Folding Forum

Guru wrote:You're honestly going to sit there and say that using the number of WUs completed is not fair, but assigning some number to a WU and passing that out is????? Um, you just said that it wasn't fair, but then you turn around and say it is? lol.... FAIR is listing what you are REALLY DOING. If WU sizes are a factor, then show an average, a minimum, and a maximum, and perhaps even list the number of small, medium, and large WUs... It's really not complicated. I don't get why you can't comprehend it.

Yes, I am going to sit here and say that. Numbers of WUs completed mean absolutely NOTHING when some WUs are 20 points and some WUs are 2000 points. Some WUs take 2 hours to finish and some WUs take 2 weeks to finish. How can you not comprehend there is no correlation there? No way to make equitable comparisons? It's like you are holding 2 pennies in one hand and 2 quarters in the other say saying they are the same. I can't comprehend how you can say 2 pennies equals 2 quarters?

Guru wrote:Don't be a fool. There are people interested in how much time is being dedicated to the service, and it's a very real factor in terms of contribution.

Contribution? Yes. Foolish, no. Because time has nothing to do with the amount of science completed. Without a rate, time does not equal distance. (distance = rate x time)

Guru wrote:No, you're wrong again. All factors, when put together, make up a good comparison. I'm blown away that you think a random scoring system is better than comparing factual data. It this strictly because you've spent a lot of money on video cards?

Stanford evaluates the science completed, and then sets a benchmark score, then all work units are scored against that benchmark. Hardly random. BTW, I have ZERO GPUs folding.

Guru wrote:My idea is awesome, but you're not being logical...

Sorry, but when you said that CPUs and GPUs were about the same in that previous post, I'm doubting any logic you post.

Guru wrote:LOL So they ARE weighting GPU scores!!! Thanks for the confirmation... I'll bet it has a lot to do with financial dontations as well... Hey, I was right about the assumption that they were favoring the GPUs, so that just strengthens my odds of being right about this as well...

Yes, they are weighted... on scientific production. If you want to claim some nut job conspiracy, go post that somewhere else. Stanford has much higher standards than that, and your suggestion is insulting.

What I want is extremely simple--just a means to measure contributions. I don't rely on the scores because they can no longer be used to measure contributions; the division between GPUs and CPUs is just too great. So something has to be put in place that does measure contributions. Pande doesn't even have to touch the broken score system if they don't want to. The thread time can be added completely separately. It is my opinion that contributions are more important to the project than the results that their software produce. I have no control over client productivity but I do have control over how much time is donated to the cause. I have provided a solution which could work side-by-side with the existing score system or could be the foundation for an entirely new scoring system.

Only the donated time matters to me; how it is used is out of my hands. Please, at least see that much.

John Naylor wrote:It really is not. The 8800GT in all probability did 10x as much work in the same amount of time as one of your quad-core processors, and as a result it gets 10x the points. Your whole argument seems to stem from the INCORRECT assumption that a quad-core processor can do as much in one hour as a 8800 can do in one hour, just because they both draw roughly the same amount of power. This is, without putting too fine a point on it, rubbish.

Do I have absolutely any control over that fact? No, it is completely outside of my control. If they start doing more ALU calculations in the future, the GPU will be obsoleted. I am merely providing the tool inside of an agreement of how it can be used. It is up to the software developers to decide how effectively to use that tool. My processors shouldn't be belittled just because the developers decided to go in a different direction. That is completely out of my control.

I know it's outside of your control. But the fact is that without the GPU they would not have been able to do that much work in such a short space of time, so they are giving points based on the ridiculously huge speed advantages they get from using GPUs. You provide them with a supercomputer on a card, they give you points accordingly. You did 10x as much work as a quad core, so they give you the points as if you'd run the quad core for 10x as long. The system is as logical as it gets, can't you see that?

I see it and I disagree with it. It is favoring the client efficiency over user contributions. That has only become obvious after GPUs inflated the results.

