Folding Forum

orion wrote:
So what's next? People folding on HPCS.

F@H #1 folder 343 clients running on the cloud. http://folding.extremeoverclocking.com/ ... p?s=&srt=1 http://fah-web.stanford.edu/cgi-bin/mai ... me=_s3v3n_ but who cares about that it is not de-valuating anybody efforts right. And they are only breaking a few moral and ethical rules while the rest of the folders out there have to compete with them.There are several people on that team alone that are running multiple instances of the cloud how many on the rest of the teams out there are doing the same. But who cares about that lets just look at how the bigadv folders are de-valuating everybody. (Sticks head in sand)

orion wrote:

7im wrote:Orion asked how many times is this going to come up?! As many times as it takes to solve the wheat and chess problem with the current QRB system. It's not going away unless you can convince Intel to stop double the core counts every 18 months. And running away or ignoring it won't solve it.

I didn't realize that this was intel's doing. All along I thought that it was PG. So here's a novel idea...why don't the complainers take it to an intel forum and lobby intel to stop doubling core counts every 18 months

Where were some of you when BA’s and their QRB were implemented back in '09? I don’t recall one iota about this back then but now it’s a problem. All I heard back then was how slow 2p and x6’s where slowing the science down and that the shortages of BA's were because of them. Nothing about points being too great for fast rigs that did BA's in a short amount of time.

So what's next? People folding on HPCS.

Tbh this was discussed before as well. And there is a point made which seems pretty valid, even while it's not relevant right now it will be in the near future so having it as a point of discussion isn't a bad idea.

As computational power keeps increasing, we'll start seeing people with 10.0000 ppd or more, and this can lead to point tally's which need to be shown as xxxx to the y in the stats unless you prevent it by normalizing them against the actual increase in computational power.

That part does make sense.

Well, I have to disagree with my esteemed teammate.

It has indeed been debated Ad nauseam.

However, so has world population growth, deficit spending and nuclear deterrence and non proliferation.
To date, I have not heard any resolution to any of these problems that has been the result of an "impending urgency argument".

What it comes down to is that PG at some point in time will need to "blow-up" the system and start over.
7im is correct that the "problem" has not changed, but further debate won't change the parameters.

PG created the system, and only PG can make the necessary changes. I haven't heard them asking for input on the topic lately.
Until they show some inclination to tackle this unpopular conundrum, having these debates only leads to consternation.

I feel your pain Grandpa. What is the point in attacking people who follow the rules and make significant contributions.

OK. You've all expressed your opinions several times ... both in this topic and in previous topics.

I don't know about you, but I get tired of perpetual discussions that accomplish nothing.

Topic closed.

I have been asked to reopen the topic. I must admit: perhaps I acted too rashly.

I can't really say "Please stay on topic" because there are already two topics being discussed in this thread ... but please don't wander too far-afield.

Just reading through this and old saying come to mind, if you cant run with the big dogs stay on the porch. Thats where I am on the porch,when it is no longer worth it to run what I have I will turn them off move them to the shed for builds for family, neighbors, video surveillance, sell them or what ever.

Not putting any more money in it myself, been dumping tons since 03 and have not seen anything that has impressed me as to cures etc. Lost my uncle back then to cancer got on the band wagon to help find some cures, now mom is gone feb 25 i see no reason to throw more money in this myself... Cant afford power to run what I got and its only fixing to get worse with $6 a gallon gas coming this summer... Let the big dogs fold on with the clouds etc. I hope they find something...

OK,
Say you were setting up an endowment: Like the Getty trust fund, but much smaller.
Instead of investing in ART, you wrote the endowment to “ensure the steady and perpetual donation to FOLDING at HOME”.

Now pick an amount, say $100,000. And for the sake of agreement, say that you get a steady 5% rate of interest.
So, how would you invest in equipment and power, and how would the points you generate in the ninth year compare with the points you generated in the first 8 years?
How about in the 20th year compared to the previous 19?
If you spent $2500 per year on electricity, and every four years spent the remaining $10,000 of generated interest on the latest equipment…wouldn’t the last year’s production after an upgrade always make your previous year’s donation look bad?
This is why I don’t see the chess board problem as a real concern.
In 20 years, if people are generating in trillions or billion trillions it doesn’t matter. We have the notation and since they are just abstractions, they don’t take up any space.

I also understand the donor fatigue expressed by road-runner. While I don’t think there is anything disingenuous about the description of the project, it is easy to hope that applications will result sooner.

