• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

VA Tech G5 Array, Folding@home !!!

L

LoStMaTt

2[H]4U
Joined
Feb 26, 2003
Messages
3,180
Within the next month or so I am going to attempt to get the techs over at VA Tech to run Folding@home for the [H]ard Team for at least 24 hours. One question. How will I distribute the program to all of the computers at once and start it simultaneously.
 
Are you going to kidnap the deans children or something? How are you planning on convincing them?
 
Originally posted by muffhead
Are you going to kidnap the deans children or something? How are you planning on convincing them?
Click to expand...

LOL, seriously man, that is a multi-million dollar SUPERCOMPUTER that you are trying to run on. Good luck!

FOLD ON!!!
 
I have connections..but anyways, say I get permissions, I will need help figuring out how to set this up. So how do I get F@H set up on 1100 comps at once, all working on the same WU's..if possible..
 
Originally posted by LoStMaTt
I have connections..but anyways, say I get permissions, I will need help figuring out how to set this up. So how do I get F@H set up on 1100 comps at once, all working on the same WU's..if possible..
Click to expand...
1. You darn well better have connections.
2. Go to http://forum.folding-community.org/index.php and contact OC-AMD
 
Its a cluster right? You run 1 instance and the cluster works on it all at once.
 
heh.. if that's the case, we'd be measuring it's output in Frames Per Second, as if F@H has become a First-Person shooter of sorts. :)

Should you get permission, you should let Vijay know. I'm sure he'd be elated to hear that.

.:Ttocs
 
I am going to make some phone calls this weekend. I will keep you updated. BTW, lets see some estimates. How many WU's do you think this array could pump out in 24 hours?
 
Originally posted by LoStMaTt
I am going to make some phone calls this weekend. I will keep you updated. BTW, lets see some estimates. How many WU's do you think this array could pump out in 24 hours?
Click to expand...

If they're tinkers? Maybe 3 ;)

Just a little tinker humor :D
 
well that is 1100 Dual 2ghz Machines....

4400ghz folding for 24 hours

if all that power works on a single work unit at a time...should chew up and spit out about 5000 WUs in 24 hours
 
you have no chance to survive! make your time!

sorry, had to... too many 24 hours jokes hehe
 
Holy poop, that's a lot of ghz. :eek: Man, I could sure use a cluster like that... :p

Maybe when I get money, I'll setup a cluster on some Xeon machines... ;)
 
Just be sure to set it up with OSUguy98 as the username.... and Team 33......... okay okay okay, someone had to say it... :D

must... resist.... urge...to...spend...money....on...boxen............
must...save....money.....for...house....so........I....can.....buy... more....
boxen....in.....the...future.............. must.... resist.......
 
Don't mean to rain on the parade, but I am a skeptic and I personally dont think this will work all that well. To start off with, programs that are run on clusters like this have to be specially designed in a manner that allows multiple processors to work on small chunks of it simultaneously. as an example, at college a couple years ago some guys set up a Linux Beowulf cluster of 8 machines. I was not really in on the whole deal, but I remember a couple of guys talked about running SETI on it and it worked the same speed as if only 1 computer were working on it, because thats what was happening.

Now, this could still be pretty successful and very cool, but I'm pretty sure you would have to run the client seperately on each machine, and the distribution of that would be beyond my Linux capabilities.

Brian Taylor
 
A G5/2.0 produces about 3.5-4 points per hour, with a range of about 2.25 to 5.25 (per my own experience). Figuring 4400 processors gives you about 200,000 points per day. This is maybe 10-15% of current total production, taking 1.3 million for the regular teams and guessing at team 0 and Google team's work. The load on the Stanford assignment servers when 4400 clients try to DL work units wouldn't be pretty.

Molecular dynamics like F@H is not a good app to code on a cluster, because the chunks of work you could parcel out to the machines are each so small that communication overhead would be too large. A shared-memory multi-processor like a standard 2/4/8 cpu box could be made to work, but not a cluster with network links, where the latency for data transfers is probably 2000 times longer (millisecond range vs. 100s of nanoseconds)

So you're stuck with WUs that take 4-12 hours each, and running 4400 of them at a time. Kind of cool, but not really interesting from a Computer Science point of view.

There's also the question if the 1100 machines are actually tied to the internet at all, or just a few "head-end" machines.
 
Originally posted by Warren Marts
A G5/2.0 produces about 3.5-4 points per hour, with a range of about 2.25 to 5.25 (per my own experience). Figuring 4400 processors gives you about 200,000 points per day. This is maybe 10-15% of current total production, taking 1.3 million for the regular teams and guessing at team 0 and Google team's work. The load on the Stanford assignment servers when 4400 clients try to DL work units wouldn't be pretty.

Molecular dynamics like F@H is not a good app to code on a cluster, because the chunks of work you could parcel out to the machines are each so small that communication overhead would be too large. A shared-memory multi-processor like a standard 2/4/8 cpu box could be made to work, but not a cluster with network links, where the latency for data transfers is probably 2000 times longer (millisecond range vs. 100s of nanoseconds)

So you're stuck with WUs that take 4-12 hours each, and running 4400 of them at a time. Kind of cool, but not really interesting from a Computer Science point of view.

There's also the question if the 1100 machines are actually tied to the internet at all, or just a few "head-end" machines.
Click to expand...

It's like when 4,400 people try making phone calls at the same time. :p If I still had DialUp, I'd get booted, regardless. ;)
 
Originally posted by LoStMaTt
Within the next month or so I am going to attempt to get the techs over at VA Tech to run Folding@home for the [H]ard Team for at least 24 hours. One question. How will I distribute the program to all of the computers at once and start it simultaneously.
Click to expand...
Supposedly, it won't be operational till May, where it will undergo month of internal testing. They may still be replacing the old G5s with the XServes and recabling the cluster.
 
Originally posted by SLee
Supposedly, it won't be operational till May, where it will undergo month of internal testing. They may still be replacing the old G5s with the XServes and recabling the cluster.
Click to expand...

I'd hate to have that job. That's a lot of CPU's to swap and a lot of cabling to do...
 
I called and their secretary said I should expect a return phone call sometime this week. She did suggest that now might be a good time since they are re-doing their cluster and aren't running any experiments right now, so hot dog! maybe we'll get some real [H]ard folding in sometime soon.!
 
You must log in or register to reply here.
Back
Top