Hey we are set to break 1 million points this week!


I'm also having a problem with my PC. I just upgraded my processor, mobo, and memory. I was happy b/c I'm running an XP 2700+ now and thought it'd be nice to get that extra processing power. However, when I run FAH console client, right as it begins work on the first step my computer freezes. I have no idea what the problem is but I want to get it fixed so I can get over that 400 point / day average...

[Edited on July 30, 2004 at 10:25 AM. Reason : more]

7/30/2004 10:23:34 AM

gahhh mine is still at 95%, it's been that way for about a half hour

7/30/2004 10:27:27 AM

^^sounds like processor overheating... since it's probably running near 100% when it's folding... just a thought. slap a larger heatsink or fan, and make sure it's properly lapped/mounted

i got another 10 tinkers this week

[Edited on July 30, 2004 at 10:33 AM. Reason : .]

7/30/2004 10:31:37 AM

maybe that extra boost will keep us in the top 100, hopefully we can keep moving up. btw you were over 400 for atleast one cycle that I noticed the other day.

today should be a good day for me considering yesterday i barely squeezed out anything. tinkers are still not worth it...

7/30/2004 11:51:55 AM

Opstand-Basically what Prospero said. If you are not overclocking, the retail HSF should be fine. Try reseating it. 1,000,000 points will be a great milestone for the team. I have been keeping my eye on it for a little while now.

To my right are 2 new P4 3.0 machines at work. Now to convince someone to run F@H

7/30/2004 1:02:23 PM

I have a Thermalright SK6 on it, much better quality than the retail heatsink. Maybe I need a better fan...I have a good Panaflo on there but I should probably get one that moves more air.

7/30/2004 1:05:51 PM

^what about case fan?

7/30/2004 1:33:52 PM

^^ What are your temps as idle and 100%?

7/30/2004 2:17:30 PM

You know you are a foldaholic when you know what core you are running just by looking at CPU temps. On my laptop:
47C = Tinker
51C = GROMACS

8/1/2004 3:27:15 PM

I have joined...

8/2/2004 12:58:17 AM

I assume you are going to put some of these to work. LOL Great to have you on the team.

8/2/2004 1:13:23 AM

yeah baby. ranked 5396! oh, and I'll hit 30k today, and we will hit a milli today

8/2/2004 1:53:10 AM

i was going to wait up and post when we hit a mil, but thinking of hitting the sack... anyhow, it will be soon. i got 50% of this last points update with 44 points, love the javascript mouse over...

Quote :

"Sun Aug 1 20:59:59 PDT 2004 team teamname score wu 59 NCSU Wolfpack 999501 28106 "

8/2/2004 1:59:22 AM

1,000,000! Great work people.

8/2/2004 8:48:29 AM

Welcome to our newest member: GrimX!

8/2/2004 10:17:08 AM

hooray

8/2/2004 3:02:24 PM

Welcome to our newest member: CarZin!

Also a pre-release of F@H v5.0 is out. Here is the anouncement:
http://forum.folding-community.org/viewtopic.php?t=8854

Quote :

"Hi, We now have available for download a prerelease of the version 5.00 client for Linux and Windows (Macintosh will be forthcoming). It incorporates many suggestions and has several major improvements. Notable changes are the incorporation of a scheme for uploading results even when a work server is down, built in Windows service support, and an enhanced Windows GUI. If you would like to try it, you may download it at http://vspx27.stanford.edu/beta/beta.html . Note that it expires August 30, and you should not feel compelled to download this client rather than wait for the final release. The Windows GUI has known issues on Windows 95/98. Below is the change log with details. -Guha Changes in F@H Client v. 5.00 -------------------------------------- - Collection server support. If a client finishes a work unit from a Work server with this feature enabled, then if 2 upload attempts to that server fail, the client will return to the work unit to a Collection server. Pending credit will be given for the unit, becoming final when the Work server comes back up and verifies what the Collection server has received. All work servers will eventually be associated with a Collection server, and many already are. - Configuration option to allow downloads and uploads greater than 5 MB in size, which some new projects will require due to the size of what they are simulating. - G@H configuration option reworded to be "No deadline" option. - Print executable name and launch directory to log. - Fix for bug where queue slot could get stuck "Fetching" - MyFolding links updated. Linux specific: - OpenBSD support: If give flag -openBSD, client will automatically make elf2olf system call on downloaded Cores. Windows specific: - Configuration option to pause work when the computer is using battery power (for laptops). - Behavior in response to a shutdown or logoff of the system should be improved (exiting nicely). - Console: Built in service installation support through Configuration, to have the program automatically begin at machine startup. If one wishes, the service can be further configured and interacted with through the Services tool (GUI) that can be accessed through the Windows Settings (under Administrative Tools in 2000/NT). The Sc command line tool may also be used. There are known issues with the Net tool. - GUI: Larger window size and more informative display. - GUI: Screensaver configuration option. - GUI: Print a message to FAHlog.txt when ask to network."

