So [H]orde, what are your upgrade plans?

[Agent_N]

I've done some reading over on the FF and I want to point out some things. Some of the regulars made some great points that many don't want to acknowledge in their bloodlust for more points.

1. I know that you guys are always trying to find ways to beat the -bigadv deadline with the least amount of hardware and dollars spent. I would like to run -bigadv myself, but I don't have the funds to put together a system outside of an i7 to run those WU's. Technically, an i7 is capable of meeting the deadlines, but should you when you ought not to given the 8 core restriction that PG has placed on the -bigadv WU's for specific reasons? Are points so important that you will find any way possible to share in the -bigadv points pie? It's like the poor complaining about the wealthy having more money; -bigadv is for expensive server hardware, and many are using non-server hardware (i7's and the like) and whining that they want more points. It smacks of Socialism to make all WU's equal in points for fairness, as some have suggested on the FF's.

2. Many of you may disagree with me on this, but where is our focus?? Is it on points, or science? Granted you can argue all day long that because an i7 is capable of -bigadv that you are doing more science in a manner of speaking, but are you doing it for the science, or the points? I know how competitive we can get accumulating points, but PG has put tiered point levels in place for a reason. If you have the money to burn for server level systems for -bigadv, then go for it, but if your hardware is regular SMP only, then be content with that. If you don't like the points that you are getting, save up and buy more hardware as you are able. -bigadv will never be enough because you will always desire more and more. I started off with 300-500PPD on two single core CPU's years ago, and moved up to 2,000PPD with an OC'ed X2, and now I have an OC'ed X3 with a 9800GT putting out between 8,000 and 9,800PPD, more if I never touch it. I would like to net 20,000PPD or more, but I don't have the cash for such hardware at the moment.

3. Those of you who say that the little guys like myself and others who produce less contribute just as much as the big dogs with 50,000PPD or greater output should put your money where your mouth is; if the accumulation of points is so important, then what of the science? "Yeah, the science is important...but those points, I gotta get a higher points fix; 14,000PPD isn't enough for my addiction anymore man" Has the science become a secondary matter behind points accumulation?

Guys, remember why we fold, it's for science, and as Tobit mentioned in another thread, there are a lot of uniprocessor WU's available to fold, but we are dedicating lots of equipment to -bigadv and SMP WU's. Which is more important, the science, or the points? Choose one, but don't give me that tired argument that more science is being done because you are folding -bigadv WU's. Certain kinds of science is being done, not necessarily more. Would not folding regular SMP WU's yield more science because they are finished faster than -bigadv units anyway? Do I get an X4 or X6 to finish the regular SMP WU's faster for science than I can with my current X3 or do I get it for more points?

Like you, I have gotten caught up in the points accumulation, but what if we focused on getting science done as fast as possible by dropping this whole effort to run -bigadv on a Thuban, since it has been proven that it can be done, though just barely, and use them for regular SMP WU's which are far more plenteous and still need to be folded?

Not everyone can be an elite folder, that is why they are -bigadv is for the elite folders with deep pockets. At least they gave us bonuses for regular SMP WU's, don't forget that you would have had less PPD if it weren't for the bonus system being applied to regular SMP WU's since early this year or late last year. PG took the -bigadv bonus system and applied it to regular SMP and no one is happy about that, why? Feel free to correct me if I am wrong about regular SMP WU's and bonus system since I only started folding SMP early this year.

I'm sorry for the long post, but I want us to think about the project, not the bonus of that project we all know as points. Perhaps we should think about how to churn out WU's as fast as possible rather than churning out slower -bigadv WU's for more points. Factoring in number of WU's per dollar for efficiency of course.

I guess I need to change my sig from "7,500 - 9,500PPD for the [H]ord" to "X number of WU's for science so far!"

 

[Kendrak]

-bigadv give standford a chance to see how a few normal SMP WU interact with each other. (from my understanding)

There is more science done with -bigadv.

Also, if PG wants the results back faster, shorten the deadline. It's that easy. Buy putting the deadline at 4 days they have said do it at this speed or faster.

I understand the points you have made. However PG has always used points as a carrot to lead us where they want us to focus.

If they need a type of work done quickly, they will put out a point structure to get it done to their liking. We (the folders) are the ones lead around by the nose
EX: There was a time that they put out some single core bonus WU. You got more ppd running the standard client than running the SMP unless it was a quad core box.

 

[Tobit]

Agent_N: I couldn't agree more. I've tried to remain quiet on the issue of -bigadv on Thuban project several folks are working on because many of these folks are my friends and I don't want to come across as offending them. However, Stanford has recommended we not do this so I can't condone it either. Yes, an X6 could meet the preferred deadline with a few hours to spare but this is not why -bigadv was created and we are potentially hurting science by mucking around with this. Many people have already stated why this is and I can't explain it much better but we need to ignore this "preferred deadline" when thinking about -bigadv and focus purely on reducing TPF if we want to best benefit science.

