• 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.

Stanford getting greedy?

S

Strikemaster

[H]ard|Gawd
Joined
Mar 29, 2001
Messages
1,264
Anyone else wondering just how much power they've lost lately?

With the retirement of my Slot-A Athlon, I no longer have a sub-gHz machine in my little crabgrass Folding plot. Sure, the firewall is running a gutless 667mhz Celeron, but I consider it to be doing it's full job if I can read the [H] and it shunts WUs out the cable modem. If It eaks out a Tinker now and then, so be it. Nothing else in the place is slower than 1.2gHz, no FSB slower than 266 (tell it's an AMD haus?), 256mb minimum RAM in every machine, and I'm effectively DOWN in production. Sure, I've been hanging on point-wise, but watch tomorrow and Saturday. The trough in the graphs will make "A Perfect Storm" look like a breeze blowing across a millpond. :(

Case in point: my "flagship" machine is currently chewing on a Gromacs project, p1406_Q44x2. That boxen is no slacker; A64 3000+ @ 206FSB, 1 gig RAM and 10k SCSI. It will take three days (61:40:00) to complete, according to EM3. Other end of the spectrum, that Celery-stick has a "unknown points" p112_p678_L939_K12M_355K Tinker it got fed on the 13th, and the power company willing, it'll deliver it on the 22nd!?! I think mice have a shorter gestation period than that!

When did Stanford decide the gruntwork would get, well, so damned heavy? I can't imagine a typical user E-Machine on screensaver getting anything done in a month! :eek:

Discussion?
 
From a more science-related point of view, I'm just glad that they're continually changing/improving (hopefully) the work that is being done.

I would be more worried if they sent out similar WUs without variation.

At the end of the day, the Folding@Home project will depend on Stanford's analysis of a few good pieces or patterns of data. In other words, better quality and possible more complex WUs are potentially good.

On the other hand, an increase in WU size could mean that they're starting to brute force computations rather than taking more creative approaches.
 
I posted about this a while back. Stanford seems to be increasing the workload which I don’t mind, but, how much useful data are they actually getting back? Way too many people don’t leave their machines on 24/7 like we do. That being the case there will be more then one or two WU’s lost to shut down error. There will be a lot lost. Since they run the same WU X number of times for verification purpose it would seem they are working against themselves at this point.

To my narrow mind it would make more sense to break a big unit into say like 4 parts and get the results back faster, and actually get more results back. That make sense?

Someone (V Jay?) should realize that we (all the teams and individuals involved) are not an endless resource. If I’m going to volunteer to be used at least use me to best advantage.

Whip me, beat me and make me write bad checks, but don’t abuse me ;)
 
BillR said:
Whip me, beat me and make me write bad checks, but don’t abuse me ;)
Click to expand...

Calling all moose, calling all moose!!!!!

There's a blue light special on BillR's today.

:D :D
 
Yeah I know what you’re talking about and it’s been bugging me a little lately also. Like on my laptop, I do not fold all the time, but if it’s really cold in my room I’ll fire it up on there… but with these big jobs coming in, there isn’t really any chance of any computer ever finishing them unless the computer is 24/7 folding. Or how about this – I ran a LANparty this last weekend, and setup the dual xeon game server. The server is running 4 instances of Folding@Home (because that’s how many virtual processors it has) and was running from Friday night to early Sunday morning. None of the clients finished a work unit during that time, and since the only time that server is really used is during LAN parties, I guess all that work was for nothing.
 
i remember getting pissed about this a year ago or so.

It's as if they forget we give this production to them, asking nothing by stats and satisfaction in helping a great cause. Shelling out hard earned cash to run my own garden . Granted, my power is free and I have some free hardware... but still.

I'm now shuffling projects to afford Win 2000 or better so i can use F@H 5.03, so that I can run timeless wee point proteins on my slow(?) 1 gig machines so I dont have to worry about not making it. I'm sure there is a way to run wee proteins otherwise, but I dont dig into the program like alot do... I'm abit lazy in my philanthropy.

What can happen is that a person can get overwelmed knowing that he just simply doesnt have the hardware to make a difference. I'm begining to think a sub gig machine isnt worth the cost of running it if i want production. So does that means I get to spend money in a garden just so I can stay in the top 300 of the team, 6134 overall?

And Since I'm bitching... the new point system is garbage. Fold Fold Fold. I'm running up on 1000 proteins. A friggen thousand. I've been at this for a few years... passed and beaten to death by a new farmer/gardener on the new scale.