First of all, I'd like to thank bruce for reopening this thread - I can understand his reasons for closing it and I hope that we can continue to develop this process without having it locked again - because it is really important.

I have been asked to prove why there is a need for adjustment, and what this adjustment means. Therefore, I have created a graph based on real data.

Graph 1

What the graph above is showing is the inflation in PPD that is directly attributed to technological improvement over time. That is, in the above graph, the X axis represents time, not TPF. Because of the increase in absolute numbers I have had to log the axis. That is, due to technological improvement we are moving towards the left hand side of the graph and ppd is increasing exponentially.

There are four lines on the graph. The first, the blue line, shows the current inflation in PPD over time that result from technological improvement. As you can see, it is increasing exponentially. The red line shows us the current inflation technological improvement has on the present QRB system. The key thing to take away from this is that the red line is growing faster than the blue line. Why is this? Because under the QRB system, the impact of technological improvement is being applied twice - once in increasing the base points and then again in increasing the speed ratio.

The green line shows the effect of accounting/controlling/normalising for technological improvement. Here we can see that we have removed the effect of technological improvement from the calculation of PPD over time. However, what is interesting is that while the curve of the QRB points system has been greatly reduced by accounting for the technological improvement on the underlying points system, there is still a curve. This suggests that the current proposed system does not completely remove the inflation caused by technological improvement. What it does do is significantly reduce it and perhaps buys us some time before we have to reconsider it.

Now, it might be easy to say "well this does not work, because under this system there is no incentive to fold faster." That hypothesis, while very understandable, is not accurate. Why? Because at any one point in time, the original point curve still applies. That is, at a given point in time the point curve below still applies. Someone folding at twice the speed of another folder will still have an increased incentive greater than 2x to fold, but the value of the incentive will not increase exponentially over time.

Graph 2

My proposal of PPDn / Yn, where PPDn is the points acquired in the time period n and Y is the relative computational improvement at time period n compared with the base at time 0, fulfils the original intent of my first post:

As I see it, the easiest way to keep 'PPD inflation' from going nuclear, is to adjust the benchmark machine to the rate of CPU Improvement. At a defined interval (annually is probably too infrequent, but monthly may be too frequently), adjust the benchmark machine's PPD to reflect the improvements of computing power.

MtM suggested his proposal:

We make ba make instead off 1000 points, 100 points ( 10%), and smp instead of 100 points, 10 (10%).
We use 10% as a start as it's easy to use.

The next round, we use Y which I described should be based on the computation speed increase. This speed increase should be applied to both trajectories again, so let's assume we have a 10% speedup which have caused ba to be making 1000 points again, and smp 100 points = we normalize down by 10% for both and publish the 10% number so people can still see the how much computational/scientific effort was needed to make earn x credit. How does this influence the relative progression of a ba versus a regular smp instance? The BA instance will still earn the same amount of points more than the SMP instance relative to the total points obtainable.

First of all, I disagree with this proposal for a number of reasons, first that it is ambiguous:

1. Why are we differentiating between BA and SMP? It would be simpler to simply write PPD1 = PPD2/y1, which is exactly the concept I proposed in the original post and exactly the formula I proposed in the post prior to your suggestion.

2. By talking about BA and SMP separately, you are increasing the complexity of the adjustment - It is quite easy for people to read that proposal and suggest perhaps you are normalising BA to 1000 points and SMP to 100.

3. The 1/10th figure is completely arbitrary and appears 'simply thrown out there'. Amongst other things, massive. Far too large for a single adjustment. Why choose 10%? Why not make the value of new points half the value of original points? I have had to justify every single element of my posts including exactly how to calculate the value of technological improvement (and have the examples of how it can be calculated then used as a vector to shoot down my entire proposal) and I think so should you.

4. If what you are attempting to propose has the same effect as my original proposal, why not use the formula that I proposed in the prior post? As per point 1, my formula is much more simple.

I have been constantly asked to provide conceptual evidence and proof of my concept, while others have not done so. Therefore, below is a hopefully concise outline of my proposal:

My equation was initially written, PPD / Y

It would have been clearer, but analogous, if it had been written PPDn /Yn

where PPDn is the points acquired in the time period n and Y is the relative computational improvement at time period n compared with the base at time 0.

The formula to calculate total accrued points becomes:

sum(PPD1/Y1)+(PPD2/Y2)+ (PPD3/Y3)...(PPDn/Yn)

If you have earned 10,000 points, then those 10,000 points remain locked in and are not changed. What does change is the value of the current points that folders are earning but only in relation to other time periods and only in relation to technological improvement.