Foxery wrote:I can't believe you guys are still spending this much time humoring these two. They don't get it, and don't want to.

@Guru: Please stop using a Mad-Libs book to write about logic circuits. It's painful.

You're so clever. You resorted to throwing insults like a child. Good for you. XD

Well the whole point of supporting faster hardware on newer technologies (or just different technologies i.e. sidestepping from x86 processors to stream processors) is to speed up the Pande Group's research. yes this is a distributed computing effort, but at the end of the day all these processors are contributing to scientific papers. The quicker F@H produces scientific papers, the quicker they can be used to understand protein folding (the project's stated goal) and therefore to possibly cure diseases. Therefore the goal for the Pande Group is going to be maximum client efficiency. Of course they would be nowhere without the users, but at the end of the day they need their work doing ASAP, and for now the fastest way of doing that is on GPUs. As a result they will try and compensate people accordingly for the huge speed increases people supply by using GPU clients (and SMP clients too). Client efficiency has to be the main aim of the project for the project to achieve its goals quickly, and you seem to have forgotten that.

The fact is most people spend more money on their hardware, they get more points, tough, upgrade get more points, dont upgrade well you get it i think. Just because you run old systems doesnt mean you should get the same points as everyone else. If you want more points upgrade, if you arent prepared to upgrade then i guess you arent going to get anymore points. Also FTR, if you run -adv and do ambers, you may make more points than you do now, but if i were you id just run SMP.

OT: I wonder how long before this thread gets moved to Todays Discussion? Its definitely worthy of a move.

Guru wrote:Don't be a fool. There are people interested in how much time is being dedicated to the service, and it's a very real factor in terms of contribution.

Foxery wrote:@Guru: Please stop using a Mad-Libs book to write about logic circuits. It's painful.

Guru wrote:You're so clever. You resorted to throwing insults like a child. Good for you. XD

If you don't want this thread to be closed, I expect you to stay on topic, without starting to throw insults at each other.

Here is the root of the problem:
http://img229.imageshack.us/img229/3826/yeahzd8.png

What you see in that picture is nothing shy of disgruntling. It is extremely tempting to stop folding on those 8 cores because that 1 GPU shames it. So I guess the question that comes from this is: Does Pande really want to lose those 8 cores (already lost two) or not (for the sake of GPUs)?

Every single contribution made gets the scientist one step closer to a result. The source should not matter. The points system, ultimately, does not play in their favor. You don't make friends by belittling their attempts to help.

7im wrote:

Guru wrote:You're honestly going to sit there and say that using the number of WUs completed is not fair, but assigning some number to a WU and passing that out is????? Um, you just said that it wasn't fair, but then you turn around and say it is? lol.... FAIR is listing what you are REALLY DOING. If WU sizes are a factor, then show an average, a minimum, and a maximum, and perhaps even list the number of small, medium, and large WUs... It's really not complicated. I don't get why you can't comprehend it.

Yes, I am going to sit here and say that. Numbers of WUs completed mean absolutely NOTHING when some WUs are 20 points and some WUs are 2000 points. Some WUs take 2 hours to finish and some WUs take 2 weeks to finish. How can you not comprehend there is no correlation there? No way to make equitable comparisons? It's like you are holding 2 pennies in one hand and 2 quarters in the other say saying they are the same. I can't comprehend how you can say 2 pennies equals 2 quarters?

So the number of things you do, regardless of size, don't count toward being a possible cure for cancer? A small one might be what it takes... Besides, remember that my suggestion was to list all applicable attributes of the folding as factual information, rather than some biased scheme as to what they think is worth more points...

7im wrote:

Guru wrote:Don't be a fool. There are people interested in how much time is being dedicated to the service, and it's a very real factor in terms of contribution.