Example from Random team mate:

Total WU....... Total Score
697............. 72,963.00

me:
913............ 41,484.00

A user with no more production:
2,563........... 46,725.00


Anyone see anything odd there ?


Where is the fairness. These stats are braggin points for some, it shows how far you've come. And for a small gardener like me, it's been a fight.
 
Sorry, BillR, hadn't seen the article. Life's been a bit 'o fun lately.

It hit me when I looked again at my stats here , specifically at the 7 month average. The points were consistent with incremental upgrades, but the WUs have been trending downward.

I thought I'd see a boost with bringing the Duron dually and the AMD64 online. Aside from the heat in my basement, and a slight decrease in my fuel oil bill, nothing. :confused:
 
See, I had just the opposite reaction to the larger WUs. I felt like my machines were being under-utilized with the 33 point proteins. Granted, I do have a Win98 machine at 400MHz that is set to timeless WUs, and a big ol' protein isn't what it needs. The laptop I am on right now at work is a 1.13 GHz P3M, and I pulled in a monster Gromacs unit to work. EM3 tells me 5 days, and that will be relatively accurate as I bring it home and set it up to run FAH only. There isn't much I can do about it, and I realize that most people are getting the same units I am, therefore all stats should remain relative. But the real world is different, and some have good luck getting small WUs, some are gardening, borging, farming and then some are just plain murdering CPUs on a grand scale. I have chosen to participate, and I am free to stop at anytime.
Reading further into this forum, I saw this same type of issue discussed previously. Those of us who do this (I can only assume here) mostly for points/stats honors find the time-consuming BA WUs to be tedious. I agree, but how long will the trend last?
I just started folding at the beginning of this month, so I have no comparative basis for the new points scale you guys are referring to. Can you explain it to me just so I know what happened/changed?
I do agree with the point brought up here that smaller WUs could be completed by softcore folders more often with less shutdown error/loss. And at the same time, keeping them large enough to make a points difference should be built in to the thinking. And to those new folders coming in, the points and stats are most likely going to dictate whether they hang around or not. I would venture to guess that some build a new rig and want to flex their muscle to see what it can do. When it doesn't just rip ass and send them to the top of the chart overnight, do they stay? Probably not. Is there a happy medium to satisfy all users? We all have choices to make, and sticking through the BA WUs is my choice for now.
 
We may be stepping out of our area of expertise a little bit when we start talking about how big WU's should be and what kind we should be getting. I agree that work will be lost on the WU's that take a very long time to complete because it will take a top-end machine to get them done in a reasonable amount of time, and non-24/7 machines may never finish them. It could be that Stanford gives us WU's that fast machines of the time can finish in a reasonable amount of time, and it also could be that big WU's are the ones that are more capable of giving them the information they really need. A couple of years ago we even had WU's that took days to finish on the fastest machines - maybe they're just giving us bigger stuff now that we have machines capable of doing them. I don't agree with the way they changed the points values so that just about every WU is worth much more than it used to be based on the amount of time it takes to do it, but there's not much I can do about that. If Stanford gets more use out of the large WU's, then I'll have to defer to their judgement on which ones to send out to us.

I solve all of these worries by not paying a damn bit of attention to what WU's my machines are working on. Every now and then I look to see if the CPU is at 100% and to make sure there are no unsent WU's. Ahh, folding bliss. :)
 
My problem is not so much with the size of the protien or the point values.... it's more with the 50+min frame times... if I restart it for some reason (usually only by force, win update, etc), I could loose 49.99 min of work... everytime I see it start back up, it's starts at the last complete %...

the only other problem I have with the huge proteins are that they take forever to fold... I watch the points... I do this for science, but when I see my daily output sitting at a whopping 0 , my mind says "what's wrong!!"..... I guess I like to see the I'm doing something, and the points help me with that... when there aren't any, I don't get that "oooo a 700 point day!!!" feeling.... In the end, it doesn't really matter, the cure(s) will be found and we'll all know we helped in that...


Keep on Folding!!

 
NoXpert, one of the things some of the people complain about in the points department has to do with how Stanford has changed the value of work units and how they score them over time. While I have only been doing this since about February or March, and before that did a bunch of Genome, I remember people remarking on how their top end machines would take a day or two to fold some proteins and they would only get a couple of points out of them. In other words, instead of talking about points per hour like we do now, they were talking about points per day. For the people that started at the beginning and processed a lot of those WUs, their machines did a lot of work, a lot more than many of us newer people have done, yet they still get the same credit for all those old WUs. I can see why that annoys some of them. I'd be annoyed myself.