 

[Kendrak]

Agent_N: I couldn't agree more. I've tried to remain quiet on the issue of -bigadv on Thuban project several folks are working on because many of these folks are my friends and I don't want to come across as offending them. However, Stanford has recommended we not do this so I can't condone it either. Yes, an X6 could meet the preferred deadline with a few hours to spare but this is not why -bigadv was created and we are potentially hurting science by mucking around with this. Many people have already stated why this is and I can't explain it much better but we need to ignore this "preferred deadline" when thinking about -bigadv and focus purely on reducing TPF if we want to best benefit science.

But why not just up the deadline on the next batch of -bigadv WU to make it that you need a boxen that much more powerful?

Stanford sets the deadlines, we just follow them.

 

[Tobit]

Someone recently on FF recently asked "How is the "science" being hurt by running them on those CPU's (X6) though?"

Bruce, the site admin, answered it quite well:

Projects are assigned to broad classifications, based on concept of latency.

The latency goal for bigadv is more stringent than for SMP and similarly, SMP is more stringent than for the uniprocessor projects. The goal of the Pande Group is to do whatever they can to minimize the latency between the WU being assigned and the WU being returned for each classification of projects.

According to the topic on your team forum, the goal of your team is to maximize the latency provided you meet the Preferred Deadline. (e.g.- two hours before the deadline). That does NOT benefit science.

You've already touched on a critical distinction. When there are lots of WUs, more CPUs, even if they're really close to the deadline is a good thing. When there is a shortage of WUs, it's important to assign them to the machines that will return them in the fastest possible time. In both cases, marginal performance drags down the averate turn-around time.

What do you suggest that the Pande Group should do so that the average turn-around time becomes FASTER, in spite of people who's objective is to make it as slow as possible within the deadline.

I also see you're comfortable with the rationalization that what Kasson really meant was that anything that meets the deadline is fast enough even though that's clearly not what he said.

Just barely meeting any deadline has never been acceptable for FAH. (See numerous old topics on running two classic clients on a hyperthreaded P4. Same conflicts between scientific goals and PPD goals, and same rationalizations on the part of those who didn't care if their choices dragged down the speed of the science.)

Six core hardware can do an excellent job of improving the average latency (speeding up the average turn-around time) for SMP Projects (in a population that includes lots of quads and duals).

Latency can be managed (rather poorly) by deadlines. Latency can be managed by increasing bonuses for WUs that are returned well ahead of the deadlines. You've discovered that latency cannot be adequately managed if someone who just meets the deadline of one class of projects earns more points than someone who exceeds the deadline of a slower class of projects by a significant factor. I suppose that means that the PPD earned by bigadv when you just barely meet the deadline (say 3.9 days out of 4.0) should not be higher than the PPD you can earn when the same hardware completes standard SMP WUs. (But returning a bigadv in, say, 2 days out of 4 should get a very significant extra bonus, compared to that same hardware running standard SMP.)

Eight cores was set as an easy-to-detect criteria upon which a hardware minimum could be set. Circumventing that detection, whether by patching a client, by patching WINE, or by allowing a VM to lie for you is still circumventing the goals set by the FAH Project owners. As such, those choices are strongly discouraged, whether you choose to rationalize your way around the concept by clinging to the literal words in the EULA or you accept the choose to ignore original intentions of the FAH scientists.

 

[Kendrak]

OK.... totally get it.

Then why not just set the bonus up so that if you just make the deadline you get the same (or maybe a bit less) ppd than standard A3 and have a steeper slope for faster returns?

That way you still have a buffer if someone has a -bigadv but has to take the rig down for a few hours. But keeps anyone from thinking about using -bigadv on minimum hardware.

 

[Tobit]

Stanford sets the deadlines, we just follow them.

There are two differences between -bidadv and all other work units. 1. The points are much higher, because the work is expected to be turned in VERY quickly. 2. There is an 8 core minimum entry to the -bidadv bonus program. And #2 assures that #1 is true, regardless of where the pref deadline is set. 8 core systems were assumed to always finish the WUs with days to spare before the pref deadline, not just hours to spare.

The preferred deadline is not the line of permission, and certainly not the intention of this bonus program to have WUs turned in just before that deadline. The deadline is a safety net for when WUs get lost or deleted, so they can be reassigned. The deadline was never assumed to be a minimum performance level. And if people are playing the margins that tightly, you best not complain about missing a deadline if there is ever an upload problem. The 8 core minimum also allows for a margin of error so -bigadv can be uploaded, even if the first upload attempt fails. The auto send a few hours later will upload it with time to spare when you have 8 cores.

 

[SazanEyes]

Let me just address this point:

3. Those of you who say that the little guys like myself and others who produce less contribute just as much as the big dogs with 50,000PPD or greater output should put your money where your mouth is

I think you kind of misunderstand what we mean when we say that. Objectively, bigger producers are contributing more than smaller producers, and points are one meaure of that. However, distributed computing by definition is a group effort. I may produce 200K PPD, but that's less than 3% of our team's daily points, and it's a minuscule fraction of the total F@H daily production. The value of distributed computing is that the group effort of thousands of people provides more computing power than big university mainframes. Without the daily contribution of hundreds of thousands of clients, Folding@home would not be a successful project. That's why everyone's contribution is important, regardless of their hardware.

