What does it take?

sbabb · Post by **sbabb** » Tue Oct 21, 2003 7:46 am

What does it take to go from white to green to yellow to orange to red?

What does it take to work your way into the top 20 in the SPCR stats? How about the top 1000 in all of F@H?

The ExtremeOverclocking stats at this moment (11:45 AM EDT, 21 Oct 2003) show TRC-13 (in the #1 spot) as our only red member. TRC-13 was kind enough to post his/her setup yesterday: 22 CPUs ranging from an Athlon MP dualie through Barton 2800s down to a 1.0 P3.

bkh (#3), Lockheed (#4), and Zhentar (#5) all show as orange. The Stanford stats show them having 33, 14, and 17 CPUs, respectively.

mjrusso45 is currently yellow in the #2 spot with 12 active CPUs in the past week.

In slots #6-#10 are: tragus (yellow/6 CPUs), mike_chin (yellow/5 CPUs), Rusty075 (green, but almost yellow/2 CPUs), aphonos (yellow/4 CPUs), and CharlieChan (yellow/8 CPUs.)

Our other yellow Folders are: Cosine (#16/13 CPUs), Metaluna (#18/7 CPUs), and tangled (#19/9 CPUs.)

Way down in #76 is me: Sbabb/green most of the time/4 CPUs. I'm currently running:

Barton 2500 24/7
Athlon XP1900 24/7
Athlon XP1800 whenever my daughter has it on, which is often
Athlon XP2000 laptop ~8/5, but I've been plugging it in and running it at home lately because I'm obsessed with folding.

My daughter's machine is running the GUI 3.24 client, but the other three are running EM-III with the 3.25 console client -advmethods -forceasm.

OK, obviously it takes lots of time and persistence to get to the top 10. The longer you keep folding, the higher your score gets.

What else does it take? What are the people who are yellow with <10 CPUs running? How is Rusty075 getting almost 1000/week out of only 2 CPUs?

Is there a point at which it doesn't make sense to install F@H on a system? I have a 233MMX running Linux as a file server and a Cobalt Qube (K6-2/450) web server which runs a highly modified Linux. Is it worth making them fold, or will they never make a deadline anyway?

Obsessed minds want to know!

Thanks,
Scott

CulRidr · Post by **CulRidr** » Tue Oct 21, 2003 7:59 am

could someone explain to me what exactly this is all about, what the point is, the possible rewards? I have tried to figure out exactly what this "folding" thing is and im clueless.....

a few details would satisfy my curious mind...

TRC-13 · Post by **TRC-13** » Tue Oct 21, 2003 8:34 am

The colors are based on a given amount of points in a week.

Color codes are as follows: 0 = Grey / 1-499 = White / 500-999 = Green / 1k-2.49k = Yellow / 2.5k-4.9k = Orange/ 5k+ = Red

Post by **Rusty075** » Tue Oct 21, 2003 8:49 am

CulRidr,

check out Folding@Home's site: F@H It explains the who/what/why better than I'd be able to.

Sbabb,

What the stats don't tell you is that my two machines are each Cray-3 supercomputers that I bought off Ebay.

Actually F@H has never gotten my number of processor's right. At the moment my farm consists of:

P4 1.6
P4 2.4
P4 2.53
1Ghz T-bird

The T-bird is the only 24/7 folding box. The P4's are all borged work boxes, and they're set to only fold when idle. Their output varies depending upon whether or not my co-workers leave their machines on over the weekend.

I haven't reinstalled F@H on my main machine at home yet.

sbabb · Post by **sbabb** » Tue Oct 21, 2003 8:50 am

Folding@Home is a distributed computing project. The main web site for it is folding.stanford.edu

Are you familar with things like Genome@Home or SETI@Home? If so, then Folding@Home is the same type of thing.

The Folding@Home servers at Stanford University parcel out work units to thousands of F@H participants around the world. These many computer systems then process the work units (usually over many hours) and return the results to Stanford. The "processing" consists of running simulations of protein folding. The "work units" are chunks of data from the results of experiments conducted in labs at Stanford. What a protein manages to do in microseconds requires hours of processing time on a computer and the only practical way to process all of that data is to distribute it to thousands of volunteers PCs around the world.

The possible rewards from this entire project are advancement in understanding of many diseases like Alzheimers, cystic fibrosis, emphysema, "mad cow" disease, and cancer. Maybe we can advance the understanding of these diseases to the point of a prevention or cure.

To make things interesting, Stanford assigns a "score" to each completed work unit and posts these scores on their stats page at the site mentioned above. Volunteers can join "teams" if they want, and contribute their score toward their team's total. SPCR has a F@H team, and several of the members here contribute toward the team score.

