What if every university had a computing cluster capable of delivering a petaflop of performance (equivalent to about 20,000 typical laptops)? Better yet, what if every university department – biology, chemistry, physics, and others – had one?

I’m certain that this would fundamentally increase the pace of scientific discovery around the world.

“We will award early access to this
new Kepler-based Tesla GPU to the three best
proposals…”

Computing is now the third pillar of science, along with theory and experimentation. Giving every researcher access to a dedicated high performance computing facility could lead to breakthroughs in renewable energy, climate prediction, the development of lighter and tougher materials, or even pave the way to find cures for some of society’s most devastating and persistent diseases.

By the end of 2012, NVIDIA will launch a Tesla GPU based on our new Kepler architecture that will enable any university in the world to build a petaflop supercomputer that will fit into nearly any university’s data center and budget.

To support this launch, we are inviting proposals from researchers around the world on what you would do if you had access to a dedicated petaflop supercomputer. We will award exclusive early access to this new Kepler-based Tesla GPU to the three best proposals we receive.

What types of studies would you conduct?  Which scientific problems would you tackle?

To enter this contest, post a detailed answer to the above questions in the comments section. Be sure to include a link to your research, if available. Or if you prefer, email your submission to cuda@NVIDIA.com.  You should also include your contact information (at least email address and affiliation) so that we can reach you if you win.

NVIDIA will use the proposals to highlight the potential of petaflop supercomputing, so please do not submit anything you consider confidential.

We will announce the top three entries at NVIDIA’s upcoming GPU Technology Conference (GTC 2012) in May, and notify them directly.  Also, look for a blog post announcing the winning entries after GTC.

You can find the complete list of contest rules here.

Good luck!

  • Sumit Gupta

    Sander

    This sounds like very interesting work.  Can you make a specific proposal on the work you would do.    We are looking for detailed research proposals.

    Thanks
    Sumit

  • Sumit Gupta

    Thanks for your submission Anders.    Have you scaled your work to multiple servers with GPUs in them?    A Petaflop supercomputer with Tesla Kepler GPUs will have hundreds of servers with hundreds of GPUs in them.
    Sumit

  • http://profile.yahoo.com/NU5RPC5UCQOSHAKJ343WPTUTWM h4x354x0r

    I would use it to compute and simulate all the possible combinations, and effects on the human cannabinoid receptor system, of the nearly 100 different cannabinoids found in pot plants to develop new, safer medicines. 

  • http://www.facebook.com/people/Caleb-Sarah-Riley/1524042211 Caleb ‘Sarah’ Riley

    I would first use it to research subliminal imaging and the idea of somehow inputting information into the human brain without having to learn it WITHOUT mentally scarring the patient using ‘encoded images’ or another subliminal means. I would use it to house and analyze large quantities of encoded data. Then program the computer to analyze the data and make connections between the data inputted and other material. When activated the data would be encoded into inconspicious images and they displayed in a massive rush of information which would facilitate subliminal retention when viewed by those receptive to it. Then use the computer to build and create blueprints for a smaller computer to hold all this information and host the program. In short, I would attempt to create an ‘Intersect’ computer, as stupid as it sounds.

  • anjali dadhe

    render a life like world with more than a trillion floating particles to simulate wind and particle dynamics for next gen games .

  • Eduardo Viecili

    I would use to run an algorithm to manage all money transfers allowing the government to know how each individual/company spend their money. And with these information tax with precision, and detecting if a fraud or an irregularity was made. Also could search for patterns to generate a better economic model so the central bank could be more precise at the interventions. Detec which areas are rising and need more investments. Manage resources,labor and logistics and other more functions to improve the country to a high desired level of living standard. Predict and detec risks and solutions for this advanced society.

    Other use is for breaking the public cryptography keys to take correlated informations over the internet, because breaking cryptography is easy with a petaflop.

