Folding Forum

Well,

Awachs is the number one in a little over six months.

Currently up to 10,000,000 ppd.

Is his contribution not significant?

What is the objective of this points envy? Perhaps he should quit to make sure long term folders are not intimidated?

Yes, lets all aim low so no one feels bad.

Makes my head hurt.

Well, he has 444 active clients (past 7 days), so I tend to put him in to a conceptual 'enterprise category', as I would OATmeal's 484, or as I did Pande Lab when they running a considerable load.

I don't think anyone would argue their points contribution is not significant and I don't think anyone (other than you) is suggesting that they should 'quit'..? I'm not even sure why you would suggest that that is the implicit opinion amongst people that believe that the point system in it's current design is unsustainable?

I'm not even suggesting their proportion of points relative to other users should change. What I say is that, in my opinion, for which I have outlined the justifications, exponential absolute point systems have a net downside relative to a normalised/standardised/differential points system.

Obviously, as you have stated a number of times previously, your opinion is that exponential absolute point systems have a net upside, which is that users should be rewarded for the technological improvement of their hardware. Yet you still come out implying underlying socialist or communist ideologies and suggestions that people in favour of making the points system sustainable want big folders to quit!

smp and bigadv both have a qrb system for quite some time now. what about us gpu folders? still waiting.

jimerickson wrote:smp and bigadv both have a qrb system for quite some time now. what about us gpu folders? still waiting.

In order to do that they would have to greatly reduce the base points to something more realistic of their value.
They are Single precision wu... and are valued at about 60x ppd/atom than a bigadv.

They are good for deciding what to focus research on but bigadv does the work.

There is nothing unsustainable about tears 2 million ppd or AWACHS 10 million ppd.

What then was the point of 7ims post?

Apparently he was pointing out the high points in relation to WU.

Now, I know 7im knows better. Those WU, were very large, and took a very long time to complete.
I finish GPU WU in 4 hours. Large WU on 48 core 4p machines can take days.

7im knows that WU count has no correlation to actual work done, since the WU are divided to match the strength of the machine doing them.

Consequently, his post was really a red herring to this conversation, unless of course he was just trying to provoke discussion.

How does pointing out the ability of someone to make large donation, somehow justify changing the point system? I don't see it. The only logic is that it is somehow unfair.
The only other time I hear this argument is in politics and economics.

I've never heard someone say that a cruise ship shouldn't have 4800 HP engines...that personal bass boats will somehow feel cheated....that since it is not fair, we should rename the power to cruise ship HP where 1000 normal HP = one cruise ship HP.

There, now both boats will have 48 to 200 hp. (the cruise ship has (4) 4800 hp engines and bass boats are know to have( 2) 100 hp engines.

See we have renamed the hp, everyone is happy, the production is normalized, since it would not be fair to give the cruise ship an advantage just because it is bigger and has more money and has newer technology.
Heck, the cruise ship only goes 22 knots top speed, and everyone knows that a good bass boat can go faster than that.

Sheer madness.

Patriot wrote:In order to do that they would have to greatly reduce the base points to something more realistic of their value.
They are Single precision wu... and are valued at about 60x ppd/atom than a bigadv.

They are good for deciding what to focus research on but bigadv does the work.

i am all for reducing base points so long as we get a curve like bigadv.

jimerickson wrote:i am all for reducing base points so long as we get a curve like bigadv.

Careful what you wish for. There's crap loads of duplication on GPU WUs because of all the OCed graphics cards. People would find their QRB non existent once they fall below the 80% successful completion rule. Then they wouldn't be getting any bonus on their SMP clients either. Could be the end of GPU folding.

hmm maybe so. i don't overclock so i didn't think of that.

I have seen this points argument over and over. If they were to drop the points system and just count WU's would you quit folding?
When I srtarted I didn't even know there were points. All the points really do is give us something to compare the start point, wihich is 0, to.
If someone has equpment that can do more, get it done faster, or do a better job of it, they will make more money on a given job. Why should this be any different?

Blacksmith1 wrote: If someone has equpment that can do more, get it done faster, or do a better job of it, they will make more money on a given job. Why should this be any different?

Simply because some fill/think/believe it's not fair for some with the right kind of hardware to get more points than those that don't.

Since tear's name has been brought up in this thread I would hope that certain people would be very happy that he is donating his time, systems and electricity to F@H. Not on some other distributive computing endeavor.

orion wrote:Since tear's name has been brought up in this thread I would hope that certain people would be very happy that he is donating his time, systems and electricity to F@H.

I know I am.

