Folding Forum

ChasR - exponential bonuses are unsustainable, but even without QRB the points system is exponential over time

MtM wrote:You do raise a question "The objective is to keep people folding for the long term, right?" I would hypothesise that normalising for computational improvement would keep folding people in the long time because there would not be a situation where absolute points increases mean that a person's 10 year points total is surpassed in a single day by a current folder!

Exactly! All that is happening is that people are turning F@H off. I don't think it is very hard to understand, do you?

vbironchef wrote:

MtM wrote:You do raise a question "The objective is to keep people folding for the long term, right?" I would hypothesise that normalising for computational improvement would keep folding people in the long time because there would not be a situation where absolute points increases mean that a person's 10 year points total is surpassed in a single day by a current folder!

Exactly! All that is happening is that people are turning F@H off. I don't think it is very hard to understand, do you?

In the 8 years my team has been folding it has had 368 folders but only 25 active folders now.

People come and go for all sort of reasons...it's the way of life and the way of folding.

orion wrote:

vbironchef wrote:

MtM wrote:You do raise a question "The objective is to keep people folding for the long term, right?" I would hypothesise that normalising for computational improvement would keep folding people in the long time because there would not be a situation where absolute points increases mean that a person's 10 year points total is surpassed in a single day by a current folder!

Exactly! All that is happening is that people are turning F@H off. I don't think it is very hard to understand, do you?

In the 8 years my team has been folding it has had 368 folders but only 25 active folders now.

People come and go for all sort of reasons...it's the way of life and the way of folding.

The fact that people come and go does not mean that we should not explore how to encourage people to stay for longer.

k1wi, I agree with you to an extent. I don't think the adoption of computing power on the desktop is occurring as fast as Moore's law would dictate so the need to compensate isn't as urgent. The QRB dramatically steepens the curve so that would be the first thing to fix.

ChasR wrote:@ MtM, I'm sorry I confused you.

Simplify my quote to "I think each WU has a scientific value and that value should remain constant."

Exponential bonuses are unsustainable. That's why I came up with an alternative. For the record, I'm opposed to my own alternative. Plug the formula into Excel, it's not that difficult. I did note the "It may not overcome scaling losses", which would mean running two instances of folding may produce more ppd in some circumstances. I should have been more clear.

You're not even close to confusing me.

You're actually proofing my first hunch I had when I first replied to you in this thread was correct.

You may feel exponential bonusses are unsubstainable, I do not agree and have explained why not once but more then a couple of times already, asking you why that doesn't fix your problem. Each time you say something which comes down to: you're not getting it, it's to confusing for you, blah blah blah. But you never answer, because you know that you can't go back to claiming your standpoint after doing so.

I repeated those two things not because they surprised me, but because they fall 100% in line with what I expected and made clear in my first reply to you, not because I got confused but to emphasize your 'unexpected' conclusions. You been saying the same things in every thread, and have avoided answering any of the relevant questions. Let me know when you're ready to answer those, the discussion might get interesting.

May I suggest this idea. Points could be multiplied by amount of years folding. For example,

1-5 years ... 1 times points
6-10 years... 2 times points
11-15 years... 3 times points
16-20 years... 4 times points
21-25 years... 5 times points

This way people have a reason to stick around. Just a thought.

I would rather not, because you're using participation as factor instead of sience, but if would keep more people happy, I wouldn't mind at all.

What you would run in to though, is that no new folder would pay for a system a ( cheap ba?) because you as 20y folder running some cheaper gpu's will still make more ppd.

Both should be encouraged, new folders, and sticking around. But preferably without putting of the other

ChasR wrote:k1wi, I agree with you to an extent. I don't think the adoption of computing power on the desktop is occurring as fast as Moore's law would dictate so the need to compensate isn't as urgent. The QRB dramatically steepens the curve so that would be the first thing to fix.

I Moore's law holds true, but because of it is people are making tradeoffs between computational improvement and issues such as mobility, power consumption and price. I would suspect that at the top end it is holding very true, because mobility and price don't really change and power consumption is pretty darn complex. Of course, this is leading to the 'spread' in relative computational power.

I think what we are both discussing is in what order we solve the issues. Both 'base PPD' (or ppd without QRB) and 'PPD with QRB' are exponential and together they are compounding each other over time. Fix one and the other becomes less of an issue. Fix both of them and the issue of exponentially increasing performance is solved.

Where we differ is that I am looking at fixing the long term curve first, and then looking at issues that you raise. That is, if we can resolve the long term effects of point inflation, we can then consider the effects of the proportional distribution of relative PPD at a a given (or present) time. Which is what I feel you are getting at. My reasoning for fixing the long term curve first is that once we do that, we are talking proportional differences at a point in time of say 160,000 points vs 10,000 points rather than (eventually) 1,600,000,000 vs. 100,000,000.

