Let's sit down and discuss...

[Macaholic]

Just to let everyone know, changes can be expected in point output. One heated thread (http://forum.folding-community.org/viewtopic.php?t=17725&highlight=) in particular has brought the "patty melt complaints" to a head. So, one shouldn't be surprised by getting less points than what one would normally expect from certain units. I'm sure we will see several threads pop up in the coming weeks. You can find a few more details here (http://forum.folding-community.org/viewtopic.php?t=17721&highlight=). I would say certain practices have reached critical mass. I'm sure more changes are to come. Now discuss...

[dz_jad]

**bump**


errr....yah, even though it was the top when I came to it :p

you can tell the guy accused of cheating in the second link is just that. The way he responds to ChasR is like a FanBoy. He knows nothing about folding, or borging, or anything, and is quick to respond with accusations. yah. watever,.

[godofgorks]

Dumping in Horrible!

Server blocking/filtering is wrong!

FOLD FOR SCIENCE!

Would there be a possible way to setup a server system that server blocking could not happen. Possibly by using gateway(s) of servers or some sort of redirect so that there is a common source that cannot be block and is a pathway for all WUs?

[William Hung]

While I might have alluded to such earlier, if everyone is getting patty's, then it shouldn't matter. I've swtiched my main boxes to linux as of last night, but up to that point, I'm positive i've done more than my share of patties! And besides, they keep my boxes nice and warm! :)

I think I still have 4 of the Borgs on XP and a check shows 3 of them have patties!

My Toaster and cel Phone even have patties! Time per Frame. 8.8! Years.

[dz_jad]

lol william.

yah, that other guy has wierd logic, "its all about me" logic.
"I'm leveling the playing field" he says...wtf? by taking all of them? he's just messed up. People like him get talked about too much (and I ain't helping). People....bah...just bah.

This is just an important topic. I dont think people realize how much dumping messes up the science. if a project takes 2 days to crunch and has a 53 day deadline....that messes things up.
yah...
[/rant on many subjects]

[sno.lcn]

Cheating is stoopid.

[Rav]

They should just raise the point value, encourage people to just finish this project, and put all this controversy behind us. Raise it to 600 points I say. Then there is no reason to cheat and people will let them finish. What does stanford care about who gets the most points? As long as the science gets done and people are happy.

There have been WUs like this in the past, but the melts are particularly bad because they take so long and ruin the PPD of high end CPUs like C2D's. If people still ran P4's and XP's, we wouldn't have even taken note.

[Sleepy_Steve]

A few things...

A few of my old chips... 1ghz roughly i find to not even be worth folding. Less than 100 PPD out of all three of them. So they do nothing atm. Although I was considering using one of the cases for a Pint D 805 / 820 based SMP box.

My other rigs running more modern 32 bit single core chips... P4 Presscot @ 3.6, and even the Pint M notebook, and XP 2400+ do put out a small but worthwhile PPD. But to me, its not worth the effort to switch them over to Linux just to avoid melts.
As for my conroe... It crunches on its fair share of melts because i use it a lot for gaming and other work. But if i really want me some good points, i fire up Ubuntu / SMP and do work on my laptop so i can get 2k + PPD from one rig.


But i always wondered why there didn't seem to be melts on Linux. It didn't seem right to flood all the windows clients with such poor producers. Personally it almost seems as though Stanford is trying to fight some sort of points inflation. Frankly if thats what they are trying to do - Its the dumbest thing ever because its free for Stanford to award ever increasing points to keep all the stat monkeys happy.

[sno.lcn]

At this point, raising the point value just justifies the cheating, like something good came out of it and the cheaters got what they wanted. In my opinion, the people cheating should have their points zeroed. I think that would be the best means of discouraging future cheating. Even if those cheaters who lost all their points were to get mad and leave the project, getting the WU's done in the correct sequence by the rest of us would probably be better for the science anyway because they put those WUs out for a reason, most likely because they need US to get them folded!
And I know it's been said a million times before, but it bears repeating: People need to stop whining so much about these WUs because everyone is getting them, so the point loss averages out between everyone! I don't know why this concept seems so hard to grasp for so many but it's not exactly rocket science. It's like people just refuse to believe it, and think they are getting screwed somehow. I can understand people wanting to get better WUs and hell, these things aren't my favorite either, but the faster we finish these, the faster they will go away! And if it's that big of an issue, go smp or gpu. Don't wanna switch to linux for smp? Use vmware, there's plenty of threads around here on how to set that up.
And to those who only care about having their names on a stats page and don't really care how much science gets done (ie. cheaters), go find another project to crunch for.