    Name: Eduardo Brendler Viecili
    email1: edu@evless.com
    email2: whitezur@gmail.com
    University: UDESC-BR (www.joinville.udesc.br)

  • http://twitter.com/SgtQuackers Brian

    Try to take over the world…..ahahahahahahahahahahhaah (evil laugh)

  • Marshall Barth

    PTL!!! Set up the most complex Christian ministry, covering the gambit from simple Bible studies to one of the most intensive disaster relief operations ever dreamed of :{)))

  • http://www.facebook.com/profile.php?id=726731978 Carlos David

    I make a really useful web page server, and the same time, same computer to play videogames maxed detail that  a great screen and a nvidia gpu can offer, tand hear music =D not to mentions decodig my videos and phots and make them 3D in my huge 240hz 62″ screen. OOh lord i need to change my pants…

  • http://www.facebook.com/bozmanbeyond Robert Fairchild

    A Holodeck is all I can think of, bringing what was once a fantasy to life.  

  • http://twitter.com/mund0x Ed

    I would download a car, what else?

  • Sumit Gupta

    Thanks for your submission Eduardo.    

  • http://www.facebook.com/cameron.barron Cameron Barron

    i would game on it!!!! and also do my school work

  • Maciej Świechowski

    I am already a scientist working in area of AI in Games.
    However, I don’t create algorithms aimed at playing specific games (like Deep Blue and chess – you know) but playing any types of games instead.
    What my algorithm does in runtime is to:
    - consume rules of the game (written in dedicated first-order logic language)
    - learn how to play the game effectively (WITHOUT any human intervention)
    - play the game

    The idea is to develop intelligent mechanism of knowledge representation, machine learning and so on. “Games” are only testbed – games can refer to modelling stock, voting behaviours, natural evolution… and so on. As long as certain phenomena can be describe as set of game rules it falls under my project.

    I would use Petaflop supercomputer for extensive internal simulations of games in order to reach as many states as possible.
    m.swiechowski@mini.pw.edu.pl

  • http://www.facebook.com/people/Marcel-Hetzel/100000716792296 Marcel Hetzel

     RENDER THAT DAMN THING. I would render – infinite – and trying to get all from C4D whats possible – blowing up the limits of reality!

    In thanks of such a prize I would make a cool spot for nvidia for free :-) – I promise!

  • Sumit Gupta

    Thanks for the great responses here and at cuda@nvidia.com    Please keep them coming.    We are looking for researchers who would transform their research by having dedicated access to such a powerful supercomputer, instead of having to apply for access to one of the shared supercomputers.
     

  • Sumit Gupta

    Thanks Koji.   I really wish we could learn more about global warming and similar environmental problems.   Have you seen the work going at Japan’s NIES (National Institute of Environmental Studies) on the use of GPUs for C02 modeling?
    http://www.iemss.org/iemss2010/papers/S00/S.00.14.Using%20NVIDIA%20GPU%20for%20Modelling%20the%20Lagrangian%20Particle%20Dispersion%20in%20the%20Atmosphere%20-%20JIYE%20ZENG.pdf 
    This is for the GOSAT project http://www.gosat.nies.go.jp/index_e.html

  • Aron Miller

    IBM’s Watson technology merged with Google Search would be a nice future service. Watson helps with elaborating the real meaning of the question, and then the Google subsystem does the more exhaustive search over the volume of the internet. In order for such a system to be valuable it may need to be as fast as possible otherwise the human will do the work of the Watson component manually (as they do now). A Petaflop at 12 times Watson’s hardware may be able to give seemingly real-time results at Jeopardy level performance allowing for an interactive dialog.

  • http://www.facebook.com/MADBOOT Michael Dillon

    I would devote my time to developing an AI,that could do all my work for me :)

  • Sumit Gupta

    Aron
    Don’t know if you read the book “Hitchhiker’s Guide to the Galaxy”, but this really super-smart GPU-accelerated Watson+Google might come up with the answer to the ultimate question of life, the universe, and everything … answer is of course “42″, just don’t know what the question is ;)
    http://en.wikipedia.org/wiki/Phrases_from_The_Hitchhiker's_Guide_to_the_Galaxy 

  • http://www.facebook.com/people/Brian-Prince/1495725840 Brian Prince

    Funny Show, but I don’t think i’ll ever waste my time calculating all the way back to the ‘big bang’. God is no less realistic an answer then that “beginning time Singularity” that you append unto big bang to counter the God theory.