Blacksmith1 wrote:I have seen this points argument over and over. If they were to drop the points system and just count WU's would you quit folding?
When I srtarted I didn't even know there were points. All the points really do is give us something to compare the start point, wihich is 0, to.
If someone has equpment that can do more, get it done faster, or do a better job of it, they will make more money on a given job. Why should this be any different?

That doesn't change under the normalisation proposal. If anyone is arguing that powerful folders are going to be unfairly affected by this proposal and have their contribution negatively affected I'd love to see the working behind it...

The proposal in this thread has nothing to do with measuring based on WU counts, but it is apparent that any discussion of a different system brings in discussion of all sorts of other ways of generating points. Which is human nature I guess. All normalisation does is remove the expectation that points are easier to get simply because the good people at Intel, Nvidia, AMD make exponential gains in computing performance

I just want to quote from one of Pande Group's articles: Folding@home: lessons from eight years of distributed computing.

"The statistics system also allows volunteers to compete against each other individually and in teams with the credits assigned for completing work units, feeding the primal need to compete against others. This aspect drives recruitment of new volunteers by spreading word of mouth and other forms of viral marketing. Volunteers are often fiercely loyal to projects like Folding@home once they get involved, and many have team websites, recruting efforts, and even merchandise of their own. In an effort to boost their statistics they often upgrade their hardware more aggressively than those not involved in volunteer computing. This serves as an additional positive feedback mechanism for the self-upgrading aspect of the system."

The key concept I take from this is that the statistics system, however it is designed, should encourage folding. It is clear from that paper that the points system has an important marginal influence on participation. If, through keeping points in a relatively 'narrow' but still proportional band in a regular systemic manner (so that faster computers still earn more points than slower computers at any given point in time and in the exact same proportion, but through the system the psychological 'barrier' of massive differences in absolute values is avoided), a normalisation system encourages more people to fold, then I am 100% for it. F@H is about progressing the science behind protein folding, but it is also about progressing our understanding of Distributed Computing. I'd say that studying the effects of the points system falls under that category and to simply say "the points system has always been like this, therefore it should always be like this" is failing to scrutinize a very big area of potential progress.

I do feel that debate as to whether it is better or worse is for another day, because until this proposal is developed into a workable, final product with a formula that can be applied to the points system (which to my knowledge has yet to be done despite all the discussion around 'fairness' and the points system), much of the debate is simply attempting to drown out development/run interference.

In this thread I wanted to develop the methodology for achieving such a system - to expect an individual user to develop an optimal & complete formula is too big an ask and in my mind, ignores one of the fundamental advantages of DC - collaboration, hence my decision to take this route. I have attempted to explain the concepts behind my proposal to those who have questioned it, because I appreciate that hearing their opinions are valuable in deciding how to explain the system's strengths when the time comes, but I would much prefer people contributing to this thread focus on improving the methodology of my proposal, rather 'advocating' their own philosphical position over and over!

No one's name was brought up in any specific way. I picked 2 random fah accounts close to each other in ranking, but with largely different WU counts. I could have just as easily picked these two below and shown the same disparity. And no one questioned their contributions in any way, or suggested a reason to quit fah... if you thought so, that is your own unreasonable attribution, not mine.

17 AtlasFolder 403837339 797480
19 sfield 367580115 7038

Yes, it was posted to get people thinking about PPD. BigAdv WUs are not that big, and GPU WUs are not that small. BigAdv are not 1000 times bigger, and do not take 1000 times longer to fold, nor do they take a machine that is 1000 times more expensive to build. Anyone who can look at 800 work units vs. 800,000 without pause to consider there might be some kind of imbalance, or not want some kind of full explanation, is surely the mad one.

That is a very narrow minded thought. What was the total cost to the donor for the equipment etc. to run those WU's what was the investment in time what is the initial investment. And how are you determining the science value of each WU. Only Stanford knows the Value of a given WU you do not neither do I. Stanford says WU X is worth X amount so that is what it is worth. You keep saying you know the value is wrong how do you know this, is it secretly your project. If it is then fix it and let’s see how much more science gets done.

Yes, it was posted to get people thinking about PPD. BigAdv WUs are not that big, and GPU WUs are not that small. BigAdv are not 1000 times bigger, and do not take 1000 times longer to fold, nor do they take a machine that is 1000 times more expensive to build. Anyone who can look at 800 work units vs. 800,000 without pause to consider there might be some kind of imbalance, or not want some kind of full explanation, is surely the mad one.

Well, since you know the full explanation;
Then you should start a thread advocating the elimination of the QRB.

This seems to be your underlying concern.

However, when I look at the list of recently announced papers and developments, I have to ask; How many of theses advances would have been made in the last few years without the QRB?