8/4/2004 8:40:46 AM

looks like we might hold down 99 for a few days... hopefully we can move up/ hold steady a while until school starts back up.

8/4/2004 10:13:29 AM

we are playing flipidy-flopedy with the ALS Attack Squad, we are currently at 100... also CarZin, you still in? i was going to post questions about how the machine that you had built was working out...

8/5/2004 10:57:56 AM

Hey. I am still in.

8/5/2004 11:36:33 AM

I havent optiized anything yet. Its only using 50% of my HT CPU. Isnt there a way to bump that higher? And is it just me, or do some WUs take a lot longer than some other ones.

[Edited on August 5, 2004 at 11:51 AM. Reason : .]

8/5/2004 11:49:24 AM

^i have an HT as well and i run 2 consoles instead of just one, to use 100% utilization

8/5/2004 12:01:15 PM

the HT can get a roughly 15% boost by running two instances of the program. At least one would be the console version (the other could be console or graphical). The main thing to note in setting it up is to use the -local flag. The rest of the installation is the same, just install into different folders. The -local flag makes it where they work out of the individual folders, as opposed to working on the same work unit (which does not work).

here is a link to the best tweaking guide I have found http://www.maximumpc.com/features/feature_2003-06-10.html It should optimize things very well.

Also some take days as others could take hours. They have it set up where slower CPU's run long, and faster run faster work units. This is not always the case as sometimes the only work units left on the servers are the slow ones (ussually tinkers). The points are almost worked out where slow WU's get you more points (but depending on the processor it could work out differently), and faster ones less points. You should be turning in more stuff soon, and move up the ladder...

hopefully that makes sense...

[Edited on August 5, 2004 at 12:02 PM. Reason : ]

8/5/2004 12:01:44 PM

The maximum PC guide is quite good. There is a link to it on our team website also:
http://www.genejockeys.com/Fold&Gene@Home/
One thing to remember about running 2 instances on a HT, is that sometimes Windows has a hard time releasing CPU time from F@H for other CPU intensive programs. There is a link to a discussion about this on our website. I'm not sure how common this is, but it has been observed.

I just borged 2 3GHz HT machines, but since they are used for very specialized software for community use, I only ran 1 instance to avoid any possible conflicts.


CarZin-I see you have had 4 instances return WUs so far. Good work!

8/5/2004 1:23:55 PM

just curious are they under your tag or a different tag?

8/5/2004 1:29:32 PM

I have 6 machines running WUs right now. The slowest one is a 2.4 P4 with a gig of RAM. Two of the machines have yet to complete their work unit and have been churning on it for over a day. And one of those is a 3.0 P4 with a gig of RAM. I just restarted the process with the -forceasm tag and did a few optimizations. I have 15 high end work stations at my disposal, but I am only going to push my luck on 6 or so. I might be able to get some time on one of the research computers...

8/5/2004 1:46:48 PM

w0000, 30K!

[Edited on August 5, 2004 at 2:03 PM. Reason : ]

8/5/2004 1:54:26 PM

i actually only have two computers over 2 ghz. the rest take their time. the following flags are the ones I use...

-local -advmethods -forceasm

this site describes most of the flags. http://folding.stanford.edu/console-userguide.html Sometimes I will run one unit on another computer over night with the -oneunit flag. This is very useful sometimes. It gets addictive ones it gets going...

8/5/2004 2:01:27 PM

I installed the 4.0 client before the prerelease. i didnt jump on installing 5.0, because it looks like its a lot of GUI changes which I dont care for.

8/5/2004 2:03:15 PM

^^^They are under KRUZNBY, but I doubt they will run 24/7, so I only expect a minor increase.
^^It can take a few days for some WU and your active processors number only reflects different instances that have returned WU. Sounds like you are well on your way. It's probably best to try it out on the few machines you have first so you can get a feel for it, before rolling it out to others.
^There is a v5.0b console, but since there is no expected performace gain, I'm waiting until it is out of beta. Also it expires at the end of the month.

[Edited on August 5, 2004 at 2:07 PM. Reason : Hot topic, well you know who you are "^"]

8/5/2004 2:04:36 PM

okay well CarZin has finally cashed in some WU's up to 10, so it is working... up to 824, w/ 5 proc's showing, number 77 on team. nice...

i guess i was a bit impatient this morning

8/5/2004 2:30:37 PM

nevermind... they dont post the source.

[Edited on August 5, 2004 at 2:50 PM. Reason : .]