  • http://www.facebook.com/people/Brian-Prince/1495725840 Brian Prince

    Will it play WOW ???

  • henna_programer

    If I had any idea how to use petaflop machine, I would definitely NOT participate in this contest.  Ideas can be very valuable these days and cost much more than NVIDIA propose to pay for them in this contest.  This contest is, in reality, a scavenging for new ideas.

  • Sumit Gupta

    I appreciate your viewpoint.   This contest is designed for researchers who believe that having access to fast supercomputer will revolutionalize and fundamentally speed up the pace of their research.  Or enable them to simulate things that they couldn’t before.

  • http://www.facebook.com/erkaninho Erkan Tairi

    If I had a supercomputer capable of delivering petaflop of performance probably I would do two things with it. First of all I would do simulations and calculations about quantum computing and quantum cryptography, in order to create a compact quantum computer and unbreakable quantum cryptography. Therefore, changing the world how people do computing, and introduce them the new era of quantum computers. Who wouldn’t like to have an unbelievable fast and secure computer?

    As my second research I would focus on string theory, which is somehow related to quantum mechanics. It is at the moment closest theory to describe our universe from the smallest particles to the enormously big galaxies. The only problem is that currently it is just a theory. It works with the miniature strings of energy, about billions of times smaller than an atom, which make them unable to be seen with the experiments. And in my opinion the only solution to play around with the strings and experiment with them is to use the power of simulation and the third pillar of science, “computing”! I think we have waited long enough for the mystery of the universe. It is time to unify our universe under a single expression (formula). Who wouldn’t like to discover the universe?

    Name: Erkan Tairi
    E-mail: erkan.tairi@gmail.com
    Company: Progex Systems
    Department: Progex Research Center

  • Tyler Gubb

    GPU computing would be beneficial in solving the wave-functions for single crystalline (nano-domain) particles, and to simulate their synthesis and behavior.  These highly engineered materials are finding numerous applications as of late.  Noncrystalline ceria (cerium-oxide) may be used for enriching fuel mixtures for ‘perfect’ burn and delivering drugs across the blood-brain barrier to effectively eliminate the cause and effects of Alzheimer’s disease.  Nano-domains may be used (via quantum dotting) to produce highly efficient materials which convert light into electric current and visa-versa.  These examples of highly engineered materials only represent the tip of where the future of materials science is headed.  However that future starts with a powerful laboratory computer performing the theoretical calculations and simulations needed to engineer the product.  That is where I would like to focus my future research and thesis work.

  • Tyler Gubb

    GPU computing would be beneficial in solving the wave-functions for single crystalline (nano-domain) particles, and to simulate their synthesis and behavior.  These highly engineered materials are finding numerous applications as of late.  Noncrystalline ceria (cerium-oxide) may be used for enriching fuel mixtures for ‘perfect’ burn and delivering drugs across the blood-brain barrier to effectively eliminate the cause and effects of Alzheimer’s disease.  Nano-domains may be used (via quantum dotting) to produce highly efficient materials which convert light into electric current and visa-versa.  These examples of highly engineered materials only represent the tip of where the future of materials science is headed.  However that future starts with a powerful laboratory computer performing the theoretical calculations and simulations needed to engineer the product.  That is where I would like to focus my future research and thesis work.
    Tyler A Gubb
    tag3@alfred.edu
    Alfred University

  • http://www.facebook.com/people/Royce-Wegner/510528245 Royce Wegner

    I would design a %100 physics based world, where everything is rigid bodies,soft bodies,cloth,fluid dynamics,particles & more.The full physics simulated world could be used in a variety of businesses. This is has never been achieved yet as there has not been enough computing power to drive massive amounts of physics simulations at once.Still …today almost every simulation is static.This would be a universal idea for a vast amount of applications, for both software and hardware.Even a petaflop might not be enough power to drive a simulation on this big of a scale.