*This is not necessarily directed at anyone here, I love you guys!

*Edit: And if it would help any, instead of dumping the pattys just send them to me and I'll stop downloading my own work and fold those instead. But I have to finish the ones 2125's I'm already folding first :D

Is anyone here dumping them anyway? I highly doubt it :)

[Rav]

I wonder if people realize the GPU_lambda's are the same as the lambda_melts. It makes the CPU version seem like an even bigger waste of time since GPUs are doing more work in less time at 6x the PPD.

[benbaked]

Well said sno.

[Wiggles]

Is anyone here dumping them anyway? I highly doubt it :)

Not I. I just let em' ride. They eventually finish, even if it almost takes 4 days on my rigs here at work. Oh, and on a good note, I finished my 1000th WU yesterday :bday:

Wow! Completed 44 WU's between just the 2124 & 2125 (don't look at that stats page often).

[dwschoon]

It seems that there are cheaters in just about anything. I quit playing some online games because I got tired of everyone cheating. Those are just games though. FAH has real world implications and it makes me sick to think that people would risk not finding a cure or a drug that will help people, and for what, a few extra points on a stat sheet? I have an idea, why dont they stop assigning all wu's except for patty melts until they are caught up and people quit complaining about the points. Or they could reduce the points of all the other wu's until they equal the output of the patties. While this isnt feasible, and it isnt fair to hurt the other projects, sometimes you need to take a hard line approach in order to make people realize that points are secondary. Also, I agree with sno, zero the cheating punks.

[Leviathan41]

Unfortunately, the only true solution to this problem is for Stanford to adjust the point values for various WUs. The ideas listed in the F-C.org thread may help to an extent, but not as much as changing the point values would. It's sad, but if Stanford truly wants to get the best results (for science) from the FAH project, they will have to adapth and stop this dumping/blocking trend by making it not worth doing.

I fold for science and realize that the points are just a way to keep things interesting. However, not everybody thinks that way and something has to be done to keep this from happening.

It's a shame that people are doing this, but I don't see any other way to truly stop it.

[WarriorII]

Very well said Sno. :thup:

These patties are just like thos old 44 pointers we use to get. :bang head

If we can just get through them.... they'd be gone. :confused: :rolleyes:
-go figure.


Cheaters... BAH! Zero their points!

Fold On Brotha's !
:attn:

[marty9876]

Well, I would but the FCF mods no love me anymore! :) Hopefully Mac still has a place in his heart for me.

I still think the whole mess is from the patty melts being too long in duration, nothing with regards to PPD. From a PPD, they are actually just fine and inline with all the other regular 130 point Gromacs/what not. It's just the one takes a day and the other takes 4 days and people get sick of them. Had the patties been of the same size/durations of all the other Gromacs WU's this would never have been an issue.

Will the tet point decrease fix anything, I doubt it. But good for Kasson for having a big pair of balls if nothing else! :) In the end there never will be a perfect solution, pick a plan and stick to it. I'm getting to the point where I've had just about enough of all this. I don't even look at my daily totals anymore, just reading all this crap is just getting to me.

[Sorin]

Version 6 being able to detect hardware better and act accordingly sounds nice at first. But I'm worried it's just going to give the melts and other 3-day-WUs to the fastest processor families under the belief that they'll get done faster.


I still think the whole mess is from the patty melts being too long in duration, nothing with regards to PPD. From a PPD, they are actually just fine and inline with all the other regular 130 point Gromacs/what not. It's just the one takes a day and the other takes 4 days and people get sick of them. Had the patties been of the same size/durations of all the other Gromacs WU's this would never have been an issue.

That's a big part of it. Most work I've gotten takes significantly less than 1 day. That's fine. Ribo projects take 2 days, but since they're worth 600 points, that's fine too. Melts take 3 entire days however, and are worth jack in comparison (insert old and tired "they're the 'normal' work units, get used to them" speech). It's kind of annoying, but they need to be done. The biggest reason I joined the project was to use my computer's power for something worthwhile. So since melts are obviously a worthwhile science venture for stanford, I leave them alone and let them melt away on my boxes.