8/5/2004 2:39:41 PM

Unfortunatly the code is not open source. This has been discussed on the official F@H board. IIRC, the main concern for allowing it to be open source, was that people may be able to cheat the system and give bogus results that would hinder the project. It would be great if there was some way to get it to work on your clusters.

8/5/2004 2:51:19 PM

It would be nice if they could put together an MPI friendly version of the code then...

8/5/2004 3:04:48 PM

pegging along am I.... Hopefully make it into the top ten in a short time.

If things go right, I might just do it in less than 200 total WU's.

8/5/2004 6:00:56 PM

^good luck, i'm just about to break the top ten in 2 more days, and i have 417 WU's and i thought i was doing pretty good.

this might give you a better idea
http://folding.extremeoverclocking.com/user_overtake.php?s=&u=84926

btw, how did you pull this off?
08.05, 3pm 231 1
08.05, 12pm 298 2
08.05, 9am 221 1
08.05, 6am 146 1

the best i've gotton on core79 is ~90pts./WU
unless you've been turning in all your tinkers all at once... that's pretty impressive.

[Edited on August 5, 2004 at 7:06 PM. Reason : .]

[Edited on August 5, 2004 at 7:06 PM. Reason : .]

8/5/2004 7:00:59 PM

^hehehe, pretty sure that is a lot of tinkers at once. I raided some of the work machines that I use on a daily basis a few days ago. They all started on tinkers.

Also I forgot to mention that they are all the same pc basically. Each one is exactly the same, down to the same programs on the hard drives.

As far as the Extreme OCer's overtake page, it doesn't take into consideration that I'm constantly machines into the battle. It only projects what I'm doing right now. I have two personal machines that fold, the four machines at work, and hopefully a friend of mine is going to go live with 7 servers, each with dual 3.6ghz P4's, all in my name.

ALL IN TEH NAME OF SCIENCE!!

[Edited on August 5, 2004 at 10:14 PM. Reason : more info]

8/5/2004 10:09:54 PM

^Sweet! I noticed you have been steadily increasing production. Honk when you pass me.

8/5/2004 10:46:35 PM

I like the new last 7 months production for the EOC stats. I guess they will be more useful as time goes on since there was a weighting adjustment for TINKERS awhile back, so it is difficult to make comparisons directly.

8/6/2004 2:40:46 PM

woopty-woop... number 3 producer for the team

8/6/2004 9:13:03 PM

yeah! 10th under 480WU's

8/7/2004 8:55:35 AM

damn you... passing me

8/7/2004 3:37:11 PM

just jumped up from 130 to 89th

8/7/2004 5:27:31 PM

the first 100 or spots fly by... it's hacking through the top thirty that gets kinda slow, regardless of your production.

8/7/2004 7:59:03 PM

i just had a record high today, 958 points in one day

8/8/2004 1:04:00 AM

I had to shut down 5 of my workstations for a full day. I have 2 machines which will cash in on a 1000 WU protein fold within 24 hours, and 2 other 400 WU proteins in the same period. I expect to get close or over 1000 points on sunday.

8/8/2004 1:20:47 AM

Why is it my ratio of points to work units is really high? Its take me on 18 work units to get 2000 points, when the next closest person has a LOT more.

8/8/2004 2:33:20 PM

http://vspx27.stanford.edu/psummary.html

this is a site that has a break down of the points. if you are getting tinkers or larger gromacs that take longer and give more points, that is your situation.

about 3 months (guess) ago the tinkers took forever and gave a horribly low ratio to points. there was a flag to avoid getting these. so people would try and did get the smaller WU's which gave lower points, but higher ratios per WU. So people used to work this way. now that the points are more balanced people are turning in less WU's per amount of points. hope that somewhat makes sense...

8/8/2004 2:47:27 PM

There is also a bit of benchmarking that goes on so Stanford can make decisions about which machines should get which work units. I really don't know the details of this process. It really does not matter too much right now, because the value pts/time is faily even now as LeGo mentioned. There are differences, but now where near the way they used to be.

Also a bit of terminology:
Work Unit (WU) - The actual data assigned to you to process and return. A list of current projects is at the URL LeGo showed.
Frame - It's a fraction of a WU. Think of it like a frame in a movie, it is a snapshot in the process of folding a protein. Sometimes there are 100 frames/WU other times there are 2000 etc.
Points - Numerical value assigned to each WU based benchmarking it on a dedicated 2.8GHz Pentium 4 machine with SSE2 disabled.

The true test of the performance of a given CPU and WU's is the pts/time ratio.

Hope this helps.

8/8/2004 3:12:55 PM

Welcome to our newest member: matt!

8/8/2004 10:31:49 PM