 

[Agent_N]

But why not just up the deadline on the next batch of -bigadv WU to make it that you need a boxen that much more powerful?

Stanford sets the deadlines, we just follow them.

Kendrak, it amounts to acting upon principle rather than because you can get away with it if only just barely. I personally don't have qualms with those running SR-2 boards crunching -bigadv, but the principle remains. More science gets done because we reduce TPF by upgrading, and we get a bonus of more points. As I asked, do you care more about the points, or the science? I will admit that I have cared a lot about points up to this point, but now I want to reduce TPF of regualr SMP WU's because that is of greater benefit being that they are crunched and returned faster than -bigadv on Thuban's and i7's.

Upping the deadline is like having more police pull you over for speeding when you shouldn't be speeding in the first place as it endangers lives. Acting on principle means that you obey the posted speed limit for your sake and the sake of other motorists; not pushing the envelope because you can ((folding -bigadv on non PG approved hardware).

If you want to fold for points, you can certainly do so, but ultimately what is more important to you, points, or science?

 

[Agent_N]

Let me just address this point:



I think you kind of misunderstand what we mean when we say that. Objectively, bigger producers are contributing more than smaller producers, and points are one meaure of that. However, distributed computing by definition is a group effort. I may produce 200K PPD, but that's less than 3% of our team's daily points, and it's a minuscule fraction of the total F@H daily production. The value of distributed computing is that the group effort of thousands of people provides more computing power than big university mainframes. Without the daily contribution of hundreds of thousands of clients, Folding@home would not be a successful project. That's why everyone's contribution is important, regardless of their hardware.

I understand what you are saying and always understood that was what individuals like yourself meant, but I was directing that comment at the point I was making: which is more important, points or science (i.e. WU's completed)?

 

[Kendrak]

OK, wow.....

I posted above, "I get it" (after reading the FF quote)

And as a SR-2 owner (someone who is trying to push TPF as low as possible), I think your throwing the stones at the wrong person. I was just trying to voice the other side.

But what about my idea? None of you are touching that. It's simple, why don't they just do that?

Then why not just set the bonus up so that if you just make the deadline you get the same (or maybe a bit less) ppd than standard A3 and have a steeper slope for faster returns?

That way you still have a buffer if someone has a -bigadv but has to take the rig down for a few hours. But keeps anyone from thinking about using -bigadv on minimum hardware.

 

[Tobit]

which is more important, points or science (i.e. WU's completed)?

Ideally, it's a combination of both. I've always said that you best benefit science by running as many different clients as you can, then just maximize what you have to get the most PPD out of it all.

 

[SazanEyes]

On the issue of bigadv, quoting from the FF quote above:

Latency can be managed (rather poorly) by deadlines. Latency can be managed by increasing bonuses for WUs that are returned well ahead of the deadlines. You've discovered that latency cannot be adequately managed if someone who just meets the deadline of one class of projects earns more points than someone who exceeds the deadline of a slower class of projects by a significant factor. I suppose that means that the PPD earned by bigadv when you just barely meet the deadline (say 3.9 days out of 4.0) should not be higher than the PPD you can earn when the same hardware completes standard SMP WUs. (But returning a bigadv in, say, 2 days out of 4 should get a very significant extra bonus, compared to that same hardware running standard SMP.)

The bolded part above is the problem. PG screwed up the bonus system, and apparently they don't want to fix it. I've said before that I think the PPD of all bigadv units should be set to the low P2684 level, to reduce the incentive. Changing the bonus so that quad i7s earned less PPD on bigadv than standard SMP would be even better. They set the rules; we just follow them. Their admins complaining about user behavior on FF accomplishes nothing. They have the power to fix this, and they should do it.

 

[Kendrak]

If you want to fold for points, you can certainly do so, but ultimately what is more important to you, points, or science?

There is no reason why they cannot/shouldn't be aligned.

On this I will point the finger back at PG for not setting up a proper point system that has people behave in the way they want them to. Do they not have Psych 101 at Stanford?

 

[Agent_N]

OK.... totally get it.

Then why not just set the bonus up so that if you just make the deadline you get the same (or maybe a bit less) ppd than standard A3 and have a steeper slope for faster returns?

That way you still have a buffer if someone has a -bigadv but has to take the rig down for a few hours. But keeps anyone from thinking about using -bigadv on minimum hardware.

Kendrak, so long as individuals value points over science, deadlines are meaningless and changing them is meaningless if you can barely make them with an i7 for example. Individual who value points will go to whatever length, as is evidenced here in this very thread with testing Thuban's to see if they can handle -bigadv deadlines for more points, to just barely meet the upped deadline. If the deadline is upped, then instead of an i7 qaud people will try with an 980X to barely meet the deadline. It won't stop utnil you act on principle. It's why I keep mentioning "principle"; so long as you violate the speed limit, even just barely, you are still breaking the law technically even though they allow a 5MPH grace buffer for speedo's that doen't always read accurately. The speedo in my car is 2-3MPH slower than what it reads according to radar trailers that are set up to make you feel guilty for speeding.

