• 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.

Data Points

G

gnewbury

[H]ard|DCer of the Month - September 2007
Joined
May 4, 2001
Messages
2,544
I'd like to start another thread on data points primarily for OSU98, but of benefit to all of us.
I've a
PIIIM 1.113 that is cranking a p1475 in an estimated 45:33, frame time 27:20, 354 points
Sempron 2400, p1475, 51:45, 31:03, 364
all on v5 bigunits.
I bring this up because that little PIII had been running TT's and getting 40ppd, now it's cranking 191ppd. When I set it up on TT's about 8 months ago both TT's and Gro were getting about the same ppd - obviously that has greatly changed and I missed it,

If we could decide on a format,
CPU, protein, WU time, frame time, points
Then it would make it easy for us nonfancy peeps to cut and past into a text file, then import to Excel via CSV
 
Thanks for the info,etc... The best format I can think of (for me) is the one that LPerry uses in EMIII... I already have a spreadsheet set up with my personal boxen and all I do is just highlight the html file that EMIII produces and paste it into Excel... I have all the boxes checked, with the exception of the protein image... That way, EMIII does most of the hard work... and everyone would have the same format... Then it's just a matter of collecting all those files from people that are interested... and then just data manipulation and summarization...

Though, if we want to create a basic "CPU, protein, WU time, frame time, points" spreadsheet, could we have someone throw together a quick webpage that you could just enter the info? That way, it would all be formatted the same way, and it could just build a text file which would could import into Excel quite easily....

I dunno, just brainstorming here...


Keep on Folding!!

 
If you could get it hosted somewhere a simple mail script is pretty easy to set up. Places that give good hosting while allowing CGI scripts for a sensible price on teh other hand are a different matter. What would be easier is probably a perl script to parse email sent to a dedicated email address if someone has a domain and doesnt mind instaling it. Not really my forte though.
 
Herulach said:
If you could get it hosted somewhere a simple mail script is pretty easy to set up. Places that give good hosting while allowing CGI scripts for a sensible price on teh other hand are a different matter. What would be easier is probably a perl script to parse email sent to a dedicated email address if someone has a domain and doesnt mind instaling it. Not really my forte though.
Click to expand...


I have a website we could do that on, I'm just a noob when it comes to scripting.
 
Which? The email script or the parser? THinking about it a bit more the best way to do it would prob be in php, that way you could write it into a proper database and then do pretty much whatever you want to as far as sorting goes. Let me have a think, have a fair few commitments w/ family +girlfriend over next couple of weeks but i might get some time to look at something.

Thinking something along the lines of.
Page 1.
php Script, with a form on, hit submit bangs things into a database.
page2,
another script, or several, probably one for neatnesses sake, that could take what you want to sort it by as an argument, then output into a table, xml, csv, however the hell you wanted really. Its all fairly simple, like i said, bi just havent got a lot of time on my hands, especially seen as i want to get this lcd scroller app finished.

In the true spirit of the forum, anyone know how you make system tray apps in VB 6?
 
What we should do is have a sticky that someone keeps up to date with the latest recommendations about how the project should be run (switches, config settings). There's always going to be some variability, but if we had recommendations that were reviewed every month or two, I'm sure we could churn out a bunch more WU's. I know that OCAU was better at getting the word out about the -forceasm and -advmethods switches back when they came out, so it would be nice if we could keep the team up to date somehow so we can optimize the amount of work we're getting done. For example, I wonder how many people don't have the large WU's enabled that could be doing them?
 
That's a good idea... DR did a good job of putting more info in the Stickies... Maybe it's best for him to make a Linux based recommended setup... and windows recommended... (Not saying you should make it DR, just saying that your a mod and could sticky it and update it as necessary)

Sorry Herulach, don't know how to do that... VB is lost on me... anything past ANSI C is lost on me!! :D


Keep on Folding!!