Will the tet point decrease fix anything, I doubt it. But good for Kasson for having a big pair of balls if nothing else! :)

Run that by me again? They did *what*? I went and checked the summary page and lo and behold, they're worth less than 300 points now. I hope there was a legitimate reason and not just so people will like them less or something. If the latter part is the case and they're going to play that way, they can darn well raise the points on the melts. If it's the former however (like a re-benchmarking thing, as for 1498), then nothing to see here, carry on!

[ChasR]

The first step were going to see as a result of the cheaters is a reduction in the bonus for large memory requirements. P149x and 2006 today went from a 100% bonus to 50% bonus. A C2D will still produce 200% more than the benchmark in terms ppd/GHz but a socket A or P4A, C, or Early E is going to produce only 50% more. There will still be a large incentive for the C2D to dump normal WUs so not much will be achieved. Zero the cheaters points and it will stop.
Marty is spot on on the duration of the p212x simulation. If it were half as long, the patty melts wouldn't be such a problem.
Raising the points on normal WUs devalues everything done to date and many of us would be opposed to that. Eliminating or drastically reducing the bonuses is more palitable.

[Leviathan41]

Raising the points on normal WUs devalues everything done to date and many of us would be opposed to that. Eliminating or drastically reducing the bonuses is more palitable.
That's a good point. It sounds like reducing bonuses are a better plan than raising points on the WUs being dumped. I do think Stanford is going to have to adjust the points in some manner in order to stop this problem.

[William Hung]

What %-age do most people get for patty's on XP?
I'm at 500 WU's and about 20% have been Patties. Actually it'll be a little higher %-age since I run one mac and a linux SMP box.
P2124 55
P2125 34

20% really isn't that bad. 1 in 5...

[leelegend]

never dumped anything, you can never guarantee what is coming next anyway, so what is the point?

i think it would be better if they cut the amount of frames on the Pattys, just so people can see a bit of movement....i can understand the frustration, but it is still not justification for messing with the whole project!

i dont think it should be tolerated at all, you sign up for all units, not just high scorers, i have mainly mac's so i very rarely get bonus units, you dont hear me crying about it.......zero the points and kick them off.....

lee

[jws2346]

never dumped anything, you can never guarantee what is coming next anyway, so what is the point?

i think it would be better if they cut the amount of frames on the Pattys, just so people can see a bit of movement....i can understand the frustration, but it is still not justification for messing with the whole project!

i dont think it should be tolerated at all, you sign up for all units, not just high scorers, i have mainly mac's so i very rarely get bonus units, you dont hear me crying about it.......zero the points and kick them off.....

lee

Yeah, I have to agree, what is the point in dumping your first WU because you never know what you're going to get next :eek: , in my opinion "not too smart".
Now the blocking server thing or whatever I know nothing about :shrug:

[leelegend]

how about if they change the way the flag system works?

make -advmethods dish out nothing but patty's and for that you get a bonus for the amount of time they take.......every other unit gets a fixed lower amount, but based on the benchmark make the same amount of points over the same amount of time.

so if you dont want your machine tied up for days on 1 unit fair enough.....if you dont mind, you get the bonus.

that way, if you sign up for it....shut up and get on with it????

or split the patty's up into different length units so that everyone still get's pattys, but -advmethods get longer units....and so the bonus (a bit like the ribos).

lee

[marty9876]

Even if you took all the bonus away from P149x wouldn't they still fold just about the best (PPD wise) on a C2D system anyways? Just a huh thought.

One note, the larger the WU (patty melts being larger so to speak) the less disk space the results take up. From Stanford's end, this equates to a heck of a lot of disk (spendy server disk) space. Larger in duration WU's save a ton of disk space on Stanford's end. Not saying good or bad here, just this could be a motivating factor.

[dz_jad]

Marty is spot on on the duration of the p212x simulation. If it were half as long, the patty melts wouldn't be such a problem.

yeah, not just with points, but because they take so long, any my machine has been transient the past month and a half, I've had a handful EUE on me. Actually, I just turned in a 3 point WU, no clue what that one was :p

[pscout]

We have had other somewhat tiresome projects ... but nobody complained as much since the wu's finished a lot quicker.