Once you get into the "competitive" aspect of trying to run up your score and your team's score, you start using the Extreme Overclocking F@H stats site to keep track of who you're chasing and who is creeping up on you in the standings. The SPCR team is currently #28 in the team standings, but we'll get overrun by another team that's putting up scores faster than us soon. A month or so after that, we should pass another team, so we'll be back to #28 again, and climbing. We were at #30 when I started folding a few weeks ago (not that my ~3000 points contributed since then is the main factor in our move up, but a little bit from many people adds up to a LOT!)

If you're into competition and bragging rights, then the scores and teams will probably lead to an obsession with folding and running up your score. I'm not in much of a position to brag right now. I've moved up 7 spots in the last 24 hours, but 6 of the people I passed aren't actively contributing to the folding effort any more, so I've really only passed one active folder. But now I have DryFire in my sights, and we've been a close match in terms of score production. He's firing up a new machine soon, so I may not catch him, but right now the stats say that I'll pass him and Mr_Smartepants within the next 36 hours or so. Then I'll have to try to maintain my position ahead of them, I'm not sure what Mr_Smartepants is running, but DryFire is definitely not going to let me slack off if I manage to pass him.

See all the fun we're having? It's like being able to race go-karts and have it feed the starving!

Check out the Stanford site above for more info about Folding@Home.

Hope this helps,
Scott

sbabb · Post by **sbabb** » Tue Oct 21, 2003 8:59 am

Rusty075 wrote: What the stats don't tell you is that my two machines are each Cray-3 supercomputers that I bought off Ebay.

I'm actually sitting in the middle of the equivalent of a Cray-3 here at work. If F@H ever releases a SPARC/Solaris client, I've got a few dozen SPARC CPUs I could borg here

I'd use cron to make them only fold after hours, but 12 hours/day from a few dozen CPUs would certainly help me out.

miker · Post by **miker** » Tue Oct 21, 2003 9:24 am

Well for the sake of info, here's my "farm"

P4 2.0A x 4 (Misc "borged" PC's from around the office)
AthlonXP 3200+ (My silent home PC, a 2500+ OC'd to 3200+ specs)
AthlonXP 2100+ ("The Wife's" PC, a 1700+ OC'd to 2100+ specs) This one is only on 1/10 the time.
Dual Proc Xeon 2.8 with HT (Counts as 4 procs)
Dual Proc Xeon 2.0 (Counts as 2 procs)
PIII 700Mhz Laptop

CharlieChan · Post by **CharlieChan** » Tue Oct 21, 2003 10:39 am

sbabb wrote: See all the fun we're having? It's like being able to race go-karts and have it feed the starving!

Now calm down!

The machines I have running F@H are

1 Dual P3 1G RH9/Wine
1 P4 2.4G win2K
1 P4 2.0G winxp
1 P4 2.4G winxp (my kids computer not always on)

and tuck away in the garage is a linux folding farm comprising of

1 P3 700Mhz
1 Dual P3 866Mhz
1 Duron 1.3G

CulRidr · Post by **CulRidr** » Tue Oct 21, 2003 11:04 am

rusy075 and sbabb, thank you very much for the links and VERY detailed information.....being in the obscure sucks

but ya, what a great way to contribute to the further advancements of science. Unfortunately, only having a dialup connexion until xmas will unavailable until then...
the 2500+ O/C'ed to 3200+ would make some points for the team, and the amd500 (soon to be turned into a duron1300) would make a few more....but nothing compared to some of the combos you guys have....

neway thx a bunch, and good luck to the SPCR team

DryFire · Post by **DryFire** » Tue Oct 21, 2003 11:59 am

well I'm just running a barton 2500+ at 3000+ specs.
my sis has a 1.7ghz p4 wilmate. it is on like 6 hours a week or at least last week.

I was going to borg a ton of computers are schoool but they're 1ghz celerons and i decided the consequences for getting caught were just not worth 1 wu a month.

If i could get my 1800 back and up 24/7 I'd be green with only 2 cpus.

wussboy · Post by **wussboy** » Tue Oct 21, 2003 2:35 pm

Maybe this will help.

I seem to really ride the line of white/green. I am green about half the time, and white half the time (usually varies every few days). I have:

1 AMD T-bird 800 on ~12 hours a day.
1 1800+ on 6-12 hours a day.
1 P4 2.4 on 24-7

WarpedPlatter · Post by **WarpedPlatter** » Wed Oct 22, 2003 6:22 am

DryFire wrote: I was going to borg a ton of computers are schoool but they're 1ghz celerons and i decided the consequences for getting caught were just not worth 1 wu a month.