 

[Agent_N]

OK, wow.....

I posted above, "I get it" (after reading the FF quote)

And as a SR-2 owner (someone who is trying to push TPF as low as possible), I think your throwing the stones at the wrong person. I was just trying to voice the other side.

But what about my idea? None of you are touching that. It's simple, why don't they just do that?

Kendrak, I'm not throwing stones at you, I want you to consider why you fold, what is your main drive/motivation for folding? And, I did answer your question with the speeding analogy. If upping the deadline means that an i7 is no longer able to do -bigadv, but now a 980X is just barely able, nothing has changed except steeper requirements to barely meet a meaningless deadline. Does this make sense? As I said, I have no qualms with SR-2 owners as they can get -bigadv back much sooner than an OC'ed i7 qaud.

 

[Tobit]

But what about my idea? None of you are touching that. It's simple, why don't they just do that?

Because Stanford rarely changes the rules after they've been implemented. One of my gripes with them. Sure your idea might make sense but..

The big issue is that -bigadv was designed before Core i7 CPUs existed. They were supposed to take advantage of the mega-servers out there with multiple sockets . The regular SMP clients could not serve them well because the A2 cores were limited to 4 cores and the A1 didn't scale well above two cores. So these WU's were developed to take advantage of the possibilities of their extreme capabilities. They were given the first bonus points because they could do more work faster than anything else by quite a bit. Then i7's were released and it was found that they could finish the bigadv's with a little room to spare. -bigadv was never intended for these processors or x6's.

 

[Kendrak]

Kendrak, so long as individuals value points over science, deadlines are meaningless and changing them is meaningless if you can barely make them with an i7 for example. Individual who value points will go to whatever length, as is evidenced here in this very thread with testing Thuban's to see if they can handle -bigadv deadlines for more points, to just barely meet the upped deadline. If the deadline is upped, then instead of an i7 qaud people will try with an 980X to barely meet the deadline. It won't stop utnil you act on principle. It's why I keep mentioning "principle"; so long as you violate the speed limit, even just barely, you are still breaking the law technically even though they allow a 5MPH grace buffer for speedo's that doen't always read accurately. The speedo in my car is 2-3MPH slower than what it reads according to radar trailers that are set up to make you feel guilty for speeding.

You don't even understand what I posted.

If they set the point system up this way it would not be an issue.

OC-ed i7 @3.7 current

A3 WU : 17k ppd
-bigadv: 25k ppd

What it should be

A3 WU :17k ppd
-bigadv : 15k ppd

But 16+ threaded system would get similar points because the "K" factor is higher.

This would solve every issue that you have by setting the point up with the science.

And telling people to fold by principal on hardware that they donate the time/power on isn't going to fly to far.

(And speed limits is a bad example, it is a way to keep gas prices down and insurance companies happy. Most car wreaks are caused by slow drivers. Same thing with seat belt laws, we wouldn't have them if it weren't for car insurance co.)

 

[Agent_N]

There is no reason why they cannot/shouldn't be aligned.

On this I will point the finger back at PG for not setting up a proper point system that has people behave in the way they want them to. Do they not have Psych 101 at Stanford?

Oh, you had to go and mention psychology! I am a psych major and I'm telling you that upping the deadline does not change much of anything as I have already explained in other replies to you. There will always be those who don't think that it is fair that they can't get 4,000-5,000 more PPD folding -bigadv than regular SMP. Deadlines don't change that desire to push the envelope for more points.

 

[hiodie]

Kendrak, I'm not throwing stones at you, I want you to consider why you fold, what is your main drive/motivation for folding? And, I did answer your question with the speeding analogy. If upping the deadline means that an i7 is no longer able to do -bigadv, but now a 980X is just barely able, nothing has changed except steeper requirements to barely meet a meaningless deadline. Does this make sense? As I said, I have no qualms with SR-2 owners as they can get -bigadv back much sooner than an OC'ed i7 qaud.

But now they would be using the hardware needed in the first place, would they not? Upping the deadline would eliminate the people using hardware to get it done just in the 4 days time....according to your idea, all deadlines should be eliminated.

 

[Kendrak]

But now they would be using the hardware needed in the first place, would they not?

Bingo

 

[Agent_N]

You don't even understand what I posted.

If they set the point system up this way it would not be an issue.

OC-ed i7 @3.7 current

A3 WU : 17k ppd
-bigadv: 25k ppd

What it should be

A3 WU :17k ppd
-bigadv : 15k ppd

But 16+ threaded system would get similar points because the "K" factor is higher.

This would solve every issue that you have by setting the point up with the science.

And telling people to fold by principal on hardware that they donate the time/power on isn't going to fly to far.

(And speed limits is a bad example, it is a way to keep gas prices down and insurance companies happy. Most car wreaks are caused by slow drivers. Same thing with seat belt laws, we wouldn't have them if it weren't for car insurance co.)