Therefore, we take User A. At the start of 2012 they have a total of 10,000 points and they presently earn points at a rate of 100 per day on a new computer of 'relatively decent performance at this point in time'. This 100ppd happens to be about average across the entire community.

PG uses a range of tools and statistics to measure the technological improvement in computers over a 6 month period and determines that this value is 4% - because of the improvement in computers points are 4% easier to earn. Therefore, it devalues the current value of additional points by 4%. The reason why technological improvement can be treated as a continuous variable is because people do not all go out and purchase a new computer when a new generation is released and the price of the new generation hardware does drop during while that generation of hardware is still the most up to date generation.

In any event, User A still has their accrued 10,000 previously earned points, but their present ppd drops from 100 points to 96 (because they haven't acquired additional computers). This isn't unfair though, because the 4% drop affects ALL users equally, so no user is worse off. Under the old system, User A would still be earning 100ppd, but on average (as more users adopt newer generation hardware), the average has increased by 4% to 104. It isn't altering the 'scientific value' of points any more than the existing system does in hyperinflation.

We extend the time frame out, to say 3 years, User A now has 18,000 points, (I haven't calculated the exact amount because it would require a lot of calculations and assumptions). In this scenario, because computers have increased in power at an average rate of 100% every 18 months. over 3 years this is a total of 400%), the accumulated rate of normalisation/depreciation has resulted in User A's 3 year old computer now currently earning 25 ppd. The average user is folding 100ppd - or 4 times more.

However, User A chooses to buy a new computer, a new computer with 'relatively decent performance at this point in time'. Because of the underlying improvements to computers, this computer is 400% more powerful than his old computer. Because we have accounted for this increasing performance improvement, his new computer earns 100ppd - the same as the 'average user'.

Where this is different to the existing system is that under the old system, the first computer would earn 100ppd indefinitely. The new computer would instead earn 400ppd - four times as much. User A's original computer perpetual earns 100 but the average user is now earning 400 ppd, so the point accumulation rate of their original computer is still 1/4th of the average, or 25/100. Under the existing system, each new generation of computers earn exponentially more points. Ppd is eventually measured in trillions. Under the new system, this does not happen.

I wanted to get that point across before looking at the QRB system, but everyone kept pushing their own agenda and polluting the point with FUD. My theory is that once we resolve the wheat and chessboard effect of underlying PPD, we will be left with wheat and chessboard effect of QRB - which is apparent under graph 1 above. That is, if you look at the graphs above, then the line of PPD without QRB over time becomes horizontal, and the line of PPD with QRB becomes roughly similar to the old PPD without QRB, although it is a little lower. If that does not make sense please help me to convey it better.

My own opinion has developed to think that 3 monthly (quarterly) is probably a good period in time from which to run a re-calculation. It becomes regular enough that is becomes ingrained in the FAH framework (just as reserve bank monetary meetings are ingrained in the financial system) and the amount of change is relatively small. It is infrequent enough user's current ppd is 'stepping' down in absolute terms at a rate where they can still provide important advice regarding the beta-testing process.

If there is anything above that anyone would like me to explain better than I am more than happy to do so.

k1wi


[Add]mdk777:

I understand your point of view, but would argue, the only reason why last year's donation looks bad is because the current system rewards technological improvement. Remove that and last year's donation only looks bad if you are putting relatively more effort into this years donation (say because you have increased your donation from %5,000 a year to $10,000 a year). As to measuring PPD in billions or trillions of billions, I think there is a fundamental psychological impediment to using either factors or raw values so large - even using factored measuring. Even with factors there is an issue with exponential growth because every 18 months the ^x value will also double!

Where is the incentive for me to spend the $$$ for a MP rig I do not see it, perhaps I am missing something. Under normal circumstances a person buying a MP rig on the retail market is going to spend 15 X to 20 X the amount of a person buying your daily driver where is that addressed. Most people are not going to spend the $$$ for a MP rig if there is no Incentive to do so.

The QRB return curve is still going to need to be substantial 5 X to 10 X to incite people into these purchases. If PG decides that the QRB is no longer needed the all is fine because there is no longer a need for MP rigs, but as it currently stands there is a need. I know allot of people do not like to hear that but that is just the way it is, if a person is going to spend 10 X to 15 X more most are only going to do it if there is inadequate reward there.

It still is not going to get rid of the problem of unhappy people if you adjust for this the only thing that this will do is curb points inflation what is 1,000 today is 100,000 tomorrow. It still will not change the perspective some have between the so called haves and have nots, which is what most of the griping in the folding community is actually about.