  • Anders Eklund

    That would be nice, but my department can’t afford that.

  • Sumit Gupta

    These are great entries.   Please keep them coming.   Submissions are still open.

  • Mario Antonioletti

    We are currently working with oncologists at the Western General Hospital in Edinburgh, Scotland who are trying to do analysis of CAT images to identify tumour regions to irradiate. They currently have a 32 node computer cluster they are trying to use with a MATLAB bit of code but do not have GPU resources for this kind of analysis as far as I know. It would be great to provide them with this resource to develop their algorithms and code. The ideal result would be to develop algorithms that work in real time so that the prerequisite image analysis can be done as the patient is being irradiated. The fact that the analysis could be done in house (it is not easy – or one might even say possible – to export data to external computational resources due to the patient confidentiality of the data) and in real time would be a major benefit to us all (if we were unlucky enough to develop a cancer). I think it would be cool to give them access to such an impressive resource.

  • Lorenzo Rovigatti

    I am a Ph.D. student in physics and I would use it to study self-assembly of DNA structures into novel materials like equilibrium gels (i.e. material that can support stress but are very lightweight) or re-entrant gels and crystals (i.e. materials that return in a liquid state both on heating and on cooling). In particular, I am one of the main developers of oxDNA (http://dna.physics.ox.ac.uk), which is still in a very early stage of development but, as an unsupported feature, it can run on NVIDIA GPUs (see for example http://arxiv.org/abs/1204.0985). We have a nice >25x speed-up over a single core, but we cannot simulate very large systems on TESLA GPUs yet since the DNA coarse-grained model employed is very complicated and requires a lot of computational power.
    At the moment I am trying to study the phase diagram (in particular, the region of the phase diagram where the system phase-separates) of DNA constructs which self-assemble into a very low-dense network. I cannot afford to simulate more than 100 of these constructs (which means almost 20k nucleotides, by no means a small number) on a single TESLA GPU. While we would love to add multi-GPU support, even an increase of the performances on single GPUs, due to the new architecture, would mean a lot in terms of both research time and project feasibility.

    Name: Lorenzo Rovigatti
    Institutional email: lorenzo.rovigatti@uniroma1.it
    Website: http://dna.physics.ox.ac.uk/

  • rxeno

    What
    a thought, yes Kepler Tesla indeed. Well, I would first continue my work
    creating 3D reality live events. I believe that the media sea change will
    precipitate in remote attendance of live events to grow the performance to
    attendee ratio. (This is already happening but supercomputing will afford the
    momentum to increase the delivery quotient).

    I
    would also begine my harmonic resonance work in earnest. I have studied precisely
    the determinations and now hidden equations Nikola Tesla utilized to transmit
    free energy. I believe that just like all things, energy is at the brink of
    breaking free from it’s captors and will soon become free to all the world. In
    so doing, production and use will grow exponentially creating a better world
    for everyone. Even those that would fight to prevent its coming.

    With
    these money based activities would come billions of dollars of revenue which I
    would use to better the plight of the lesser in the world. Food, clothing ,
    shelter, medical, all of these areas would profit from my work, the outcome of
    my being allowed the power of Nvidia’s Fine new Kepler Tesla Computer.

    My
    past work, research and business background support my forward thinking plans.
    And it is true that I will achieve my goals regardless of the platform (I
    currently have approximately 30 computers and servers doing my work) but with
    this most excellent of newest pieces of futureware I would come to my solutions
    much much sooner. I would suspect, as well, much more efficiently and
    productively. All systems change through time/use/activity. In my case, the
    sooner the better.

    Thank you for your time and the opportunity to server mankind.
    What
    a thought, yes Kepler Tesla indeed. Well, I would first continue my work
    creating 3D reality live events. I believe that the media sea change will
    precipitate in remote attendance of live events to grow the performance to
    attendee ratio. (This is already happening but supercomputing will afford the
    momentum to increase the delivery quotient).