My 1.2GHz Celeron machine completes about 5 units a week, generating roughly 200 points. So, a bunch of 1GHz Celerons can really help out, especially if they're the Tualatin core version. Still, you shouldn't risk getting into serious trouble.

Wrah · Post by **Wrah** » Thu Oct 23, 2003 2:49 am

Oh how quickly I've fallen from the top-10 now I've lost my internet connection (delibaretely, I'm moving house). And I probably never will get back in there in the future since I was producing just enough to maintain my position.

Until 2 weeks ago I was still yellow with this setup:

- Barton 2800
- T-Bird 1.4
- Tualitin Celeron @1.5
- 2x P3-866
- p2-450

The barton was producing half of the points on it's own. The p3 are at work so I don't have to pay for the electricity, otherwise I wouldn't bother with them. They need up to over a week to complete a WU. So electricity-wise their WU's might cost up to 5-10x more. If you have a lot of these slow things folding, it might be cheaper in the long run to just dump em and replace them with one XP 2500 on a cheap mobo or something.

dukla2000 · Post by **dukla2000** » Thu Oct 23, 2003 3:27 am

sbabb wrote:What does it take to go from white to green to yellow to orange to red?

My 'domain' is on the cusp green/yellow. To be yellow I need 1000 points/week, which is reasonably possible with my 3 boxen:
XP2400 (+- 60 points/day)
XP2200 (+- 55 ppd)
P3 1.2M (+- 40 ppd)

I get demoted from yellow to green if I get slugged by too many Tinker, or my P3 is too mobile and can't get to the web, or I take out an Athlon for a rebuild or 24 hour Prime testing etc.

But if you ran your daughter 24/7 you should maintain yellow.

Wrah wrote:Oh how quickly I've fallen from the top-10

I remember it well: I peaked at 9th (for a few hours) but never regained a place since I took a 2 week vacation!

CharlieChan · Post by **CharlieChan** » Thu Oct 23, 2003 6:34 am

Wrah wrote:Oh how quickly I've fallen from the top-10

miker · Post by **miker** » Thu Oct 23, 2003 7:36 am

Wrah wrote:Oh how quickly I've fallen from the top-10

Well it looks like I also finally fell from the No 1 spot, never to return. I just don't have the PC power to compete

sbabb · Post by **sbabb** » Fri Oct 24, 2003 7:22 am

Woo-hoo! I'm yellow!

1032.32 points in the past week. Don't know how long it will last, but at least I got there once!

So if my 4 CPUs can get 1000+ points/week, then ~29 more CPUs from the whole SPCR community will keep us from being overtaken by the slo-tech team. If anyone is on the fence about folding for SPCR, now would be a good time to join!

Scott

Zyzzyx · Post by **Zyzzyx** » Fri Oct 24, 2003 8:29 am

I've got 4 systems going right now.

P4 2.2 - 24/7 (work)
AMD 1600 - 20/7 (home main/gaming system)
AMD 1700 - 24/7 (secondary home system)
P3-566 - 24/7 (home backup server)

That's been putting me at 700-900 pts/wk depending on what I get and how much they're up.

And I've noticed that you (Sbabb) are on my Threats list now.

sbabb · Post by **sbabb** » Fri Oct 24, 2003 8:46 am

I wouldn't put too much faith in that date, though. Before we get to that point I should be going on vacation, which will cut down the contribution from my XP2000 laptop. I'm also hoping to water cool my fast machine (XP2500) before then, too, so it'll be down for awhile.

Zyzzyx · Post by **Zyzzyx** » Fri Oct 24, 2003 10:04 am

Heh...

And I just finished setting up a new computer here at work. Celeron 2.2ghz. Got Folding going with FireDaemon as a service, all hidden.

Shouldnt' be a problem, as I'm one of the two IT guys here.

And I should be able to do the same with some new systems coming in the next couple weeks.

sbabb · Post by **sbabb** » Fri Oct 24, 2003 11:00 am

Borg 'em up, my friend! We need all the CPUs we can get to fight off the slo-tech team.

If they'd only release a SPARC/Solaris client...

miker · Post by **miker** » Fri Oct 24, 2003 11:54 am

Just borged another Dual-Proc Xeon 2.8Ghz IBM server.

That should help.

Semm · Post by **Semm** » Fri Oct 24, 2003 12:40 pm

I have a couple laptops coming that I'm supposed to fix for friends. They'll both be borged (after getting permission, of course). One celeron 750, and one P4M-2.0-ish. Not much, but every little bit helps, I guess.

sbabb:
Running a linux emulator an option? http://www.ugcs.caltech.edu/~steven/lxrun/

Can't exactly test it, but it would be great if you could get them folding, even at reduced speed...