Solutions to the cheaters via desing and programming on clients and assignment servers are not easy ... and rightfully Stanford is likely favouring develpment work on the scientific end instead of the logistical aspects that address policing.

So like the rest of our world, we can't really change the fact that some percentage of us need some degree of policing and also that we can never afford the overhead of enough of it. Competitve creatures that we are, there still needs to be an incentive system to maximize out efforts. Reducing the incentive to cheat also risks reducing the incentive to borg and farm.

So there just is no perfect answer :shrug:

But i for one would contribute to some effort on an effective way to reduce the impact of the cheaters on the cause.

Publication of the offenders would probably be a start, and some points zeroing or fines ... too many people see others getting away with it so they 'forget' that it is bad to copy them. Most people do not want to be considered a cheater by their peers ... except the hopefully very small minority who might feel proud for being a better cheater than the next person.

And like any good justice system, there needs to be a balance. I am sure a reminder to some who get overzelous is enough but others who who are incorigible need to leave.

Minimizing the reasons for people to forget their ethics, and reminding them from time to time of the purpose of folding is probably the best approach :)

[LandShark]

The first step were going to see as a result of the cheaters is a reduction in the bonus for large memory requirements. P149x and 2006 today went from a 100% bonus to 50% bonus. A C2D will still produce 200% more than the benchmark in terms ppd/GHz but a socket A or P4A, C, or Early E is going to produce only 50% more. There will still be a large incentive for the C2D to dump normal WUs so not much will be achieved. Zero the cheaters points and it will stop.
Marty is spot on on the duration of the p212x simulation. If it were half as long, the patty melts wouldn't be such a problem.
Raising the points on normal WUs devalues everything done to date and many of us would be opposed to that. Eliminating or drastically reducing the bonuses is more palitable.
I didn't have time to read the whole FCF's threads yet. but according from what people discussing in here, I'm 100% agree what ChasR and Marty said!!

(now need to spend time to read the whole threads.....)

[ChasR]

It's a 12 page thread to fix a problem basically caused by one type of WU on one type of processor. It could be fixed by making more bonus type projects so that eveyone that wanted them had a steady diet of them. No reason to dump, no reason to cheat. It could also be fixed by not making anymore projects that work like p1499.

[Silver]

It also could be fixed by upping the points on the 21xx series and there, done, no more issue. Gee, that was easy.:)

[ChasR]

Silver,
I disagree with you on raising the points for a perfectly normal WU that uses less resources than any WU in recent memory. It is absolutely unnoticeable running on just about any machine, as it should be, and thus deserves no bonus. Raising its point value would devalue everything all of us have completed to date.
BTW, I'm fine :beer:

[AlabamaCajun]

That blew me away all those patty-melts in teh trash :cry:. After reading that, I rescued 1 patty and 1 vcillin from my AM2 dualie that I converted to SMP linux. I was not dumping these but I waited until the machine sent in prior finished work units but then it downloaded these 2 just before shut down. 1 is folding away now and by Sat. I'll queue up the other. I still have 3-4 boxen and borgs munching on patties which is fine they make nice point bumps every few days.

[muddocktor]

I just replied in that thread at the community and this is what I said:

Either drop all cpu work based bonuses or make and adjustment to the present system on regular wu's by assigning a bonus for quick completion and return of regular wu work. If you finish and return a regualr unit in 100-97% of the preferred deadline, give a 15% bonus to the wu. 96-92% of the preferrred deadline gets you a 10% bonus and 91-88% gets you a 5% bonus on the wu. This makes the quick return of any regular wu a real and noticable points-earner. and if they combine this with a drop in assigned bonuses given (like they have just done with the 149x/2060) it makes the points disparity much less between the patty melts and the bonus wu's.

It's not like Stanford hasn't ever made any adjustment in the points system. When I started participating in the project in 2001, a large wu was worth 2 points and most were worth under 1 point. So there is precedent for change.

EDIT: And this system would also reinforce the theme that all work units are important and not just the ones with BigWu bonuses attached.