    I
    would also begine my harmonic resonance work in earnest. I have studied precisely
    the determinations and now hidden equations Nikola Tesla utilized to transmit
    free energy. I believe that just like all things, energy is at the brink of
    breaking free from it’s captors and will soon become free to all the world. In
    so doing, production and use will grow exponentially creating a better world
    for everyone. Even those that would fight to prevent its coming.

    With
    these money based activities would come billions of dollars of revenue which I
    would use to better the plight of the lesser in the world. Food, clothing ,
    shelter, medical, all of these areas would profit from my work, the outcome of
    my being allowed the power of Nvidia’s Fine new Kepler Tesla Computer.

    My
    past work, research and business background support my forward thinking plans.
    And it is true that I will achieve my goals regardless of the platform (I
    currently have approximately 30 computers and servers doing my work) but with
    this most excellent of newest pieces of futureware I would come to my solutions
    much much sooner. I would suspect, as well, much more efficiently and
    productively. All systems change through time/use/activity. In my case, the
    sooner the better.

    Thank you for your time and the opportunity to server mankind.

    Robert Xeno
    Radio-X Technologies
    Earth Hertz Productions, LLC
    Nashville, TN USA
    radiox@radiox.net

  • http://voxelium.wordpress.com/ Attila T. Áfra

    I am a PhD student in computer graphics, working on real-time massive model visualization using ray tracing. My approach is currently capable of interactively rendering very complex scenes consisting of hundreds of millions of primitives, with global illumination, on a single desktop PC: http://voxelium.wordpress.com/category/graphics/voxlod/

    If I had access to a petaflop supercomputer, I would focus my research on exploiting its power to render on holographic displays like HoloVizio (www.holografika.com) using ray tracing. Such light field displays have very high resolution (more than 60 Mpixels), which dramatically increases the number of rays that must be shot. Rendering high-quality, noise-free images using physically-based Monte Carlo ray tracing methods (like path tracing) is a very challenging problem even in case of 2D visualization. A petaflop supercomputer could render photorealistic images on light field displays in real-time.

    The research would be based on my current out-of-core path tracing algorithm. Although ray tracing is inherently parallel, incoherent memory access patterns due to random rays could compromise the scalability of the algorithm. It is important to note that the scene would not be able to fit into the memory of a single GPU. Thus, optimally exploiting the memory hierarchy of the system would be one of the primary challenges of this research.

    The results of this work could have important applications in CAD, medical and scientific visualization, entertainment, and other domains.

    Attila Tamás Áfra
    attila.afra@gmail.com
    Budapest University of Technology and Economics, Hungary
    Babeș-Bolyai University, Romania

  • http://www.facebook.com/rdavidbarr David Barr

    I would use it to help solve some of the more complex chess moves.  Not so much in brute force but natural thinking. 

    Thanks,
    David Barr

  • http://www.facebook.com/profile.php?id=100001409064577 Sander Peters

     Dear Sumit,

    As a second year student Industrial Engineering, I am working hard on my simulation skills and it will take a year or two to fully develop them by following courses at the Technical University of Eindhoven. Never the less I have a research proposal for simulation, if I had an super computer I could conduct that in two or three years.

    In today’s industries more and more chemical processing is done by companies. This processing brings along with the benefits of those new techniques and materials also danger of a major chemical disaster. Some people can remember probably the biggest industrial disaster in Bhopal, India. Where 3000 people died of an accident at the Union Carbide chemical production plant. Because highly intoxicating gasses where released in the atmosphere. Today local communities are facing the choice between economic profit and possible danger. Also industries need to worry about there safety procedures, and emergency services need to know how to act at those large scale chemical disasters.

    In order to help both of them I would create a visual 3D simulation of all the possible disasters that could happen if you build a chemical plant at that site. Fires, toxic clouds and even terrorist attacks could be simulated and the outcome of undertaken actions could be recorded. In this way emergency services could better decide how to under take action and local communities can decide better if they want that chemical plan so close to an inhabited area.

