Bigadv points change

[snowbird48]

This is a great debate going on here, however, with much remiss, I respectfully disagree with much of it. I would liken it to a bubbling cauldera with a thin skin on top. I often see rants on here regarding donation of time and computer resources. Conversely, no one is forcing anybody to do this. The real value of anybody's contribution should be a feeling that comes from within rather than looking at points spreadsheets over at FAHmon or EOC. The idea of DC is holistic and that should be remembered. I believe the largest pool of points posted is from unnamed, unnumbered contributors. Tells me the majority of donors are not that concerned about QRB. After trying different combinations of clients I tried core A5 and continue to run it, not for the big points. Surprisingly, it puts the least amount of stress on my gear and that is what matters to me. My OC'd 920 is a smurf compared to those four G34 chipped brutes. I have no problem with the points they get. The objective here is to get science done and they do that very well. I admire a person who will put up the money to build a specialized machine like this and then pay the power bill to run it. How will they react when someone drops the bomb and shows up here with four or more of those 10 core Intels and a gaggle of Tesla cards to fill in the doubtless short intervals between completion of core A5 wu's. I think Stanford should should be benchmarking with very current gear and extend the K factor to more of the Gromacs unicore wu's. Those who are motivated by points only should venture out onto the web and register on a MOBA. You get the added dimension of honing your operator to keyboard skills and there a legions of like minded people who will satisfy your need to compete for points.

There seems to be a lot of donors on here who take issue with 7im's responses. Perhaps they should read his footer when he posts. The moderators on here have to go through a lot of repetitive querys. I think they handle it very well and keep the atmosphere of this forum as a place of science well maintained. I will now risk 7im's tritness as I have a question for him. If you have disdain for both AMD and Intel what are you running, an abacus?

[MtM]

This is a confusing situation.

a. PG decides the scientific value of a work unit
b. a point system should reflect scientific value

This is what we always had and need to have right? So with QRB PG needs to confirm the slope adhears to scientific value.

TBH, yes I can't run a bigadv system, and yes I noticed those who can are the majority of the people passing me on the statistics side of things, left and right even. But why would it make me leave if PG has decided the quick return is really that important to the science? People who are complaining now are looking at it from the wrong perspective.

If an 48 core machine can do work which is as important as an entire year folding from a couple years back, fine. That's called progress, and we should embrace it.

Even more, those who folded years ago and stopped at some time should feel an incentive to join up again and double their points as quickly as possible, instead of having to fold for years to match their old output and double it. The big points can be motivational at least as much as it can the opposite for those few who are not thinking this through without emotional attachments. Because imo, anyone who is against the hard connection between science and points is playing a game with the minds of all donors, a game aimed to fool people to participate by inflating their personal contribution.

Now, I'm not saying the current point system is working, as I am not sure the exponential increase in points is an exact match with scientific value, if it's not the formula needs adapting to the real situation. If it is, no change should be done at all. If people leave because they feel unappreciated let them leave because trying to keep them folding by adjusting the point system to artificially increase their contributions, doing so automatically decreases other people's contributions.

How about this: I am not leaving even if I drop to the last % of donor's contributing, my contribution is my presence and my machine(s) no matter how old, I'm here because I lost many family members to cancer's, because there are many people with diseases which involve the folding of proteins and because I think this project is on the foreground in research that can help millions of people eventually, the points are second nature even if important as a measurement. That gives a sense of pride, which would be taken away if PG bows down to those complaining they aren't getting enough points for their contribution.

It all comes down to this: instead of asking us for workable figures, tell us the true scientific bonus of returning a unit quicker. It's not the job of the donors to decide what their contributions are worth, it's your's ( directed at Vijah Pande ). Do no move away from scientific value should equal points, that would be wrong. If the current method isn't working right because of the variance in donor machines then adjust the benchmark machine to be more representative of what donors are/should be using, but keep true to the principle that points=value from your end. And I never included 'donor perception' as being something which has to be thrown in the equation which leads to attributing scientific value to a work unit, even when from a management standpoint that might be right ( if you really loose allot of donors science will be hurt more by not listening to their 'demands', thus one could see donor happynes as a factor in assigning a scientific value to work units.. I'm just not sure I like that concept as it leaves PG open to 'blackmail' ).

[Punchy]

Isn't the "flat rate" award already there - the number of completed work units?

[MtM]

Isn't the "flat rate" award already there - the number of completed work units?

Well, no not really if you take into account the run time of gpu wu's versus smp wu's.

When smp was new, you could see people with avg points but huge wu count's, and you would know those people had been in the fold for a long time from just their wu count. But that changed in time when work unit's got really short compared to the old day's. When you now look at just the stats there is no clear way to tell.

Mind you, I'm reasoning from the opposite perspective right now, I don't have a problem with points being awarded based on science, no matter how inflated current contributions might seem compared with the contributions made years back. But wu count being equal to a flat rate would only hold true if all work unit's were equal in run time, classic/smp and gpu.

