Chemical Forums

Please login or register.

Login with username, password and session length

Sponsored links

Pages: [1]   Go Down

Author Topic: Quantum supercalculator @ home  (Read 1605 times)

0 Members and 1 Guest are viewing this topic.

nicoM

  • Very New Member
  • *
  • Mole Snacks: +0/-0
  • Offline Offline
  • Posts: 2
Quantum supercalculator @ home
« on: November 15, 2011, 10:33:24 PM »

Dear everyone,

I have a little project that might be of some interest for researchers here.
Let me introduce myself first : I am a french master student in chemistry in Paris' Diderot University and we lack a good supercalculating infrastructure which would be versatile enough to be used daily.

So discussing on the web and around me as well I (we !) found the following idea doable: To create a cheap supercalculating device that would be accessible by SSH to send calculations onto it. The location of the computer would be home.

I order to achieve this I made some investigations on the subject. Fist I really found CUDA as being of great interest.
But, the only program I've found to do quantum chemistry is Petachem wich to work only on special graphic cards Tesla (and they are not cheap nor petachem is).
I wonder if gaussian on linux could adapted to provide the same but on a cheaper Nvidia graphic card.

I still lack of knowledge on the system's architecture. I think linux would be valuable as they are examples of domestic clusters done with it. Then that would afford a great evolution capabilities to the overall platform.

The aim of this message is to build a (ambitious) little project and provide a way for all chemist to do fast complex calculations for less than 3K€. Thus I think the little time invested will be profitable for everyone.

So if any ideas come up... ;-)

The goal is to do fast calculations on > 1K g/mol with metallic centers / metallated-porphyrin based arrays.

Nicolas.
PS: I apologize about my definitely bad skills as english writer ;-).
« Last Edit: November 16, 2011, 06:56:49 AM by Borek »
Logged

Arkcon

  • Global Moderator
  • Sr. Member
  • ***
  • Mole Snacks: +354/-113
  • Offline Offline
  • Posts: 5032
Re: Quantum supercalculator @ home
« Reply #1 on: November 16, 2011, 12:20:13 AM »

Quantum computers remain only a theoretical technology at this point in time.  So you should just leave it to science fiction stories, for now.  You appear to need more computing power then you can afford to purchase.  A common theme on the internet is "the beowulf cluster", a massive network of inexpensive machines networked together.  You should try to work with your university's IT (intelligence technology) groups.  They may be able to help, and might welcome the opportunity to work on this project for their own benefit.

http://en.wikipedia.org/wiki/Beowulf_cluster

http://fr.wikipedia.org/wiki/Cluster_Beowulf (sorry, this one has much less information)
Logged
That all depends on how reasonable we're all willing to be.  I just want my friends back, except for Cartman, you can keep him.

fledarmus

  • Chemist
  • Sr. Member
  • *
  • Mole Snacks: +198/-28
  • Offline Offline
  • Posts: 1679
Re: Quantum supercalculator @ home
« Reply #2 on: November 16, 2011, 01:54:23 AM »

I believe the subject is referring to calculators for solving quantum mechanical problems rather than calculations using quantum devices...

There are several molecular modeling service providers and virtual screening providers that have already set up this type of architecture. They are currently fairly competitive with those that have access to supercomputers. It is my understanding that this type of architecture is also used in rendering digital animation projects.
Logged

nicoM

  • Very New Member
  • *
  • Mole Snacks: +0/-0
  • Offline Offline
  • Posts: 2
Re: Quantum supercalculator @ home
« Reply #3 on: November 16, 2011, 02:18:29 AM »

Yes I was referring to build computers able to solve hartree fock equations in a short time lapse.

In fact I think the cluster part would be an evolution of the actual concept. The goal is to build the least expensive computer able to have a good yields in term of calculation difficulty/time.

@Fledarmus: Have you any sources I could refer to as an approach of this types of architectures ?
Logged

fledarmus

  • Chemist
  • Sr. Member
  • *
  • Mole Snacks: +198/-28
  • Offline Offline
  • Posts: 1679
Re: Quantum supercalculator @ home
« Reply #4 on: November 16, 2011, 02:35:59 AM »

I'm afraid that as an old bench chemist, the world of computational chemistry is somewhat beyond my expertise. However, I do have a couple of articles that might at least point you in the right directions for search terms. The first is a dedicated system using parallel computers, the second uses available hardware, and the third is an initiative to borrow computer time from volunteer computer owners.

http://www.netezza.com/documents/Computer-assisted-Drug-Discovert-w-NZ-Architecture.pdf

http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=4539327

http://drugdiscoveryathome.com/

A search for "parallel processing drug discovery" will lead you to a wide variety of other types of architecture from parallel supercomputers down to office networks, and the types of problems they solve in drug discovery.
Logged

juanrga

  • Full Member
  • ****
  • Mole Snacks: +16/-11
  • Offline Offline
  • Posts: 230
    • juanrga - sharing unified knowledge in pure and applied sciences
Re: Quantum supercalculator @ home
« Reply #5 on: November 16, 2011, 09:31:55 AM »

Not interested in computing now, but some time ago I read news about a supercomputer done with cheap game consoles PS3 and running Linux

http://www.wired.com/techbiz/it/news/2007/10/ps3_supercomputer
Logged
Sharing unified knowledge in pure and applied sciences
Pages: [1]   Go Up
 

Mitch Andre Garcia's Chemical Forums 2003-Present.

Page created in 0.047 seconds with 24 queries.