[jws2346]

I just replied in that thread at the community and this is what I said:

Either drop all cpu work based bonuses or make and adjustment to the present system on regular wu's by assigning a bonus for quick completion and return of regular wu work. If you finish and return a regualr unit in 100-97% of the preferred deadline, give a 15% bonus to the wu. 96-92% of the preferrred deadline gets you a 10% bonus and 91-88% gets you a 5% bonus on the wu. This makes the quick return of any regular wu a real and noticable points-earner. and if they combine this with a drop in assigned bonuses given (like they have just done with the 149x/2060) it makes the points disparity much less between the patty melts and the bonus wu's.

It's not like Stanford hasn't ever made any adjustment in the points system. When I started participating in the project in 2001, a large wu was worth 2 points and most were worth under 1 point. So there is precedent for change.

EDIT: And this system would also reinforce the theme that all work units are important and not just the ones with BigWu bonuses attached.

Yeah, I can see that, I may be cutting my own throat, but making it a point bonus across the board for early completion of WU's certainly, in my book, makes a lot of sense, no more trying to pick out the best WU bonuses, no more of this certain server blocking stuff, etc. :thup:

Back in 01' sheesh, they didn't even have trees back then (only kidding :p )

[Spion]

Just fold them!!! I'm been scarfing down melts for while now. they're not the best to eat but its all the cook will serve me :beer:

[ihrsetrdr]

It is discomforting to me that this issue has even come up, but human nature being what it is, this was bound to happen. What-ever Pande Group decides to do about the dumping, I hope it is sucessful. I would like to see 'equity' in the points system that takes into account the relative CPU time spent on a given WU.

Regarding the patty melts, I wonder how much new code would have to be written to run them in SMP. Perhaps it would only be a matter of "plugging-in" the specifics of the melt simulation to the SMP architecture...shortening up the completion time would surely make them less painful and less prone to dumping(in some cases).

[ChasR]

I'm not sure eliminating the bonus entirely would put an end to cheating. With NO bonus an E6600 makes 220ppd/GHz on p149x compared to 63ppd/GHz on p212x. That's 157ppd/GHz/instance gain (It's probably a somewhat less than this in actuality when folding two instances of p149x due to cache contention and bandwidth limitations). I recall when the ribos ran out, folks with fast AMDs gobbled up the timeless tinkers to gain 25ppd/GHz over normal gromacs. The bonus reduction won't have the desired effect.

Ribosome project points were reduced from 600 to 500 points today.

[ChasR]

I don't agree that the time it takes to fold a WU should affect the points awarded. The time it takes to fold a p212x has little to do with the complexity of the WU or the resources used but much to do with the number of nanoseconds being simulated. It takes 14 days for one of my p3s to fold a patty melt, should I get a boatload of points for that? I think not. All my rigs fold 24/7 what does it matter how long one simulation is compared to another as long as I'm awarded the same ppd? If I turn in a 100 point WU in 1 day I get 100ppd. If it takes me two, I get 50 ppd. The faster I turn them in the more points I get each day.
I don't think the value of a perfectly normal WU should go up and devalue everything I've done to date just because C2Ds make too many points on one type of WU.

[Silver]

Silver,
I disagree with you on raising the points for a perfectly normal WU that uses less resources than any WU in recent memory. It is absolutely unnoticeable running on just about any machine, as it should be, and thus deserves no bonus. Raising its point value would devalue everything all of us have completed to date.
BTW, I'm fine :beer:

Good to hear on your well being. Been awhile.:beer:

I would agree on the resources though I find little (that stanford does not send out with 'unknown errors') that is really all that resource heavy. New systems have kind of taken care of what was considered resource heavy so maybe that should be redefined. I would how ever argue that time in and of itself is a resource that has a value. If a project takes 1.5 days to do based on whatever and on another machine takes 1 day then I would argue that time has value (even more so in science) and that the one day should get an example base of 100 points for the days work while the first should get 150 for the 1.5 days work (all things being equal on the computers).

Point taken ChasR. Agreed. Mr Mudd seems to have a partial worthy solution but the disparity in time has to be resolved.

The baseline cpu is just wrong. The p4 @ 2.8 is outdated.

[psyshack]

Its a shame this even has to be chatted about.

I will keep folding and play by the ?? un writen rules I guess you would say. As I see it fold what you were served and like it or shut your machines down.

Shoot I have problems with the program myself. But I found away around it. I know the drug company's will make billions off my hard work and electricity down the drain. So I went and bought stock in a few company's.