[mdk777]

And I never included 'donor perception' as being something which has to be thrown in the equation which leads to attributing scientific value to a work unit, even when from a management standpoint that might be right ( if you really loose allot of donors science will be hurt more by not listening to their 'demands', thus one could see donor happynes as a factor in assigning a scientific value to work units.. I'm just not sure I like that concept as it leaves PG open to 'blackmail' ).

I understand your point. One thing to consider is that in the last few years, 90% of the work has been done by the top 10% of the donors.
This is just a fact. While it would be nice to think that a donation of an old p4 is making a contribution, in reality these older machines are less than a rounding error in the statistics.
While the "politics" of keeping these donors( the top 10%) happy should not over-ride the interest of the research; It cannot be excluded from realistic consideration either.
In the end, we all know that "points" are of course totally arbitrary. They only stand for whatever we choose to have them represent.

From the lack of consistent Big adv WU availability, it seems the current point system is actually working too well. The supply of folders is actually out-pacing the supply of WU. It is hard to make the case that the system needs to be scrapped, when "points" cost nothing, and the objective result is beyond the expectation.

Can it be improved? Sure. But talk of "capping" is a disincentive, a throw-back to class envy and socialism.


No one wants to spend time and money ($2000-$20,000) to hear that their investment and design is "too good" ; that like a NASCAR race, they should have to install artificial intake plates to limit their performance.

As VJ said, make a recommendation on a working formula, or stop the discussion. FUD will only reduce, not increase donor participation.

[MtM]

There been formula's posted, just not complemented with examples of how it would pan out.

( K * Deadline / WUTime ) ^ (1/3) * Base

*Image from 7im


Falling back to the 3rd root will make that slope flat out no? I don't have the spreadsheet here now which calculates ppd or I would give some examples with the 3rd root formula just to look how the spread would be. Looking if I can find the file will edit later ( pm me if you have it and want to save some time ).

[7im]

And I never included 'donor perception' as being something which has to be thrown in the equation which leads to attributing scientific value to a work unit, even when from a management standpoint that might be right ( if you really loose allot of donors science will be hurt more by not listening to their 'demands', thus one could see donor happynes as a factor in assigning a scientific value to work units.. I'm just not sure I like that concept as it leaves PG open to 'blackmail' ).

I understand your point. One thing to consider is that in the last few years, 90% of the work has been done by the top 10% of the donors.
This is just a fact. While it would be nice to think that a donation of an old p4 is making a contribution, in reality these older machines are less than a rounding error in the statistics.
While the "politics" of keeping these donors( the top 10%) happy should not over-ride the interest of the research; It cannot be excluded from realistic consideration either.
In the end, we all know that "points" are of course totally arbitrary. They only stand for whatever we choose to have them represent.

From the lack of consistent Big adv WU availability, it seems the current point system is actually working too well. The supply of folders is actually out-pacing the supply of WU. It is hard to make the case that the system needs to be scrapped, when "points" cost nothing, and the objective result is beyond the expectation.

Can it be improved? Sure. But talk of "capping" is a disincentive, a throw-back to class envy and socialism.

No one wants to spend time and money ($2000-$20,000) to hear that their investment and design is "too good" ; that like a NASCAR race, they should have to install artificial intake plates to limit their performance.

Again with the incorrect usage of the term arbitrary. Points are tied to scientific value, they are NOT random as you would imply.

If points cost nothing, why do they have such a high value? And no offense, but maintaining a fair and workable points system does cost a lot of time and money. How much times does Vijay spend on topics like this? And work units have to be benchmarked, on benchmark computers. Points cost nothing? Hardly. You're simply wrong about that. Points have intrinsic cost and value, however small that may be.

And if the points system works too well, why all the complaints?

The top 10% got 90% of the points because the points curve is too steap, not because they did 90% of the work.

Capping is socialism? This isn't a political forum. Try to keep things real here. The choice of the points curve is either infinite, or not infinite. Infinite is not realistic. So there needs to be a limit, or a cap. I didn't suggest stealing points from the rich 48 core people to give points to the 1 core P4 users. I did not suggest a flat limit either.

To be realistic, the line of the curve IS the cap. All I am suggesting is that we make the curve proportional to the amount of hardware being donated. If you're a big folder and want to donate a $20,000 to folding, please do! You should get $20K worth of points. But with the current points system, you're getting $100K worth of points.

Reducing those points is not socialism, but it is fairness, to those who have contributed before, and to those who will follow.

[MtM]

7im I think the real answer can't be found this way, it's first up to PG to qualify the added scientific value of quicker returns is it not? If the curve represents scientific value, there is no discussion left other then a discussion to move away from the solid connection between scientific value to PG, and points awarded to donors. Like I said above, that line should be maintained.