Grandpa - graph 1 shows change in PPD over time. Graph 2 is the distribution of points at a given point in time 'Normalising' the points in the way I propose does nothing to alter the proportional allocation of points at a given time zone (if you earned 14x the ppd for 4x the power before the system you still earn 14x the ppd afterwards).

At any given point in time Graph 2 is identical before and after my adjustment, because we are adjusting how points change over time, not the distribution of points at a given time.

And in regard to your final paragraph - as I said, most of the issue is a psychological one, 10ppd vs. 140ppd (14x proportion) looks far less unfair than 100,000 vs. 1,400,000 (14x proportion).


Edit: fixed arithmatic)

And - regarding your middle paragraph (apologies for splitting this out into three posts!) - I'm still attempting to work out how to explain things accurately.

Please accept my apologies in needing to take some time to concisely phrase it!

The QRB return curve is still going to need to be substantial 5 X to 10 X to incite people into these purchases. If PG decides that the QRB is no longer needed the all is fine because there is no longer a need for MP rigs, but as it currently stands there is a need. I know allot of people do not like to hear that but that is just the way it is, if a person is going to spend 10 X to 15 X more most are only going to do it if there is inadequate reward there.

Okay - In graph one, the ideal situation is that our final line is horizontal. Why? Because the goal is to remove the inflation caused by technological improvement over time, which is the cause of the exponential growth in ppd. A horizontal line says "you do not earn more points in the future simply because computer power is relatively higher, rather, but you earn more points in the future because you are contributing proportionally more than you are today."

Even if we make the curve in graph 1 flat the PPD curve in graph two will still exist at any one point in time - which is the incentive to fold more. That is, if in time period c your computer 4p folds 4x faster than a 1p computer (because it has 4x the power) it will take 1/4 the time to complete, and according to graph 2, you will still earn more than 4x the PPD than the 1p computer in period c - you will still earn exactly the same proportionate increase in PPD.

The reason why my calculation does not create a horizontal line for PPD + QRB is because under the QRB inflation resulting from technological improvement is included twice - once in terms of the base points, and again in regards to the speed ratio. The first stage of my proposal accounts for that first inclusion, but it doesn't remove it from the speed ratio aspect - which is why the graph is curved and not horizontal. Ideally, there would be a way to normalise the speed ratio, to keep the reward of a 4x higher than average computational power earning y times higher ppd... The idea being the speed ratio should also accommodate general improvements in computer power.

I think, although I am not yet able to concisely word it yet, the way to do so would be to take the current formula:

QRB = ppd = base_ppd * speed_ratio * max(1,sqrt(x*speed_ratio))

And adjust it as follows: PPDn = base_ppd * (speed ratio / Yn) * max(1,sqrt(x*speed_ratio)) / Y

I am not sure whether you would also need to apply the / Yn to the max(1,sqrt(x*speed_ratio)): PPDn = base_ppd * (speed ratio / Yn) * (max(1,sqrt(x*speed_ratio)) / Yn) / Y

I guess my supposition is that PG's points system works well at the present time, therefore, if we can in effect 'freeze' the relative distribution now, then we let the points system work well indefinitely.


[ADD:] PPD + QRB should be base PPD + QRB - I think!

There is a fundamental problem with the proposal as I see it, and it actually is not the proposal, it is that it does not adjust the MP to where it should be. Right now there is a rather unique situation in the MP arena and quite a few are being built and used. But that is because of the fact that the last generation of AMD processors is better for folding than the current generation of processors, as the big super servers are changing out there MC processors for Indelargos there is a surplus of MC processors which can be bought for a reasonable price. But just in the last month a 6174 has gone up 20% a piece on the used market and still rising. With the current point system a person folding a 4P will make 5X the points folding on 6903 or 6904 over my 970 or 980X folding the same WU which is not to bad since you can currently build a 4P rig for reasonable $$$ and they only consume slightly more electricity to operate. But the new bigadv WU only gives you slightly more than 1/2 of the points of the 6903 and 04 which quit frankly I can not find anybody that is willing to fold them at this time.

Why because the economics just are not there if the reward system does not address this then you will see less and less MP rigs being built and used for folding. And for some reason people want to ignore it but I know people will not buy / build MP rigs for folding if the cost them 10 X as much to purchase and I only receive 4X or 5X as much reward for using them. So how is this issue going to be addressed. As you have it set up does it deal with the issue. Is the QRB expected to deal with it. Is there a way to figure in cost factors along with scientific need because simple economics say if the economics are not there it is not going to happen.