 
Its king whos the Admin over at hardfolding yes? Because they already use php and i assume either mysql or postgresql, so it wouldnt be too much of a leap to get it set up over there, assuming hes up for it. I certainly agree probably a lot of people arent runnign big packets that could, but i imagine most of them have reasons too.
 
OSUguy98 said:
Sorry Herulach, don't know how to do that... VB is lost on me... anything past ANSI C is lost on me!! :D
Click to expand...
I'd be inclined to argue that VB6 is behind ANSI C, not ahead of....

Nope, don't know how to do that. Linux programmer here. I might be able to help you with extracting/parsing the emails, though...
 
Yup, King_N is the hardfolding.com webmaster... being the "lurker extrodinaire" he is, he may chime in...

The only problem with all of us entering our own data would be "early end units"... Someone would probably have to review the data and delete any entries that are WAY off base...

Okay okay okay... Anything BUT ANSI C..... and a LITTLE bit of C++

Keep on Folding!!

 
I too would agree C is infront of VB by a long stretch, but being as how the only other languages i know are fortran and php, neither of which itsdoable in as far as i know. The same thing could prob be accomplished w/ fortran but i really cant be bothered trying to figure out how to do windows api calls. Let alone how to access com ports.

The more i think about the parsing email idea i think its prob a bit daft, especialy if we could get something like i outlined with the database hosted on hardfolding. As far as early end units go, if it was set up rgith itd be pretty easy for someone to sort it by core and time, then get anything ridiculously low. Although if we were just doing frame times as early ends wouldnt matter quite as much. Although they prob wouldnt be as useful
 
I do not like the idea of using the stats generated by EMIII for one reason. Many of the computers crunching this are not for folding only. Actual computer use will skew the results of the time to complete the unit. Some frames will take longer to complete and others. My main computer would actually work well like this for the most part since it crunches the vast majority of it's units without anything else going on. Hopefully, I'll change this soon by moving it back in with me.

For the most part, leaving the computer alone to do nothing but chew on a couple of frames without interruption and noting the frame times would probably be inefficient but about the best way to do this. If there is a "large" discrepency between a couple of frames, then check a couple more to see if it settled down.

I'm not sure about the best way to collect the information. I don't think having people post the info in a thread is a very good way, though. We like to get off topic too much and it has a tendancy to get cluttered. If someone or a couple of people didn't mind setting up an email address for submissions and send in the results that way it would probably be easier. I suggest a couple of people have access to the account so one person doesn't have to trudge through all the info. Later on they could get together and combine the info they each have to compile the list and put it in whatever format seems to be the best for it.

 
I wouldnt mind doing the whole manually sorting through emails thing, or at least helping, im certain we all have a few (hundred) gmail invites that could be used. It just seems so, cludgy.
 
Hito, do you have a mail server running on your domain? If so, this would be really easy...
 
SmokeRngs said:
For the most part, leaving the computer alone to do nothing but chew on a couple of frames without interruption and noting the frame times would probably be inefficient but about the best way to do this. If there is a "large" discrepency between a couple of frames, then check a couple more to see if it settled down.

Click to expand...


I would agree with you but that QMD cores totally screw this up. Their frames are all over the place.
 
Yup, QMD would royally screw up the mix... that's why I was suggesting basing it on total time of completion...

Would it be entirely impossible to just post a selection from your log file? and use some kind of program to parse it out into the info we need.... Then, we could have the best of both worlds?


Keep on Folding!!

 
Please create a "data" user (data@yourdomain.com) and I'll whip up a perl script to retrieve their mail through pop3 and parse it. Do you have perl installed (you probably do)?
 
Code: 
#! /usr/bin/perl
use Net::POP3;
print "Content-Type: text/plain\n\n";

# Constructors
$pop = Net::POP3->new('localhost');
if ($pop->login("data", "password") > 0)
 {
  my $msgnums = $pop->list; # hashref of msgnum => size
  foreach my $msgnum (keys %$msgnums)
   {
    my $msg = $pop->get($msgnum);
    print @$msg;
    print "---------------End of Message--------------\n";
#    $pop->delete($msgnum);
   }
 }
$pop->quit;
This will do the following:
a) print a http header
b) log into the local machine via pop3
c) send user "data" and password "password" (you can change this)
d) get all the messages and print them out with the "---" line as a seperator
e) logout.