So damit fold what is sent to you so I can get rich!!!!!!! :)

psy

Humm would that be insider trading? eh

[Silver]

psyshack, why did'nt I think of that.

Anyone know what the requirements are on the 3028 to 3037's? Love to get hold of some of those.:beer:

The way to morally duck the 21xx's, run smp/gpu. Works and no one questions your integrity.

[Seven]

It saddens me to see people getting ahead like this, but I guess it's bound to happen.

Truthfully, I think that in a way they need to standardize the points system with some careful thought. The bonuses began to coax people into folding bigger units that took up more system resources (remember when QMDs came out?), those kinds of things became a major drain on a casual folder unless they were running a nice rig. I think yes, certain things should be worth more points, but the system should basically give premiums on the units they need the most and the ones that take the most processing power to do. Simple as that.

7

[ihrsetrdr]

I would like to see 'equity' in the points system that takes into account the relative CPU time spent on a given WU.

And then

I don't agree that the time it takes to fold a WU should affect the points awarded. The time it takes to fold a p212x has little to do with the complexity of the WU or the resources used but much to do with the number of nanoseconds being simulated. It takes 14 days for one of my p3s to fold a patty melt, should I get a boatload of points for that? I think not. All my rigs fold 24/7 what does it matter how long one simulation is compared to another as long as I'm awarded the same ppd?

Perhaps my statement was a bit too vague; wasn't advocating rewards for slower production at all, just equity as in time=resources, basically.

[jws2346]

I don't agree that the time it takes to fold a WU should affect the points awarded. The time it takes to fold a p212x has little to do with the complexity of the WU or the resources used but much to do with the number of nanoseconds being simulated. It takes 14 days for one of my p3s to fold a patty melt, should I get a boatload of points for that? I think not. All my rigs fold 24/7 what does it matter how long one simulation is compared to another as long as I'm awarded the same ppd? If I turn in a 100 point WU in 1 day I get 100ppd. If it takes me two, I get 50 ppd. The faster I turn them in the more points I get each day.
I don't think the value of a perfectly normal WU should go up and devalue everything I've done to date just because C2Ds make too many points on one type of WU.

Hmmm? Would you agree that the point system fosters competition? Would you agree that in competition (especially sports) that the biggest, the fastest, the strongest are the winners and are generally more rewarded ? (ie $, fame, etc or more literal the ones with the most points?) What are they gonna do tell all major league pitchers to only pitch so fast because it's not fair to the less fortunate pitchers (notice I said "less fortunate", not "less talented") I think, it's only my theory, that mankind striving for some goal is more successful and reaches that goal faster through competition. Oh yeah, to coin a wore out phrase I think you can support science and be competitive. (hey, what do all those heart surgeons know trying to be the best and all those college students trying to get the top grades?) :)

Only my $0.02 cents worth (without inflation) :)

[nahmus]

I don't usually chime in on stuff like this since my ppd has gone down and I'm way out of touch with the cutting edge setups for max PDD and what the projects are. I do agree that the point system fosters competition its just human nature. the point system has helped the project more that it has hurt ( i think :D ).

I think that the project has not kept up with the pc side of it and the human side of it. They are concentrating on "the project" but the project has grown and like true nerds have not been keeping up on the human side. The points system should be overhauled to reflect what is going on in their userbase in terms of hardware. If they cannot or rather do not want to manage the user side of the project then they should get people (elections of users?!?) that are in touch with the human side of the project and can keep an eye on whats going on and can report back to the Pande group and make recomendations. They need to keep the project balanced. ignore one side or the other and the project as a whole suffers.

Just me out of the loop 2c

[muddocktor]

The reason I say for Stanford to give a bonus for timely return on work units is because they are always saying the sooner they get results back, then the sooner they can get more work out and the sooner they can see how the series is doing. This was one of Stanford's arguments about only folding with 1 client on HT-netburst processors. But there is presently no incentive whatsoever to actually return work as fast as possible. As long as I beat the preferred deadline by 1 minute I get the same amount of points as by returning it with 99% of the preferred deadline remaining. With the machines I now fold with, I could run multiple clients on each core and still beat all the preferred deadlines. If you want timely return of the wu's, then put your money (or extra points in this case) where your mouth is, so to speak. Doing something like this insures that the greatest majority of folks will try to run their clients with the fastest return on work possible rather than running multiple clients per core/processor.

