Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I want to buy a $40000 1 tera flop cluster to keep it in a room. What are the standard configurations?

Cluster is supposed to do molecular dynamics simulations on biological systems.

I'm proposed a 4 pc with 8 cores each by the selling company I'm deadling with. It looks like I also need infiniband. Does some one has an experience -- what phisical memory should I buy etc?

I know things change very quickly... Still there might be a point or two to state.

Edit:

Custer is supposed to run linux, and gromacs in it.

share|improve this question

closed as not constructive by ewwhite, Greg Askew, ThatGraemeGuy, SvW, HopelessN00b Sep 28 '12 at 11:05

As it currently stands, this question is not a good fit for our Q&A format. We expect answers to be supported by facts, references, or expertise, but this question will likely solicit debate, arguments, polling, or extended discussion. If you feel that this question can be improved and possibly reopened, visit the help center for guidance.If this question can be reworded to fit the rules in the help center, please edit the question.

1  
This is quite a specific question so I would be interested if anybody has any answers to this. –  tombull89 Sep 28 '12 at 9:05
    
@tombull89: Please up vote -- other wise the question might be closed. –  Adobe Sep 28 '12 at 9:10
    
Question score doesn't make a difference to the question being closed. –  tombull89 Sep 28 '12 at 9:15
    
@tombull89: still there's a certain psycology... –  Adobe Sep 28 '12 at 9:18
    
Please don't crosspost between sites. If you already have a system recommended by a dealer, or any system at all, there should be a list of certified ram that should work with that specific models, and requirements. This really feels too open ended to trally answer. –  Journeyman Geek Sep 28 '12 at 9:21

1 Answer 1

A single AMD FirePro W9000 GPGPU can deliver a teraflop, they're ~$3.5-4k - could you not just but one of those and stick in it in a generic PCIe 3.0 based machine? You could then give me the remaining £35k and call that consultancy costs :)

While that IS one answer you need to know what APIs are to be used and what clustering mechanism will be needed before anyone can answer this. If you really do want to go down this route you're going to need a LOT more knowledge before we can give you any better answers.

EDIT - thanks for letting us know about the GROMACS thing, reading their documentation it can actually use CUDA so a top-end NVidia TESLA M2090 card (or more) should be able to give you what you need without having to resort to multi-system clustering (which is the complex bit). So get a quote for something like a top-end HP/IBM/Dell workstation with one or two of those in it (you'll need BIG power-supplies to handle two) and you'll be within budget.

share|improve this answer
    
Please, give me some readable sources. There're gigabytes of text information out there. –  Adobe Sep 28 '12 at 9:29
    
In gromacs I'm going to use Martini force field -- and this one is not supported by CUDAs. Reference. –  Adobe Sep 28 '12 at 10:12

Not the answer you're looking for? Browse other questions tagged or ask your own question.