So, copy this to your cgi-bin directory as script.pl, and send data a message. Then go to yoursite.com/cgi-bin/script.pl.
 
Hito Bahadur said:
I would agree with you but that QMD cores totally screw this up. Their frames are all over the place.
Click to expand...

Is it really worth possibly skewing a lot of the data for just one core, though? I would prefer to have accurate results for the rest of the cores and leave the QMD out completely than have so many of the results skewed. If you wanted to do the QMD core different from the rest, that would probably work out. I just wouldn't use the same process for all cores.

 
That's why I use a manual method. I really think there needs to be official benchmarking tools.
 
I'll write the parser to get an average time for QMDs, but it'll rely on people to upload results only from idle machines. I can't tell the difference between "I played a game during this frame" and "This frame took extra long". Unfortunately, the Mind::Reader interface hasn't been completed yet. ;)
 
unhappy_mage said:
I'll write the parser to get an average time for QMDs, but it'll rely on people to upload results only from idle machines. I can't tell the difference between "I played a game during this frame" and "This frame took extra long". Unfortunately, the Mind::Reader interface hasn't been completed yet. ;)
Click to expand...


Any way to id the computer so people can mark them as idle or partial use....Although with enough input, and some graphing we should be able to id computers that aren't in an idle state......
 
GAWD!!!!
Once again y'all have taken a fairly simple thing and crafted it into a goddamn major project, probably requiring a $10 billion dollar funding line!!
I was just attempting to start a SIMPLE thread to collect some data points.
Other than OSU everybody has contributed CRAP to this thread.
 
Hey, screw you, okay? I know it'll take some setup, but once I get it going, you won't even need the simple thread! Just mail your logs somewhere at the end of each week, and we automatically get stats!

Not trying to get confrontational, but I'm working on a solution. Saying I suck, and should be doing something else, is throwing the glove at my feet.

I would give you some data points right here, except that right now I can't due to being behind a proxy so I don't have any WUs to compare. Since I can't fold right now, I'm spending my time on other things folding-related. If I set up a system to graph speed of clients, that will help the community in general and make it easier for you.
 
unhappy_mage said:
Hey, screw you, okay? I know it'll take some setup, but once I get it going, you won't even need the simple thread! Just mail your logs somewhere at the end of each week, and we automatically get stats!

Not trying to get confrontational, but I'm working on a solution. Saying I suck, and should be doing something else, is throwing the glove at my feet.

I would give you some data points right here, except that right now I can't due to being behind a proxy so I don't have any WUs to compare. Since I can't fold right now, I'm spending my time on other things folding-related. If I set up a system to graph speed of clients, that will help the community in general and make it easier for you.
Click to expand...
Yeah, you are getting confrontational.
It will be great if you get the "mail logs" to work, but we've been dickin around with this since the genome wars.
(note - mail logs, dickin - BAD)
I just wanted to get a thread where results were posted by honest people. It seems that Stan4rd is once again screwing around with the points such that they need their chain yanked. The same PIIIM that was getting 191ppd is now on a p724 for 70ppd, which represents ABOUT 1/3 less value to SCIENCE.

As far as you INFERRING THAT I SAID THAT YOU suck - I did not even think that you were worth my effort to pick you out from the crowd of other people saying what we could do but not posting any stats.
 
gnewbury said:
snip
Click to expand...

I suggest you go back and read your first post. Myself and many others saw it as a question posed to see how we could figure out a good way to submit the time it takes to finish WUs with different hardware. Ideas and the reasons for them were given on different ways to do this.

You did not tell people to post their results in a specific format to this thread. You also left out relevant information such as using the results only from idle machines which would have skewed any results you got if people submitted info from machines that were used for other things.