[Silver]

I do remember early on thinking that I would get fewer points as I had taken too long turning some work units in. I just naturally assumed that this would happen.

[TollhouseFrank]

people get too hung up on the point system.

[psyshack]

Lot of good comments in this thread.

It would seem our team,, The best in the world I mite add. :beer: Is very grounded and commented to the effort. This fills me with pride!

I can understand the poster in the thread that feels he/she has a rite to block and fold what and when they want. Really I do. You have a few newer higher end machines laying around the house folding. You have some money tied up in the effort. Not to mention time. Does that make it right? No,,, but I can understand it. Stanford has the largest supercomputer basically in the world doing work for them for free!

Yeah know a gentalman down the street died of cancer around 12:30 am today. His wife gave him a drink of water. He said thank you honey and passed away. It took him over two years to pass on because of his tumors. It has consumed the family. They don't have the money to bury him. I will see to it that he is buried and take care of the family's needs at this point.

Deep in my gut I'm sick of watching friends and family die of cancer. SICK OF IT! Makes me want to block all work units but the ones that pertain to cancer. Would that be right as we know it in the project? No. But it would stand up in the long run in the court of public opinion. If this debate was ever seen by the mass of humanity.

The project attracts all types. You have the folks purely interested in the science. Peeps folding for humanity. And yes the points. IMHO you have to have a ground in all three at some level. Or you wont stick with it. Or you will become seen as a problem by the group as a whole.

Ive been around seti and fah long enough to have seen this issue come up before. It always seems to leave a bloody gash in some users. I'm sure at some level it hurts the project. But at what level,, who knows.

I hope as a team we can stay focused. Weather the storm. And move on.

psy

[benbaked]

It would seem our team,, The best in the world I mite add. :beer: Is very grounded and commented to the effort.

Indeed, we have a great group of people here. Forget the [h] and Google, we ARE the best team. :D

I hope that this whole points thing get straightened out and people don't quit folding in disgust or whatever, but if people quit folding because of the points then whatever...that's too bad, but I don't think there will many on Team32 that will give up folding because of that reason.

I'll fold melts all through '07 if I have to, I don't care. I am not shutting down my computers.

[jws2346]

Indeed, we have a great group of people here. Forget the [h] and Google, we ARE the best team. :D

I hope that this whole points thing get straightened out and people don't quit folding in disgust or whatever, but if people quit folding because of the points then whatever...that's too bad, but I don't think there will many on Team32 that will give up folding because of that reason.

I'll fold melts all through '07 if I have to, I don't care. I am not shutting down my computers.

Touche' benbaked, I have no doubt that Team 32 is the best out there and I do agree with you we have a great group of people. I wouldn't worry too much about some people becoming disgusted with this point deal, I think that's probably human nature (look at how many countries were founded by disgusted people, people disgusted with their government)
Nope, I don't really see many people quitting folding, I think the true folders are too stubborn to give up and most have deeper seated desires to see cures for these diseases (some have lost loved ones) Maybe someone will stop folding because of this point dispute, rather than trying to work it out like adults, they'd rather "take their marbles and go home" like children, unfortunately there's a 1/10th of 1 % in everything. :)

[ChasR]

Here's an example of what happens if you change the benchmark machine, as some of you have proposed, to a E6300 (1.86 GHz) which makes approximately 65ppd/GHz/instance , 121 ppd/instance on Gromacs. The benchmark ppd becomes 121 ppd up from 110. The 6300 produces a p1495 in .59 days. The appropriate non bonus point value of p1495 becomes 72 points. Lets say we use the current 60% bonus and give p1495 a value of 115 points. The E6300 makes a respectable 195ppd/instance, down from the 610ppd it made on one instance before the bonus reduction. But how does this affect the other machines running p1495. An A64 @ 2.4 takes .82 days to complete p1495 so it would make 140ppd/instance, down from the 444ppd/instance it used to make. The current benchmark machine takes 1.67 days to complete p1495 so it would now make 69ppd, 41ppd less than its own benchmark on normal WUs, down from 220 ppd it currently makes. An A XP (Barton) @ 2.0 takes 1.46 days so it would make 79 ppd, down from the 250ppd it once made. Changing the benchmark machine will change the "inequity" from one type of machine getting too many points on one type of WU to most machines getting far too few points on p149x.
@jws
Sorry, but I don't have a clue how your post relates to my quote.
@ HT
So you're saying that p2124 should be worth more than 392 points because it takes twice as long to fold ss p3038, even though they produce the same ppd and your machine didn't run one minute longer to produce 392 points on 1 p2124 than it would to make 392 points on two p3038s? You furnished no more resources and incurred no more expense. You should get no more points in my book.
@muddok
A days work deserves a days pay. If you do it in 1/2 a day you make a lot per hour. If you do it in a month, you make very little per hour. The points system works exactly like that and I don't see anything wrong with that. KISS principal at work. Early completion bonuses have been discussed before and dismissed as too difficult to implement serverside and the complaints about points awarded would be never ending.