Sorry, I guess I did misunderstand you then. I realize that principal won't fly very far, but is that not the reason that we are in this mess besides PG's setting up the point system as they have? You are still hung up on points as far as I can tell and not on WU's processed as quickly as possible, or a balanced approach as Tobit menitoned above. I feel good about having upgraded from an X2 to an X3 to meet deadlines faster, but if they keep putting out these 921 and 1395 (or whatever that one is) WU's then I will have to upgrade to get those in faster since those take more than a day to fold on my PC. I can send the 481 and 470 point WU's in fast enough so far...

Anyway, as I said, if you value points more, then none of us can stop you from folding for points, regardless of how messed up the point system is now.

 

[Cypher-]

You don't even understand what I posted.

If they set the point system up this way it would not be an issue.

OC-ed i7 @3.7 current

A3 WU : 17k ppd
-bigadv: 25k ppd

What it should be

A3 WU :17k ppd
-bigadv : 15k ppd

But 16+ threaded system would get similar points because the "K" factor is higher.

This would solve every issue that you have by setting the point up with the science.

And telling people to fold by principal on hardware that they donate the time/power on isn't going to fly to far.

(And speed limits is a bad example, it is a way to keep gas prices down and insurance companies happy. Most car wreaks are caused by slow drivers. Same thing with seat belt laws, we wouldn't have them if it weren't for car insurance co.)

This exactly. This I think would fix everything.

 

[Agent_N]

But now they would be using the hardware needed in the first place, would they not? Upping the deadline would eliminate the people using hardware to get it done just in the 4 days time....according to your idea, all deadlines should be eliminated.

Is forcing the exclusion of certain systems the best way when you can choose not to for the sake of science by folding regular SMP WU's for faster turnaround times for fewer points? Go back and reread Tobits lengthy quote post. I am not advoacting the elimination of deadlines, but what is in the best interest of science. They set the minimum limit to 8 cores, though certain Intel CPU's can circumvent that restriction, which will always be a problem unless they steepen the requirement to 16 cores/threads instead of eight. Deadline changes are much less effective than hardware requirment changes.

 

[Kendrak]

Sorry, I guess I did misunderstand you then. I realize that principal won't fly very far, but is that not the reason that we are in this mess besides PG's setting up the point system as they have? You are still hung up on points as far as I can tell and not on WU's processed as quickly as possible, or a balanced approach as Tobit menitoned above. I feel good about having upgraded from an X2 to an X3 to meet deadlines faster, but if they keep putting out these 921 and 1395 (or whatever that one is) WU's then I will have to upgrade to get those in faster since those take more than a day to fold on my PC. I can send the 481 and 470 point WU's in fast enough so far...

Anyway, as I said, if you value points more, then none of us can stop you from folding for points, regardless of how messed up the point system is now.

Points should = science

That is my point. The fact that they don't is a big mess up on PG part.

And yes I car about points. It is the reason [H] is ramping to try and forestall EVGA.
It is the reason I got a SR-2. Shorter TPF=higher ppd
It is the reason I have spend what I have, for friendly competition

It is a motivational tool, and the fact that PG can't get it set up properly is their fault.

I also don't have a problem with more hardware coming online (that wouldn't have otherwise) to fold WU of any kind as long as the science is getting done.

For PG set up a broken system and them whine when people try to find ways to up their points is asinine.

 

[Tobit]

There are inequalities all over with Stanford's point system. However, -bigadv is a truly unique project and shouldn't be compared to any other SMP projects. Also remember that -bigadv could disappear or change radically at anytime. -bigadv is still a trial, test, project. With all these people publicly posting about how they hacked the client to circumvent the 8-core requirement (violating the EULA) and also using WINE to get around the requirement (not a EULA violation but Stanford doesn't like it), don't be surprised if Stanford actually does make some changes (yes, there's a first time for everything).

Those of you playing with -bigadv on X6's, be careful with meeting the deadlines and borked WUs as you could easily find yourself dropping below the 80% success rate thereby losing bonus points on all your systems.

 

[Cypher-]

Anyway, as I said, if you value points more, then none of us can stop you from folding for points, regardless of how messed up the point system is now.

I think you are missing the point that without the points system there wouldn't be this huge race between EVGA and [H] right now. Yes I understand that we all should be folding for the science but without the points system not nearly as much science would be getting done. I think people see it as the more points they are getting the more science they are doing. The current points system has skewed that though. Stanford needs to fix it so that the points system matches what they want done.

Edit: What Kendrak said again^

 

[Agent_N]

You don't even understand what I posted.

If they set the point system up this way it would not be an issue.

OC-ed i7 @3.7 current

A3 WU : 17k ppd
-bigadv: 25k ppd

What it should be

A3 WU :17k ppd
-bigadv : 15k ppd

But 16+ threaded system would get similar points because the "K" factor is higher.

This would solve every issue that you have by setting the point up with the science.

And telling people to fold by principal on hardware that they donate the time/power on isn't going to fly to far.

(And speed limits is a bad example, it is a way to keep gas prices down and insurance companies happy. Most car wreaks are caused by slow drivers. Same thing with seat belt laws, we wouldn't have them if it weren't for car insurance co.)

Reqoute from Tobit's post in case you missed it.

