 2013
Messages
26,919
Likes
98,471
Country flag
India adds another supercomputer at Mohali under national mission


3 minutes


Under the National Supercomputing Mission, India has increased its capability by adding a 650-teraflops system at the National Agri-Food Biotechnology Institute (NABI) in Mohali. The system was inaugurated by the minister of state for science and technology Jitendra Singh on Tuesday. The facility has been established at the cost of ₹20 crores.
 

sorcerer

Senior Member
Joined
Apr 13, 2013
Messages
26,919
Likes
98,471
Country flag
Petascale Supercomputer “PARAM Ganga” established at IIT Roorkee under National Supercomputing Mission

Availability of supercomputer to accelerate R&D activities in multidisciplinary domains of science and engineering


Posted On: 08 MAR 2022 12:06PM by PIB Delhi



The National Supercomputing Mission (NSM) which is being steered jointly by Ministry of Electronics & Information Technology (MeiTY) and the Department of Science and Technology (DST) and implemented by Centre for Development of Advanced Computing (C-DAC) and Indian Institute of Science (IISc), Bangalore, has progressed significantly. The four major pillars of the NSM, namely, Infrastructure, Applications, R&D, HRD, have been functioning efficiently to realize the goal of developing indigenous supercomputing eco system of the nation.


C-DAC has been entrusted the responsibility to design, development, deployment and commissioning of the supercomputing systems under the build approach of Mission. The Mission plans to build and deploy 24 facilities with cumulative compute power of more than 64 Petaflops. Till now C-DAC has deployed 11 systems at IISc, IITs, IISER Pune, JNCASR, NABI-Mohali and C-DAC under NSM Phase-1 and Phase-2 with a cumulative compute power of more than 20 Petaflops. Total 36,00,000 computational jobs have been successfully completed by around 3600 researchers across the nation on the NSM systems to date. The supercomputer infrastructure installed at various Institutes across the country have helped the R&D community to achieve major milestones, objectives and products for scientific and societal applications.


Under the build approach, C-DAC is building an indigenous supercomputing ecosystem in a phased manner, which is leading to indigenously designed and manufactured supercomputers. It has designed and developed a compute server “Rudra” and high-speed interconnect “Trinetra” which are the major sub-assemblies required for supercomputers.


Some of the large-scale applications which are being developed under NSM include the following.


  • NSM Platform for Genomics and Drug Discovery.
  • Urban Modelling: Science Based Decision Support Framework to Address Urban Environment Issues (Meteorology, Hydrology, Air Quality).
  • Flood Early Warning and Prediction System for River Basins of India.
  • HPC Software Suite for Seismic Imaging to aid Oil and Gas Exploration.
  • MPPLAB: Telecom Network Optimization.

As part of its tireless journey of success, NSM has now deployed “PARAM Ganga”, a supercomputer at IIT Roorkee, with a supercomputing capacity of 1.66 Petaflops. The system is designed and commissioned by C-DAC under Phase 2 of the build approach of the NSM. Substantial components utilized to build this system are manufactured and assembled within India along with an indigenous software stack developed by C-DAC, which is a step towards the Make in India initiative of the Government. Availability of such a supercomputer will accelerate the research and development activities in multidisciplinary domains of science and engineering with a focus to provide computational power to user community of IIT Roorkee and neighbouring academic institutions. This national Supercomputing Facility was inaugurated on March 07, 2022 by Shri B.V.R. Mohan Reddy, Chairman, Board of Governors, IIT Roorkee in gracious presence of Prof A K Chaturvedi, Director, IIT Roorkee, Dr Hemant Darbari, Mission Director, NSM, Shri Naveen Kumar, Scientist D, NSM Program Division, MeitY, Prof Manoranjan Parida, Deputy Director, IIT Roorkee, Shri S A Kumar, Advisor NSM, Shri Sanjay Wandhekar, Senior Director, C-DAC, Pune and Convener- NSM Expert Group on Infrastructure, Dr Sivaji Chadaram, Scientist - F, DST along with senior officials from MeitY, DST, IIT Roorkee and C-DAC.
 

Abdus Salem killed

New Member
Joined
Mar 10, 2021
Messages
4,103
Likes
15,759
European pre exascale computer lumi
Our equivalent to this is param mahapadama
But no news on that yet
And then thier will be param mahapadama 2 and European Leonardo
I hope we achieve pre exascale in 2023
We will be in top 5 country if we do so
 

Chinmoy

Senior Member
Joined
Aug 12, 2015
Messages
8,734
Likes
22,726
Country flag
European pre exascale computer lumi
Our equivalent to this is param mahapadama
But no news on that yet
And then thier will be param mahapadama 2 and European Leonardo
I hope we achieve pre exascale in 2023
We will be in top 5 country if we do so
PARAM SHANKH to be precise. To be launched in 2024.
 

Concard

Senior Member
Joined
Jun 18, 2020
Messages
1,159
Likes
8,962
Country flag
First launch param mahapadama pls we need pre exascale before exascale it would put us comparable to Japan fugaku


I know what param shank is but thier is literally no updates from goi
Fugaku is different. It is based on ARM architecture and there are no GPU's which were used. I heard of our exascale program only today on this thread. It would be a huge boost to our scientific capability if we bring it online by 2024.
 
  • Like
Reactions: SKC

Abdus Salem killed

New Member
Joined
Mar 10, 2021
Messages
4,103
Likes
15,759
Fugaku is different. It is based on ARM architecture and there are no GPU's which were used. I heard of our exascale program only today on this thread. It would be a huge boost to our scientific capability if we bring it online by 2024.
Yes we are at 2023 we were supposed to get param mahapadama pre exascale this year it would have been comparable to lumi but no dice till now
 

Latest Replies

Global Defence

New threads

Articles

Top