Personally, I don't see why the points values have changed the way they have. Sure it looks nice getting 200+ points for Tinkers when before they were worth around 70. But what about all those 70 point Tinkers that were done. They stay the same. I don't think the points system should have changed at all unless all the old WUs points values had been changed the the accounts updated also. The points system is stongly weighted towards newer folders rather than old folders.

I personally don't mind the larger WUs. Then again, since the system in my sig runs 24/7 and it's no slouch, I don't have to worry about not finishing WUs on time. My work machine (PIII 700) in time will be a different story. It takes 150+ hours to crunch a Tinker and that's the fastest processing WU it can get hold of according to points per hour or day. Luckily I leave the thing running 24/7, otherwise it might actually have some trouble finishing off some WUs. I do have a problem with one thing. It had a 277 point Gromacs core a while back and it took around 2 1/2 hours to process a single frame. You can guess how long it took to process that protein. I don't see how this machine ever got a WU like that. That should have been sent to a machine with some speed.
 
I've always wondered why I've stayed in the top 80 for the horde for so long with just 5-7 machines hammering away, when I see so many sigs claiming 40+GHz of folding power. I'm still quite confused by it, I've got less than 10GHz total of Athlon Xp/P3 folding(They are all going 24/7), but I still manage to stay high in the standings.

u_sandmanx
 
SmokeRngs, thanks for explaining what changed, and I agree with the size of the WU; my laptop gags on those big Tinkers, and I hate to not have it finish in time.

OSUguy98, I know what you mean about the frame times. Not very often do I have to shut the system down for any reason, but when I do, I find myself waiting that extra few minutes to get that last frame done to save my place. Had I used any forethought, I would have set the program to write a checkpoint at faster intervals.

Seeing the progress is important to all of us. The points/stats tell us we are doing something besides leaving that jet turbine on 24/7.

Mattman, I agree; we were merely speculating as to the worthiness of the large WU, I think. And I like your attitude towards it. I myself welcome larger WU (on capable systems). Makes me feel like my equipment is worth something!

I didn't mean to stir the pot, really, just dropping my two cents into the ring. We'll keep folding and the results will keep appearing. Them's the facts. Fold on!
 
NoXpert.... you can change those interval times (I believe)... edit your config file.... or stop F@H and run it with the -configonly flag....


Keep on Folding!!

 
What bugs me is the variation in Point per Hour per Ghz.
You can do one Gromac and its worth 1.7 PpHpG.
Do an other and its worth 2.3 or even 4 PpHpG for any of the p13** series.
I know Intel's and AMD's fold at different speeds but I think the percentage spread for each type of protien ( Gromac, Tinker, Amber ) should be smaller.
But I've only got AMD boxen here so I don't know how Intel's fold.

As for changeing the interval times it will only work for Amber cored protiens.

Tinker.
[00:00:31] Finished a frame (65)
[00:04:45] Finished a frame (66)

Gromac.
[10:00:48] Writing local files
[10:00:48] Completed 172000 out of 200000 steps (86)
[10:19:20] Writing local files
[10:19:20] Completed 174000 out of 200000 steps (87)

Amber.
[06:32:32] Completed 0 out of 100000 steps (0)
[06:35:34] Writing checkpoint files
[06:38:35] Writing checkpoint files
[06:40:37] Writing local files
[06:40:37] Completed 1000 out of 100000 steps (1)

Unless the checkpoints dont show up on default vebosity.

Luck.......... :D
u=Tigerbiten.gif
 
NoXpert brings up a decent point:

Could Stanford set a performance datapoint based on uptime/24hr.?

A human can parse the logfile and see when the computer was shut down / started vs. folding checkpoints, why can't they build code for the WU server to toss part-timers a bone (small Tinkers)? It wouldn't make a damn bit of difference to me with a dozen boxen howling away in BoomerHaus, but the guy whos gotta wrestle with power bills, or traveling laptop users like NoXpert, could get that "warm fuzzy feeling" of seeing results despite the interruptions.

Yes, it's a "supposed" cheat for farmers, but whoinell would deliberately pulse power on a farm to garner 33 point WUs?
 
Tigerbiten said:
What bugs me is the variation in Point per Hour per Ghz.
You can do one Gromac and its worth 1.7 PpHpG.
Do an other and its worth 2.3 or even 4 PpHpG for any of the p13** series.
I know Intel's and AMD's fold at different speeds but I think the percentage spread for each type of protien ( Gromac, Tinker, Amber ) should be smaller.
But I've only got AMD boxen here so I don't know how Intel's fold.
<snip>
Luck.......... :D
u=Tigerbiten.gif
Click to expand...