"Eight cores was set as an easy-to-detect criteria upon which a hardware minimum could be set. Circumventing that detection, whether by patching a client, by patching WINE, or by allowing a VM to lie for you is still circumventing the goals set by the FAH Project owners. As such, those choices are strongly discouraged, whether you choose to rationalize your way around the concept by clinging to the literal words in the EULA or you accept the choose to ignore original intentions of the FAH scientists. "

Actually Knedrak, I would guess that the reason accident rate may be higher is because some go 5-15 faster than the speed limit which exacerbates the problem with slow drivers going 5-15 slower than the speed limit. It isn't the slow driver as much as it is the combination of slow drivers and speeders. I suppose you are going to tell me that it is safer to drive without my seatbelt on next?! I'm ribbing you here

The FAH project is not perfect, and as was pointed out in Tobit's quote post, the point system is not as it truly should be, and it seems the PG isn't going to change that, so we contineuing to hurt science, in Bruce's words, by barely making a deadline (meeting -bigadv deadline in 3.9 days rather than 2 days).

I am changing my motivation for folding from valuing points to valuing WU's returned faster. Which will require an upgrade soon if they keep putting out bigger WU's.

 

[Kendrak]

There are inequalities all over with Stanford's point system. However, -bigadv is a truly unique project and shouldn't be compared to any other SMP projects. Also remember that -bigadv could disappear or change radically at anytime. -bigadv is still a trial, test, project. With all these people publicly posting about how they hacked the client to circumvent the 8-core requirement (violating the EULA) and also using WINE to get around the requirement (not a EULA violation but Stanford doesn't like it, don't be surprised if Stanford actually does make some changes (yes, there's a first time for everything).

I welcome the changes if it brings points more inline with what science is acctualy getting done.

Those of you playing with -bigadv on X6's, be careful with meeting the deadlines and borked WUs as you could easily find yourself dropping below the 80% success rate thereby losing bonus points on all your systems.

For those who do decide to do this proper care for the WU is important. WU that get messed up, or are turned in late/never, hurts the project! I will say here the point system is a good thing here. If a borked WU or late one = no points, and thus the motivation for this goes away.

 

[Tobit]

For PG set up a broken system and them whine when people try to find ways to up their points is asinine.

Agreed. Projects have evolved nicely over the years and, for the most part, the clients are running better than ever but proper thought was never put into the points or, more likely the case, PG was/is too paranoid about making changes that would piss off certain people. Who knows.

 

[Kendrak]

OK, it seems we are saying the same thing, just on different sides of the fence

 

[Agent_N]

I think you are missing the point that without the points system there wouldn't be this huge race between EVGA and [H] right now. Yes I understand that we all should be folding for the science but without the points system not nearly as much science would be getting done. I think people see it as the more points they are getting the more science they are doing. The current points system has skewed that though. Stanford needs to fix it so that the points system matches what they want done.

Edit: What Kendrak said again^

Actually Cypher, I do understand that. Points can be a means to friendly competition, but don't forget why you fold and also EULA violations with systems that are not technically meeting -bigadv requirements. Stanford may do away with -bigadv altogether if we push our luck. Tobit makes some great points, even better than I have made.

My goal is to get into the top 200 fodlers on team 33, and points is a means to get there being that that is the measurement used to determine your standing.

I understand that the point system is messed up, so go to Stanford and fix it for them. They have other things to concern themselves with than fixing a broken point system. Had they implemented the point system correctly in the first place, then much of this may have been a moot point.

 

[SazanEyes]

OK, let me just speak for myself and be blunt: I value the points over the science. I fold for the competition and to give my PCs something to do when I'm not using them. I have run dozens of DC projects in addition to F@H, some of which have little if any 'social value." Sure, curing cancer or other diseases would be great, and my family has been affected by cancer so a cure would benefit me personally, but it's honestly not the main reason I fold. If F@H never leads to any great scientific discoveries, I won't feel like I've wasted my time.

I'm not a molecular biologist and don't understand much of the science. Therefore I rely on the point system to determine what F@H client to run, since WUs with higher point values must be more valuable to PG. SMP generates more points than "classic" uniprocessor, and bigadv generates more points than regular SMP. Therefore, on my systems which can run bigadv, that's what I run. Those systems (3.85GHz dual Xeon hex-core and 4GHz i7 hex-core) are easily capable of making bigadv deadlines. If PG doesn't like what I'm doing, they have total control over both the client software and the point system. They can have the software detect i7 processors and download only standard SMP. They can adjust the points or shorten the deadlines as already mentioned.

All this bitching about people who don't "value the science" is meaningless. If points are the problem, then remove the point system completely! The reality is that the competiton aspect is a big reason for the popularity of F@H, especially among the people who run CPU farms, multi-GPU systems, and now bigadv. If PG doesn't like the incentive system they themselves have created, they can either change it or STFU.

 

[nitrobass24]

I would like to throw my hat in the ring here.