[leelegend]

any objection to splitting the patty's into smaller junks?
or how about getting a SMP core up and running that could chew the patties.... :)

lee

[ChasR]

I proposed about a month ago that future normal gromacs be no more than half the duration of of the pattymelts. The newest gromacs are pretty close to that. I have no idea if that is a result of my comments, a coincedence, or they already had that figured out. Of course half the duration means it takes twice as many WUs to complete the project. So we each spend the same amount of time folding them.
SMP is geared towards simulating much more complex and demanding projects than the pattys.

[PaciFIST69]

any objection to splitting the patty's into smaller junks?


One note, the larger the WU (patty melts being larger so to speak) the less disk space the results take up. From Stanford's end, this equates to a heck of a lot of disk (spendy server disk) space. Larger in duration WU's save a ton of disk space on Stanford's end. Not saying good or bad here, just this could be a motivating factor.

[Sorin]

They can artificially monkey with the points of the bonus work units all they want, that's not going to change the fact that it still takes less than one day for a lot of work units and still takes 3 days or more for 212x projects. Points are only part of the equation, and not the biggest part of it either. In three days people would rather have cranked out 3-4+ other work units instead of just barely finishing one melt.

[ihrsetrdr]

@ HT
So you're saying that p2124 should be worth more than 392 points because it takes twice as long to fold ss p3038, even though they produce the same ppd and your machine didn't run one minute longer to produce 392 points on 1 p2124 than it would to make 392 points on two p3038s? You furnished no more resources and incurred no more expense. You should get no more points in my book.
@muddok
A days work deserves a days pay. If you do it in 1/2 a day you make a lot per hour. If you do it in a month, you make very little per hour. The points system works exactly like that and I don't see anything wrong with that. KISS principal at work. Early completion bonuses have been discussed before and dismissed as too difficult to implement serverside and the complaints about points awarded would be never ending.

RED=disagree

I did not make the case, as shown above in red; simply voiced a desire to see the patty melts modified in such a way as to make them more "palatable". I do not have a mastery of the point values and comparative stats associated with the varied spectrum of WU's, and doubt that I'll ever take the time to gain such mastery...I just fold 'em. ;) But, I do feel the pain that everyone else does over the patty melts, and if something could be done I'd view that as a plus. If not, I will continue to grin and bear it as I have been these past months.

YELLOW=agree

I've never been in management, just been an hourly guy, so that's my mind-set. ;)

[KingFish]

I just replied in that thread at the community and this is what I said:

Either drop all cpu work based bonuses or make and adjustment to the present system on regular wu's by assigning a bonus for quick completion and return of regular wu work. If you finish and return a regualr unit in 100-97% of the preferred deadline, give a 15% bonus to the wu. 96-92% of the preferrred deadline gets you a 10% bonus and 91-88% gets you a 5% bonus on the wu. This makes the quick return of any regular wu a real and noticable points-earner. and if they combine this with a drop in assigned bonuses given (like they have just done with the 149x/2060) it makes the points disparity much less between the patty melts and the bonus wu's.

It's not like Stanford hasn't ever made any adjustment in the points system. When I started participating in the project in 2001, a large wu was worth 2 points and most were worth under 1 point. So there is precedent for change.

EDIT: And this system would also reinforce the theme that all work units are important and not just the ones with BigWu bonuses attached.


Hawt damn! That is an excellent idea. I hadn't heard or thought of that before. That sure would fix a lot of problems. I hope the brass takes your suggestion seriously. I'm 100% behind you. Great idea!