The reason why I want to solve the long term curve first is in part because I think the second part is much harder. Basically, to adjust the proportional differences, or the curve at a point in time, we need to look at the relationship between all the different platforms (using bruce's definition of the term) and then look at all the different types of hardware (desktop vs. server vs., for want of a better term, mobile, vs. console) and study them with regard to their proportional scientific value. From that we need to consider what we want the relative proportion to be. Should it be 16x (as in the example above) or should it be more like 4x?

I apologise if that got long fast!

Faster rigs would always get more points per day. Based on time served folding if a person is running a slow rig then his points would not increase much, just enough to keep that person around. On the other hand a person may upgrade his or her rig if they know that they will get 2X the points. I think it's a win/win.

@K1w1

The second part is easier when you do the base ppd first, would be so even only because bonus is relative to base. Maybe people will drop their unfounded objections against an exponential QRB when the bonus would be smaller in absolute term?

Btw, do I need to quote your above post and then go back and quote where you already said the same thing? Not attacking, trying to safe you time. I don't think explaining yourself was needed at all. As to ChasR's objections to QRB, it's something which was discussed extensively in those other threads, and probably a part of why this thread ( strenghtend with your confusing the issue by talking about qrb and base ppd without distinquising them ) did not meet a lot of positive feedback from me at the start as I was expecting the same content as those other threads got filled with.

It's time to make new suggestions, as that's what's going to make you the one who came up with a workable solution.

vbironchef wrote:Faster rigs would always get more points per day. Based on time served folding if a person is running a slow rig then his points would not increase much, just enough to keep that person around. On the other hand a person may upgrade his or her rig if they know that they will get 2X the points. I think it's a win/win.

Proof you statement about faster rigs getting more points always:

I have a 8600m gt making 1100ppd, let's say I'm a 20y folder which I'm not but for the sake of argument. That's 5x1100 = 5500 ppd. Are you saying there are no gpu's faster then a 8600m gt which would make less then 5500 ppd?

Yes, I know it is complex.

On one hand you want to reward loyalty, continuous participation.
On the other hand you want to reward the contribution of the greatest amount of science donated in the least amount of time.

Balancing these conflicting goals is the core of the discussion.

Jumping to a discussion of "Normalizing" presupposes a conclusion that has not been really discussed:
What is the best way to represent the exponential increase in the actual knowledge base.?

"normalizing" presumes that contribution in either time, money, or electric power should be the basis of giving out points.

However, the point system to date has not subscribed to that notion.
The point system to date has subscribed to the idea that "points" should have a correlation to the "chunk of data" that has been massaged.(now with a correlation to the time in which it was promptly returned)

Jumping immediately to a technical discussion of the best "means" to achieve point "Normalization" really by-passes the philosophical, physiological, and sociological questions that really are really at the root of the problem.


I have stated before that I really find the entire discussion antithetical to progress.
The entire objective is to continuously improve, by multiple factors, the ability to describe the science.

If my computer can be 10x, 100x, or 10,000x faster....that is something I want to know, and chart and get excited about.
But that is just me.

Here is an example.

http://folding.typepad.com/news/2012/02 ... n-fah.html

According to this post, the existing GPU cards might see a 10x improvement in computational output.
If acting in direct correlation with a CPU client, that improvement might be further leveraged.

Now, How to reward those who want to participate?
You could say the points should stay the same because the hardware didn't change. You could say the investment in equipment and power is still the same. You could say that the multiplication of output was not due to any donor effort, but rather due to improved numerical and simulation methodology.

I would say progress is progress
If a certain method results in 10X or 50x improvement in computation. The donor should be seeing 10x or 50x improvement in reward.

It may not be pretty, and it might not square with everyone's ideal of fairness, but that is how you motivate people and it is how progress is achieved on a disruptive rather than glacial scale.

mdk - under the propsal discussed here they would still get a 10x or 50x improvement in reward, it would just be proportional to other users, not exponentially high.

I guess, and this is where I am finding it hard to explain it to people, I want to reward users over time for proportional effort, not reward them over time for the improvements that Intel or GROMACs make!

MtM wrote:Proof you statement about faster rigs getting more points always:

I have a 8600m gt making 1100ppd, let's say I'm a 20y folder which I'm not but for the sake of argument. That's 5x1100 = 5500 ppd. Are you saying there are no gpu's faster then a 8600m gt which would make less then 5500 ppd?

Work with me here. A 8600m gt making 1100ppd you know as well as I do that the older the graphic card is the slower or less points you get. That goes for SMP units as well. The faster the CPU or GPU will produce more PPD than a older rig. Having a multiplier would not hurt the point system much if at all. Remember a person would not get 2X the points til only after 5 years of folding. Who has a Graphic Card that is 5 or 6 years old? Ok, I do, but I don't use it. It's in a drawer somewhere. So if someone is folding for 6 years they should be given a bone for just sticking around. Only if they upgrade their rig would be where you would see a decent increase in points. Point is that it gives people that have been folding for a long time a incentive to upgrade and in turn increase productivity of science. How's that!

Sorry, not exponentially high, but exponentially high in absolute terms.