1) @Agent_N i dont think anyone is disagreeing with you on that Science is what is important. The problem is that it really boils down to the fact that most people dont understand PGs reasoning on how/why points are set the way they are. So then it becomes simple "If I am getting more points, I am getting stanford more/better information, and thus I am contributing more/better to science." While in some cases this is totally wrong (bigadv on X6 is a perfect example), but most people dont understand, and then you have some people who dont give a shit because they are in it for the wrong reasons.

2) The points system is messed up, but not because PG set it wrong in the beginning. I blame moores law. I started bigadv when i had dual quad xeons a few years ago and unless you had one of those setups or a skulltrail or some enterprise system you could not run bigadv. Then all of a sudden i7 came out and rocked our processor world. The problem now is that PG cannot go back and change the points values on existing WUs.

3) I will also say that some people dont do it for science, they do it for fun (i.e. points). I dont think there is anything wrong with that at all. Its the reason I run F@H over other DC projects. There are other projects out there that have similar goals to cure XXX or fix XXX that would be great for society, but i choose to run F@H because its also fun to do.

 

[APOLLO]

Interesting subthread. I have pondered long and hard about the exact same points being raised by all the sides in this debate throughout this summer. Much has been discussed at length in the past to no real conclusion except, 1. the points system is flawed, and 2. there are folders/teams that shouldn't be overly obsessed with the points system and concentrate on the science instead.

There's not much truly novel in this current discussion if we momentarily ignore that it centers on a relatively new project (-bigadv) and the bonus system it is blessed/cursed therewith. However, I do harbor strong opinions formed in the last two months and some rather unexpected pieces of information surfaced tonight in the posts above, but I will not inject this lively debate with my thoughts unless people are that interested to know.

 

[musky]

OK, let me just speak for myself and be blunt: I value the points over the science. I fold for the competition and to give my PCs something to do when I'm not using them.

Bravo! Someone else finally admits the reality. I'll go one further in that I buy very powerful and expensive computers that do nothing but fold. Why? Points. I have been heavily involved in trying to get X6's folding bigadv. Why? Points. Not even my points (I don't own an X6.)

If you keep score, don't be surprised when everyone tries to get the highest score they possibly can. It is human nature, and trying to make someone feel guilty about it is ridiculous. Trying to control behavior by guilt ("you are hurting the science"...give me a break) is even more ridiculous. If you don't change the scoring system, then expect more of the same the next time a "marginal" processor comes out than can make a bigadv deadline.

You may wonder why I even care about this. Everything I run these days, with the possible exception of my I7 970 system (which is significantly faster than my 8 physical core dual X5550 system btw), would meet the intent of the bigadv program. It is more principle, and the exact opposite principle that many of you are voicing. If I didn't care about the science, i wouldn't do this at all. The same is true for EVERYONE who contributes to the project. Since I am contributing, I am going to maximize my points, and I strongly believe everyone else should as well. That is going to mean run bigadv if there is any way you possibly can. If you think that is a selfish attitude, fine. Just look at the amount of work I have turned in and the amount of money it has cost me, then call me selfish...

 

[Mtnduey]

I'm pretty new to the [H] Folding team, but I'm not new to distributed computing, private and public.


So with that in mind...pardon the TLDR


I dunno about the rest of ya, but the whole "what do you value more points or science" sounds an aweful lot like a F@H relgious discussion and equally a lot like "it's not whether you win or lose but how you play the game" speech after losing the big game. The truth is people are competitive, it's part of being human and I really hope we dont need to have a discussion over that...


The point system is an empty incentive to get folks to play the game and to a degree, it works. This is where the other projects of my youth failed honestly. Science for the sake of science will only go so far before you find yourself lean on support and a project failing to attract any new members to help with the workload. Let me ask you this, just how many folks out there are still running dedicated machines trying to discover a new mersenne prime? I do but I'm a geek and I'm okay with that.


Aside from eVGA I can't think of anyone giving any sort of return on the F@H points you gather. And like many have shown, all that even gets you is a discount on stuff in the eVGA store with may or may not be overpriced to begin with, hehe.


Anyhow...in regards to this discussion and F@H


Every single WU turned in benefits the project with the most value being given to WU's turned in quickly. The workloads vary, single cpu, multi-cpu, massive-multi cpu, GPU, etc... Pande group focuses on what gives them the most benefit. Take for example the fact that the list of Operating Systems they support is limited to Windows and Linux primarily and some OSX as of more recent days. Why is that? Well, even though there are other OS's out there that have hardware that would take these WU's and absolutely WTFPWN them in impressive fashion, they are are minority of folks willing to use their spare CPU cycles to help the cause. If they supported AIX, Solaris and HPUX they would have a whole new world of hardware at their fingertips. So if for them it was absolutely 100% for the sake of science as you suggest, they would already support it on the chance that a willing partner could be found for the F@H cause. The harsh reality is they do the best they can give to Science with the limited resources and man hours they have to dedicate themselves so long as the return is deemed worthwhile and justified.