Semm

sbabb · Post by **sbabb** » Sat Oct 25, 2003 7:03 pm

Semm wrote:I have a couple laptops coming that I'm supposed to fix for friends. They'll both be borged (after getting permission, of course). One celeron 750, and one P4M-2.0-ish. Not much, but every little bit helps, I guess.

sbabb:
Running a linux emulator an option? http://www.ugcs.caltech.edu/~steven/lxrun/

Can't exactly test it, but it would be great if you could get them folding, even at reduced speed...

Semm

Every little bit definitely helps, Semm! It looks like we've added a few new people and CPUs over the past couple of days, so the Slovenian team isn't gaining quite as fast as they were, but they're still gaining on us. Their #1 folder is a user named "ELan_Belisarius" who cranks out over 1000 points/day! The Stanford stats say that 69 CPUs have contributed to that individual's score in the past week. Talk about a master Borger!

Lxrun would be an idea if they were x86 (Intel-esque) Solaris machines, but they're all 64-bit UltraSPARC CPUs and lxrun doesn't help out there, as far as I can tell.

The FAQ at Stanford says that they're working on a SPARC/Solaris client, but not much seems to have come from that. Too bad. We're pretty much totally Sun-based at work. Some of the larger servers have many (50+) CPUs in a single box, all happily playing with a massive bus bandwidth. If Stanford were to release a multi-threaded client that could take advantage of dozens of CPUs to grind on a single work unit, then those boxes might be able to punch out a couple of W/Us every hour! Hard to tell without actually having a client to play with. But my borging would probably be limited to just the single-CPU workstations in my group, but that's still a few dozen boxes...

Maybe I could hack the internet-connected boxes at the local Best Buy's and get them folding for us anonymously. They usually only have one or two internet-connected boxes, in each store's PC section, though. All the local Best Buy's use the same login and password, though, so they're easy to break out of their canned displays.

Hmmm, now I'm having evil thoughts of putting the folding client on a CD-RW and just installing it on every machine I can find in any PC store where the employees don't watch you too closesly, just so that those machines will start folding for us when they eventually get sold. SOMEBODY STOP ME BEFORE IT'S TOO LATE!

DryFire · Post by **DryFire** » Sat Oct 25, 2003 9:25 pm

hey i know the password for my local bestbuy and how to get ot of the programs. that might just be an idea.

now to go pay them a visit.

Zyzzyx · Post by **Zyzzyx** » Sat Oct 25, 2003 10:16 pm

Don't do it.

I think a basic tenet of Folding is only doing it with systems that you have permission to use.

It wouldn't look good for Stanford if Best Buy were to find F@H installed on their systems.

Semm · Post by **Semm** » Sun Oct 26, 2003 12:59 am

Lxrun would be an idea if they were x86 (Intel-esque) Solaris machines, but they're all 64-bit UltraSPARC CPUs and lxrun doesn't help out there, as far as I can tell.

*reads the lxrun FAQ* Yep, you're right, I'm afraid. And since it's in maintenance mode, it's not likely that that'll be added soon.

Not sure what to make of this on Sun's site, though.

"Sun's commitment to the Linux operating system brings additional value to customers of its Solaris/SPARC architecture. ***Already, Sun systems have built-in compatibility with Linux, so that any Solaris-based system can also run Linux applications.*** New software such as Linux Compatibility Toolkit (LinCAT), announced today, can help simplify the process of assuring that Linux applications will run on the Sun Fire family of servers. And in the future, Sun's upcoming Solaris 9 Operating Environment will provide additional built-in Linux commands, utilities, and interfaces."

The page is this one.

Also, the source for lxrun is available, but I don't know how painful it would be to port it to the SPARC architecture.

Semm

sbabb · Post by **sbabb** » Sun Oct 26, 2003 10:20 am

Interesting page, but somewhat misleading. They should have said something like:

"Already, Sun's x86-based systems have built-in compatibility with Linux, so that any Solaris x86-based system can also run Linux applications."

The page is a press release from Feb 2002 when they were releasing x86-based Sun systems. The comments about Solaris/Linux compatibility are confusing. They were definitely talking about x86 Solaris, not SPARC Solaris, but that page certainly makes it sound like they were talking about both. Searching on "lxrun" on Sun's web site brings up a lot of pages that state that it only works on x86 Solaris.

Porting lxrun to Solaris probably wouldn't be too bad, it's writing the Pentium CPU emulator to run all of the the x86 code that would be difficult, and then trying to get decent performance out of an emulator. I'll ping on the Folding folks as Stanford to see how their native Solaris client work is going.

What does it take?

What does it take?

confused

Re: What does it take?