5 replies to this topic

[Viper]

What is a computer deal right now or server deal for like 5-10 pc's for folding@home? Of course I'd love to get a quad core G5 or an intel itanium computer, but I want the best bang fro the buck.

[boltsniper]

You must have money to burn man....

[Enigma1313]

Uhh...just outta curiousity...whats "Folding@home"?

[The Infinite Shindig]

Uhh...just outta curiousity...whats "Folding@home"?

Nerfers Nerfing Disease

Read the forum's sticky!

[baelnorn]

I did some research about diskless nodes about 2 months ago. I was not going to use them for F@H, but for textual data processing, but they will work well for F@H. The idea behind them is that you have essentially just mobo's and CPU's sitting behind a main server that gives them all assignments, in this case work units. The server is just any box with a fast ethernet port. You have the nodes boot off of the lan, with the server set up to network boot each one to a barebones linux system running F@H. When they complete they send the unit to the server which sends it to the F@H servers.

Here is what I was I had as a diskless node's hardware(Prices as of 2 months ago, newegg.com)
Athlon XP 2600+ 79.99
Asus A7V600-X 58.00
Zonet Zen3300E 10/100/1000 12.50
Wintec Ampo DDR400 512mb 39.99
Spire 5F271B1L3 9.99

Total: 200.47

I think that diskless nodes might be your cheapest option, and there may be a new deal on newegg for a better proccessor, I haven't checked lately.

Hope this helps,

Baelnorn

[cxwq]

Thanks for posting what I fully intended to post a couple days ago. That's the way I would go as well.

Basically the project falls within the ideal specs for a 'compute node' farm where each PC has limited-to-zero storage, memory, graphics, etc but as fast a CPU as is cost-effective. This is exactly what Google did when they were serving millions of searches per day on a near zero budget as a Stanford grad project way back in the day. Of course they threw the Google File System into the mix, but therein lies the genius.

I like baelnorn's specs overall. In general, weigh the cost of the entire system against the performance of the CPU. Get yesterday's RAM, the lowest-cost mobo you can find from a reputable manufacturer, onboard everything, and the fast proc that mobo will support.

It will also work well as a render farm if you're into video processing.