I'm not sure a cap fits with the direct tie described endlessly before, that's why I don't like the idea. A core is something which has no relevance to performance, if you want to benchmark/cap your way then your other post made more sense. Benchmark the machine's capabilities ( mflop/tflop ) and base the cap on that, not on something as 'cores' as if you do that, you need to change the whole formula each time a new architecture is released by any hardware vendor. And even more, if you base it on an actual figure which indicates performance, it's not really a solid cap, and my objection against it is removed completely.

[7im]

Core, FPU, MFLOP, whatever. I used "cores" as a commonly understood demonstrative example, like using a simpler car analogy to explain a more complex technical concept. I don't expect the new curve to be designed for people with 4 "tires." My use of "cores" is an example, not a detailed solution.

The curve should be based on reality, not a points curve that was designed to make A2 and A3 work units score in a similar manner on 1, 2 and 4 core systems! New Points FAQ - Details

The right side of that curve for 16, 32, and 48 core systems was NEVER accounted for. This is a shortsighted FLAW in the original design and needs to be correct ASAP!

[mdk777]

The top 10% got 90% of the points because the points curve is too steap, not because they did 90% of the work.

No, I was looking at total FLOPS, not points. The real value is probably greater. I have just looked at rough estimates from the stats page(# of GPU folders + # of #PS3 folders(and their FLOPS) compared to total folders, as compared to total FLOPS). When you factor the top producers in the PC and LINUX in with the GPU and PS3, the number is probably 95% of the work is done by the top 10% of the donors. The points generated has nothing to do with actual work.


Part of fair is predictable. Knowable, in order to make accurate investment decisions.

As I am sure I have posted in past discussion, donors need to know the rules.
Setting the rules, and then changing them after people make significant investments is bad form.

As VJ said, you could just eliminate the BIG ADV or QRT bonus if everyone thought it was intrinsically unfair.
However, my point was that you run the risk of then alienating your most productive group of donors.


Finally, you are correct that we seem to have a different understanding of some word's connotation.
However, arbitrary has never meant random. Someone has to be the arbitrator, the person who makes the decision.

You are thinking of capricious.

[MtM]

That is PG territory, not mine and probably also not yours no offense. The line should represent scientific value, if PG feels that returning a unit in 2 days equals x points, 1 day x times 4 but half a day x times a hundred, I can not comment on it other then saying that while it might feel unfair, if it's true then we should not complain.

I don't mind being a footnote in the statistics compared to others, I'm just glad those others are willing to contribute what they are.

In Dutch there is a saying which says a person who makes shoes for a living should stick to his profession, the contained message is that people should be happy with who/what they are, it's as simple as that imo.

Again, if there is no added value of that magnitude to PG getting them back x amount quicker, then your point is noted, appreciated and 200% supported. I don't mind those people all running away from me on the stats, but only if their stats indicate their contribution to the 'cause'

The more I think about it, I am looking forward to your idea you talked about, no matter how this works out now I would really like to see what you have in mind.

Also, mdk777 is right imo about K being 'adjustable' which is imo the same as 'arbitrary'.

[mdk777]

This is a shortsighted FLAW in the original design and needs to be correct ASAP!

Here we agree about process.

If you are going to make changes, they should be made sooner, rather than later; not after people invest in more 2p and 4p systems.

[VijayPande]

Sorry to belabor the point, but I still don't see any specific suggestions. I know that specifics are hard, but that's the point (in a sense, "talk is cheap"). How about we avoid debating in some non-specific sense and instead debate specific, detailed suggestions? (not just the equation, but the values that go into it as well).

[MtM]

Did you read my comments directed at you Dr. Pande?

The most important one is this

Now, I'm not saying the current point system is working, as I am not sure the exponential increase in points is an exact match with scientific value, if it's not the formula needs adapting to the real situation. If it is, no change should be done at all. If people leave because they feel unappreciated let them leave because trying to keep them folding by adjusting the point system to artificially increase their contributions, doing so automatically decreases other people's contributions.

Does a unit released on day x and turned in on day y have an z value, on day y-1 a value of z * 2, but on day y-2 a value of z * 4? Yes, excuse my lacking education, math isn't my strong point, but the question I hope is still clear.

The question is about the right side of the slope, where does it if at all, stop to be proportional to scientific value.

I don't think you can expect donors to come up with a point system and later on claim it's based on scientific value. That would be paradoxical.

[VijayPande]

I don't think you can expect donors to come up with a point system and later on claim it's based on scientific value. That would be paradoxical.

My hope there was to see the candidates and choose the one which is closest to what's useful scientifically (or to suggest modifications to it that make it more useful scientifically). For example, a QRM is useful scientifically.

However, I don't want to push our current system if donors really hate it. While optimizing for science is my #1 concern, my #2 concern is to make sure that we have a lot of donors around to do the science, hence finding some compromise.