Contribution? Yes. Foolish, no. Because time has nothing to do with the amount of science completed. Without a rate, time does not equal distance. (distance = rate x time)

Who says everyone wants to measure distance? Measure time in order to know how much time is involved... You're still caught up with the competition of the work, rather than the work and contribution involved... Not everyone that participates cares the score, but rather what the score implies... It's just too bad that the score forces you to make assumptions instead of showing you facts...

7im wrote:

Guru wrote:No, you're wrong again. All factors, when put together, make up a good comparison. I'm blown away that you think a random scoring system is better than comparing factual data. It this strictly because you've spent a lot of money on video cards?

Stanford evaluates the science completed, and then sets a benchmark score, then all work units are scored against that benchmark. Hardly random. BTW, I have ZERO GPUs folding.

It's random when it is decided by whatever scheme that is thought up by some random person in charge and that scheme can change at any time... I hate repeating myself, but some people just can't get it... Get rid of the score and post FACTS...

7im wrote:

Guru wrote:My idea is awesome, but you're not being logical...

Sorry, but when you said that CPUs and GPUs were about the same in that previous post, I'm doubting any logic you post.

Oh, so they are not both processors that make calculations... I can't believe that I thought that the two were basically the same with differences in structure...

7im wrote:

Guru wrote:LOL So they ARE weighting GPU scores!!! Thanks for the confirmation... I'll bet it has a lot to do with financial dontations as well... Hey, I was right about the assumption that they were favoring the GPUs, so that just strengthens my odds of being right about this as well...

Yes, they are weighted... on scientific production. If you want to claim some nut job conspiracy, go post that somewhere else. Stanford has much higher standards than that, and your suggestion is insulting.

[/quote]

The fact that they use a scheme to score hardware in order to be competitive in such a way that money could be made from manipulating numbers does in fact suggest there there is a hidden agenda... The fact that so many of you are refusing to change the software for the better is only further evidence of the same. You're not even willing to hear logical solutions from real programmers that make their companies millions... It's ridiculous...

The reason why we're rejecting your suggestions is that they ignore the scientific value added. Under the current system you get more points for doing more science. Under your system you merely get credit for having your machine switched on, and no distinction is made for how much science is being done. Yes, every extra work unit matters no matter how much science is actually done in processing it, but under the current system those units which involve more work done are valued more highly than those involving less work. I fail to see the problem with that.

You can keep the current score system. I honestly don't care too much about it. I just want to see a thread time field added and a chart of top contributors (based on thread time). That gives me, and others, an incentive to keep folding on all these CPUs. They may not contribute as much but they are contributing nevertheless.

If you'd said that at the start then you may have found us a bit more receptive... I don't know how they would backdate it though... just for CPU clients I know we are on at least the second benchmark, maybe third... add in GPUs and SMP too and the amount of time taken to do so would be enormous (there are I believe over 1,000,000 individual usernames that would need backdating)

I don't see how a Q6600 is being outpaced 10x by a 8800GT. I have both of them myself (I own 5 quads and a 8800GT, 8800GTS 640MB, 9800GT and GTX 260). Each box has a Q6600 at 3.2 GHz and it produce about 4200 PPD while a 8800GT on the same box do about 4500 PPD. PPD is a metric which mean Points Per Day and is a average of the numbers of points generated by period of 24 hours. While the GPU can do 10 WU in the same time frame and a CPU will do about 2-2.5 WU per day, a GPU WU value is 480 points with a SMP WU is worth 1760 Points. I can pull that since each Q6600 run 2 Linux clients under VMWare to optimize the production.

That's why the current system works and old hardware is obviously a waste to fold IMHO. You can stuff 40 P4 systems in a single 8800GT for about 120$ shipped and yet, you complaint about this.

Who want my worn out dead horse beating stick ? I don't need it anymore since I cannot find any dead horse trace in the puddle on the ground.

Better late than never. Like I said before, I would use a formula something like this:

base time = current score * (average time / score point)

It won't be 100% accurate but it should be close enough. The catch is, the longer the wait to implement it, the less accurate it will be because of GPU clients (to a lesser extent, SMP clients) inflating results.