If you still dont believe me, look at the support and offerings for Linux Bigadv, GPU Folding with ATI/AMD cards...need I go on? Why does Linux get little Focus? It's nowhere near as popular as Windows, it's a simple numbers game and Windows is way ahead on user count. Linux gets semi-decent support but again, the #'s are so heavily favoring Windows it's rediculous. Why does AMD/ATI video cards get so little love? Well, again thats numbers and in this case they have far more folks sporting nVidia gpu's running the client. Sure they're working on better support for AMD/ATI as the hardware is so bloody advanced they can't ignore it any longer, but nVidia gets the focus, further proof if needed see the latest gpu3 Fermi only wu's.


Right now the buzz is bigadv, but not that long ago GPU folding was king and CPU's folding was considered weak return for points, and well it kinda was. Bigadv tiped the scales back to CPU's but only when you had enough cores at the right speed, but for those with that hardware it was and still is king. So if you have the resources to sport a bunch of GTX295's or GTX480's you can still do wicked points, or spend an equally large sum of cash on an SR-2 and some crazy dual-cpu overclocking and go for a stupid fast TPF rig.


Anyhow, much <3 to my [H] family, I dont mean to ruffle any feathers and certainly dont want to turn anyone away from the [H] team.

 

[zero2dash]

OK, let me just speak for myself and be blunt: I value the points over the science.

Bravo! Someone else finally admits the reality. I'll go one further in that I buy very powerful and expensive computers that do nothing but fold. Why? Points. I have been heavily involved in trying to get X6's folding bigadv. Why? Points. Not even my points (I don't own an X6.)

Admittedly I'd say I value the points over the science also, but only because I can see the points...I can't "see" the science. (Speaking literally here.) However building a bunch of imaginary points is not why I fold, it's because of disease histories and issues in both my and my wife's families, + having 2 kids now (essentially....daughter #2 will be here in a few weeks ).

All this bitching about people who don't "value the science" is meaningless. If points are the problem, then remove the point system completely! The reality is that the competiton aspect is a big reason for the popularity of F@H, especially among the people who run CPU farms, multi-GPU systems, and now bigadv. If PG doesn't like the incentive system they themselves have created, they can either change it or STFU.

Good comment, and I'll expand on that and say they'll never have one without the other....
granted, I'm not a "lifer", I don't know what folding was like 5 years ago let alone 2 years ago. People value the points. The issue with the program as a whole is that it's split between people who value the points, and people who don't care about the points. Now I'm not trying to start anything with the following thought, so [dons flamesuit] -

Without the points, f@h participation would not be nearly as much as it is now.

Look, let's think of what would happen if tomorrow we all woke up and there were no points. There were teams you could fold under, but there were no points, we weren't #1, EVGA wasn't #2. EVGA would save money, they wouldn't have any reason to spend money getting people to fold under their team because points wouldn't matter. Participation would suffer, therefore the science as a whole would suffer...right?

See - points + science go hand in hand.
You can't have one mutually exclusive without the other; not now at least. Again, I'm not trying to piss on anyone's leg that does fold for the science; likewise, I'm not saying I fold for the points and only for the points - because I don't. I'm just saying, if you poll and ask, my answer is - I value the points over the science because the points are something I can see, track, follow, and understand. The science? I have no idea what's going on there. I know Vijay and Dr. Kasson put papers up from time to time on where things are, but I don't read them (being honest here) because I wouldn't understand them. I tell people all the time about folding, why I do it, etc. however, what's the likelihood I can tell that person "our computers had a named participation in a cure".? No one knows. I hope one day that happens, but what I'm saying is - I don't really think one day I'm going to turn on the news and they're going to say something about a cure for something due to the folding@home project; if anything, it's like the pharmaceutical companies will take all the credit for everything even if they don't do shit for it compared to what we do. It's like the other night, they had that donation program on the major networks for an hour looking for $$$$ donations for cancer and before changing the channel I yelled at the tv and said "I've done more for cancer research than you have" (to whatever famous rich person was on the tv at the time, I don't remember).

Anyway that's my opinion on the matter.....in a nutshell, you can't have science without the points and you can't have points without the science. I think if points were gone, participation would suffer. I would hope I was wrong about that, but given the push by EVGA and other teams to try to take us down....I don't think I would be wrong. I love you all, I love our team, I'm proud to be a member of our team and I'm proud of every member of our team, rich or poor, 100 PPD or 100,000+ PPD. We are the [H]orde, we are the little engine that could. We've had no sponsorships, and we've kicked everyone's ass and been #1 for this long now...and it's because of everyone here being like a family.

 

[Elledan]

Just pitching in to say that I would gladly forego the points system if we got detailed feedback on the science. Points won't cure people or lead to new knowledge

 

[SatTech]

Gonna throw my 2 cents in. While it's altruistic to say you value the science and not the points, unfortunately that's not how the real world works. People are selfish. While it is a good cause, there wouldn't be nearly the hardware running right now if there wasn't the point system to fuel competition, both within teams and amongst teams. Would I be running F@H if there was no point system? Sure. Would I spend thousands of dollars building dedicated folding boxes and feeding them electricity each year? Nope. I'd probably have a couple uniprocessor clients running on my i7 rig, which wouldn't be left on 24/7.

Call me selfish, think less of me, I really don't care.

My name is Ben and I'm a points w[H]ore.