What Components are Needed for Building a High-Performance Computer?

AI Thread Summary
Building a high-performance computer for intensive calculations with a budget of $4000 is feasible. Key components recommended include an NVIDIA video card, which is essential for applications benefiting from parallel processing, particularly for CUDA programming. A dual Opteron or Xeon processor setup is suggested for maximizing processing power, with eight processor cores being optimal. It is advisable to include at least 8GB of DDR2 RAM and a robust power supply rated between 1000-1500 watts. For storage, a RAID 5 array with multiple hard drives is recommended for speed rather than size. An alternative approach is to consider a Beowulf cluster using used computers with Linux, which may offer cost efficiency without the need for a dedicated graphics card. Overall, the focus should be on maximizing computational capabilities while ensuring compatibility with various operating systems and software.
leon1127
Messages
484
Reaction score
0
Hi guys,

Now I want to build a computer for intensive calculation. I am expecting to spend somewhere 4000 dollars just for machine. The only thing I must need is nvidia video card. hhd size isn't important but the speed is. Do you guys have any suggestion?
 
Computer science news on Phys.org
robphy said:

Very interesting link. However I am a mathematician not a computer builder. I do find building computer entertaining but I looking for something practical and ready to be used for numerical programming.
 
leon1127 said:
Hi guys,

Now I want to build a computer for intensive calculation. I am expecting to spend somewhere 4000 dollars just for machine. The only thing I must need is nvidia video card. hhd size isn't important but the speed is. Do you guys have any suggestion?

http://www.pcworld.com/article/id,119273-page,1/article.html
 
Last edited by a moderator:
Last edited by a moderator:
Yes, with $4000, you should be able to build something nice.

With that kind of money, you should probably look at an ASUS board with dual Opteron or Xeon processors. That will give you eight processor cores. Then, you should load it up with eight GB of DDR 2 RAM and one or two Nvidia cards. Buy a 1000-1500 watt power supply.

Also, you should probably have money left over to make a five or six hard drive RAID 5 array. Be sure to buy a UPS.

For about $4000 you can build a very powerful computer, as I mentioned, capable of driving four monitors (which can be used for CUDA if necessary), with eight processor cores.

And, as an added bonus, you could choose to run 64 Bit XP, Vista, or 2008, or most Unix systems, so it should be good for whatever kind of software and operating system you need the workstation for.

Your other option, if it is suitable and you have the technical expertise, is to buy some used computers, install Linux on them, and run a Beowulf cluster, which might be more cost efficient than a single computer, and would not require a graphics card.
 
leon1127 said:
Now I want to build a computer for intensive calculation. I am expecting to spend somewhere 4000 dollars just for machine. The only thing I must need is nvidia video card. hhd size isn't important but the speed is. Do you guys have any suggestion?

Why nVidia?

Does your application benefit from parallelism?
 
Nvidia does currently have more perks when it comes to incorporating the GPU for computational power...

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

Nvidia blows out Moore’s Law with fresh Tesla
Nvidia pitches its Tesla hardware as a magical solution for the world’s toughest computing problems. Just move your code that runs well across many processors over to the Tesla boards, and Shazam!. You enjoy sometimes 400 per cent improvements in overall performance.

Despite such mind-blowing increases in horsepower, Tesla continues to occupy a space that one could characterize as ultra-niche. Only the brave few have navigated Nvidia’s CUDA programming apparatus to tweak their code for the general purpose graphics processors inside of the Tesla systems.

That ultra-niche, however, may grow into a niche over the coming year thanks to the introduction of more powerful Tesla systems.
http://www.reghardware.co.uk/2008/06/16/nvidia_tesla_10/
 
Last edited:
I am going to write apps with CUDA. This is the whole idea
 
  • #10
Get a big case, a 1500 watt power supply (try ultra), and a motherboard with 4 PCI express x16 slots, and throw in four Nvidia cards (or start out with two, and then add the others when they come down in price).

http://fastra.ua.ac.be/en/index.html
 
Back
Top