    This kind of simulation could have prevented the major environmental disaster at Chemi-Pack, Moerdijk, The Netherlands in 2011. Where fire-fighters tried to extinguish a huge chemical fire with water, instead of foam. With this simulation the outcome of working with water instead of foam could be proven before the fire, instead of concluding it in a report afterwards.

    As an expansion to this simulation there can be other safety issues modelled. Like the security at the Olympic Games or at another huge event. This all could be possible with a super computer with such capacity. Although it is a long term vision, considering those 3 years it could be a giant improvement considering safety.

    With kind regard,
    Sander Peters
    Technical University of Eindhoven
    Department of Industrial Engineering and Innovation Sciences
    E-mail: S.P.F.Peters@student.tue.nl

  • Sumit Gupta

    Please get your submissions in today.   Last call – submission deadline is tonight (see rules above to get exact details).

  • Ipbeil

    I’m a simulation specialist at the medical school at the University of Louisville. With a petaflop computer, I would analyze hyper-slow motion videos of surgeons performing detailed endoscopic and laparoscopic procedures to break the processes down into keyframe gestures and repeatable patterns. I would then build three dimensional gesture models of these, and use evolutionary algorithms to optimize workflow. I would then use these optimized gestures to create a feedback training system for surgery residents, using motion capture technology such as the Microsoft Kinect. Surgical trainees would be presented with three dimensional puzzles that would require ever-increasing speed and accuracy to the optimized gestures to successfully complete. Data gathered from multiple tests of subject performance against a control group would be used to further adjust the simulation, and as foundational data for creating an at-home training system utilizing consumer level personal computers and video game consoles to heavily supplement the current training model of lab-only practice with expensive specialized equipment.

    Ian Beilman

    Simulation Technician

    Paris Simulation Center

    University of Louisville School of Medicine

    Ipbeil01@louisville.edu

  • http://twitter.com/fengji Feng

    I am a computer science PhD student. I am more of a systems guy who is thinking about providing good tools/libraries for application developers for easier working on these GPU-powered machines. And one of my work is about having transparent data movement across CPU and GPUs through an automatic runtime system. We have built such a system, yet we’ve found its performance relies on the performance of GPU cores as much as the memory subsystem. So if there is such a petaflops system with Kepler-based GPUs, I would definitely want to try it on the Kepler-based Tesla cards, which are said to have improved memory bandwidth. These experiments, when compared against that on the previous Fermi architecture as well as other GPUs, may shed some light how such a data framework should be build as a general solution.
    Feng JiDepartment of Computer ScienceNC State Universityfji@ncsu.edu

  • Daniel Nelson

    why not attempt a shackled ai?

  • http://www.facebook.com/profile.php?id=100001820644129 Shabayer Shaukat

    1:Every thing should be in 3d and like a movie because a student learns more by a animated film.so for example if we r studying the structure of DNA then through nvidia,s graphics power the teachers should create a 3d model of DNA and then teach it to the students.like this different 3d animation can b made and then teach to the students.with the passage of time the teaching methods should b upgraded with the other developments.i would like to tackle the scientific problems like the full life cycle of a mosquito which should be teach by a 3d model.

  • VIDYA BHUSHAN SINGH

    I will use that design and apply new PRAM algorithms to find answer to some of the toughest questions, like solving the RSA Challenges, DNA sequencing, etc.

  • Josh Heystek

    I would use it to try to figure out women. I’m just not quite sure if this would quite have the horse power for that. =)

  • Anders Eklund

    When will the winners be announced?

  • http://www.facebook.com/profile.php?id=1242772565 Logan Hunter Strzepka

    I would DEFINITELY do some bitcoin mining to put the guys at the AMD camp to shame!

  • Fluidic Ice

    I’d attempt making “the game of life” I designed in a blog which would require a supercomputer or better to actually run and create this “game”: 
    http://www.fluidicice.com/1/post/2011/11/the-game-of-life.html

  • Mihail Bebrovski

    Lol I’d render a bunch of simulations with 3DS Max at the highest render settings possible