That's the core - PHG - I donate a cycle - I get a reward. That's why I don''t sneakernet anymore, no certain reward. I just float along. I used to paddle like hell - That's how I got to #1 for the [H] at genome@home. We busted butt to get Stefan to get a fair reward cycle.
Last time I had a chance to give a darn the new rules were totally capricious.
I keep em' folding, but borging is not worth the effort.
ps. what's amber ?
 
gnewbury said:
ps. what's amber ?
Click to expand...
New style core, different method of saving work unit segments (saves on percentage, and on time elapsed since last save). Drove LPerry to distraction trying to get a handle on averaging time elapsed / to-go in EM3, but he got over it :) .
 
Strikemaster said:
NoXpert brings up a decent point:

Could Stanford set a performance datapoint based on uptime/24hr.?

A human can parse the logfile and see when the computer was shut down / started vs. folding checkpoints, why can't they build code for the WU server to toss part-timers a bone (small Tinkers)? It wouldn't make a damn bit of difference to me with a dozen boxen howling away in BoomerHaus, but the guy whos gotta wrestle with power bills, or traveling laptop users like NoXpert, could get that "warm fuzzy feeling" of seeing results despite the interruptions.

Yes, it's a "supposed" cheat for farmers, but whoinell would deliberately pulse power on a farm to garner 33 point WUs?
Click to expand...

Thanks; you put what I was thinking in a more understandable way, nice job. I have 17 CPU’s; all over 2 gigs and they run 24/7, much the same as many of you. Others can’t do that which is what I was discussing earlier but perhaps not quite so clearly.

I would think they could monitor total continuous output per machine (each client has a unique ID number) and collect data on that and use that to reassign the load more evenly. Perhaps even assign specific projects to specific computers. This could create a “fine tuning” of things so to speak.

How hard could it be to add some code to determine what CPU each client is using?

As it stands we (all the teams and members) are throwing gross gigahertz at the project in a sort of wily nilly fashion. This is quite wasteful in my opinion. There are other projects forming who want to share the resources (they claim there is no competition for folders, I don’t buy that for a moment) and as such it would behoove Stanford to make some common sense adjustments to their methodology. Face it; this too is part of the science.

UD attempted to “categorize” users by using methodology I never did agree with, amount of RAM, connection speed etc, all very unscientific in my opinion. Your network speed had little to do with your computer’s math capabilities.

Ok, enough of my late night rant, I just hate military style inefficiency. I know Stanford is a liberal institution, but damn………get with the big picture. ;)
 
Tigerbiten said:
What bugs me is the variation in Point per Hour per Ghz.
You can do one Gromac and its worth 1.7 PpHpG.
Do an other and its worth 2.3 or even 4 PpHpG for any of the p13** series.
I know Intel's and AMD's fold at different speeds but I think the percentage spread for each type of protien ( Gromac, Tinker, Amber ) should be smaller.
But I've only got AMD boxen here so I don't know how Intel's fold.
Click to expand...

See this link, it might clear some of the confusion regarding points differences.

http://forum.folding-community.org/viewtopic.php?t=9104

As far as Intel vs AMD, my p4 2.8c @ 3350 mhz generally produces 5% more points than my athlon xp-m 2500+ @ 2550 mhz, all from hyperthreading.
 
ezee said:
See this link, it might clear some of the confusion regarding points differences.

http://forum.folding-community.org/viewtopic.php?t=9104

As far as Intel vs AMD, my p4 2.8c @ 3350 mhz generally produces 5% more points than my athlon xp-m 2500+ @ 2550 mhz, all from hyperthreading.
Click to expand...


ezee,

Thanks for the info. I was a little worried about my CPU overclock when that 300+ point Gromacs terminated early, but it also explained that they have a tendancy to do that. Guess I can cranck the Hz back up.



Edit to spell Ezee's name right.
 
Hito Bahadur said:
ezee,

Thanks for the info. I was a little worried about my CPU overclock when that 300+ point Gromacs terminated early, but it also explained that they have a tendancy to do that. Guess I can cranck the Hz back up.



Edit to spell Ezee's name right.
Click to expand...

Keep an eye on those terminations, once or twice a month is ok, after that I’d back down one notch at a time till it stopped. I see an early termination only once in a great while.

Luck
 
Thanks,

That was the first WU it ever received. The next two have been stable, however I did drop 100 MHz off the overclock.
 
You must log in or register to reply here.
Back
Top