Not everyone has been around here since the beginning and do not know everything you have done, how you have done it and what ideas were used for something like this. There have been many new things come around since the old Genome project and many new people have come around that may have a different idea or a way to get something working that was previously not possible.

I doubt the people that want to help with this are looking to stick something to Stanford like it seems you are. We just want something to look at to see how certain machines perform with different WUs. It would come in handy when trying to decide on a new box for folding or how to setup the client to get the best production out of it. This has many more uses than just what you want. It's also something that would have to be updated rather often with new proteins or new cores or just new hardware coming out. Because of that, it would be better to develop a way that is easy to submit the information and easy to compile and distribute it.

Get off your high horse. There have been several ideas tossed around that may make something like this happen. If all you want are a few results to do your own little project, then state so and leave the rest of us out of it. Start a thread requesting the info you want and make it specific as to exactly what you want and how you want it. Do not jump on us because you left things open for different ideas and ways of doing things. We're doing nothing but finding a way to compile and present the information in a useful way that everyone will benefit from.

 
I only run 4 machines for F@H, cause it's all I got. They are all quality machines, the slowest is the AMD 2800+ Laptop. Check my points/WU. I average triple the norm of 44pts/WU. A dozen of my teammates do even better. They gotta be running herds of monsters.

I've been working on a little project to see how identical projects run at different GHz settings. I have a very overclockable rig I've been using for this experiment. Stock 3.2E P4. I can easily overclock it on air, thanks to a huge HS, and a very loud 5500 RPM fan. I've been able to run F@H at 3.92GHz on this rig(on air) boy does it get hot. Running incrementally increased speeds, I've noticed something peculiar. As you increase the clock, and the CPU by default, the PPG/H goes up. Keep in mind that all of my experiments have been run on 1140, and 1141 projects, yes gromacs. I've performed the OC's mid WU, it's a 48 to 53 hr. project, you have time. Anyways, I've noticed that from 3.2 to 3.92GHz, the PPG/H increases by .115. Hopefully somebody here can verify this increase. It has to be bus related. If I can figure out this silly clamshell for my Vapochill, I should be able to blow past the 4.2GHz mark, and see the further effects of the FSB, on folding production. A work in progress, but I thought I'd share at this tense moment.

 
If webspace is needed, I have more than enough to host anything. Hell, I think I still have the Paris Hilton (long version) video hosted there. :)

BTW, it's my server so anything goes.
 
mr_ouija said:
If webspace is needed, I have more than enough to host anything. Hell, I think I still have the Paris Hilton (long version) video hosted there. :)

BTW, it's my server so anything goes.
Click to expand...


Tell me the site.... ;)
 
yes... by all means.... PM a linky :D lol

back on topic... I'm here to help, in whatever way I can... right now, Excel is about as much as I can help... that and providing info about my 6 boxen... I don't mean to ruffle feathers, but I will admit to making things infinitely more difficult in an attempt to make things easier.... :D Just let me know how I can help...


Keep on Folding!!

 
Send me a PM and I'll reply with addy. :) It's the long One Night In Paris video.

Also, I have 1700gb/month of bandwidth on a server box so if it is desired to make it an online thing, php & mysql I could probably do myself, I can host it as well...
 
If anyone would like to take a look at the spreadsheet I've made to track my boxen, let me know... I sent it to gnewbury at his request... if anyone else is interested, just send me a PM with your email address...


Keep on Folding!!

 
Kinda on topic here. got home tonight 3.2E@3.6 has been undisturbed for a day or so. P1910 project ran start to finish in 20:37:48. Are the folks at Stanford tweaking the QMD core? This was run with version 1.02 (mar 7, 2005). Project: 1910 (Run 8, Clone 84, Gen 5).

Maybe my math is off, but 6ppg/hr? I've never seen quite that much production on a larger WU (450 pointer). The little QMD's have been great, but the big ones getting faster? I'd like to request a -QMD flag. Kinda curious to see if these things love AMD's the way they do Pentiums.

 
You must log in or register to reply here.
Back
Top