Folding Forum

1. You do care more about the points than the science done eh?

2. We can't go by number of hours operational. If that was so, i would just load 100 CPU clients on my computer, with each using 1% of CPU usage.

3. Similarly, all the strong folders/folding teams would go around buying 50 5-year old celeron PCs for about say $3000, and then run all of them, since it is the "effort" that counts. That is absolutely detrimental to the progress of the project (not to mention on the power bill!)

4. The basic fact that the points are based on science done is an incentive that would benefit the project in the long-run. Imagine if it was based on "effort". An additional GPU client on my computer for the same points, but 100 more watts of power? Many people might not like it. Instead, they'll just run the CPU client (~30-40W extra) and we wouldn't have gotten the few hundred teraflops of power from the GPU clients.

I think this is a dangerous thread to join but anyway... I don't know how to say it in english but in Finnish there's saying something like "no matter what, we have to give honor to those who rightly deserves it". As we know PC CPU's are stucked in technology where they cannot rise anymore to be competetive. If in some extreme situations where GPU's are hundreds of times faster than CPU's and in slowest still tens of times faster it's a little miracle we now have something other work them to do than accelerate our games.

This is just way evolution and it isn't Intel or AMD fault. I think it's because the lack of competition in x86 market in it's 30 years history. There haven't been a need for consumers for faster processors than what they have been in the past. But we can see this change in GPU processing speeds has wakened even Intel to do something. And do it more quicly than they're ever done before. I'm speaking about Larrabee.

Same time when CPU's went few percent faster in every year, GPU's (at least) doubled their performance. This has been continuing about ten years now. No wonder why GPU's are so darn fast compared to CPU's?

Only thing we have to fear seriously is: Are we going to see this "justice" continuing in PPD credits? It really is the (only) equitable way to give credit to those who folds. But it will be hard to explain to various hardware partners, that's for sure.

And if one (Nvidia) GPU gets same points than 4000$ folding rig, I don't see thats a bad thing. It's quite opposite actually. You should definetely sell it for a good price and still get good profit even after buying few videocards along.

FordGT90Concept wrote:

JBurton57 wrote:If my 65nm processor can crunch numbers faster and with less watts than my Pentium 166, then why not reward it for doing so?
...
Performance/watt must somehow enter the equation at some point.

Because I firmly believe it is the effort that counts. Yes, individuals must weight the cost/benefits of running older hardware but, for those that only have older hardware, why not stand and be counted? Those older processors have just as many idle clocks for the software of their era as modern processors do. They too, could be more productive during menial tasks.

JBurton57 wrote:And if my GPU can crunch out the FLOPs even faster with proportionally less power, why not reward it for doing so?

Because the current system disowns all the CPU contributors out there.

Only in your head. Maybe we should give everyone on Earth a shiny Gold Medal just for waking up this morning, because my "effort" in getting out of bed is every bit as important as the President's! You haven't said anything that makes sense, haven't accepted our explanations, and your suggestions only serve to hurt the project.

You appear to have signed up for the sole purpose of whining, and are abusing Pande Group's good nature by wasting everyone's time with this nonsense. They won't say it, so I will: If you hate the project and can't be reasoned with, then take your attitude problem somewhere else.

Apologies to the mods for fighting fire with fire, but there you have it.

Well under 1% of F@H users have computers built to fold.

This 1% (if accurate) may be producing 50% of the science. check some of the teams via Kako Stats or Extreme Overclocking. You will see that one f@H addict is producing enormous points(science).

Is performance/competition more important or is sheer numbers?

Better, Is performance/competition more important or is Science?

On a side note, you disrespect the enormous amount of work that has gone into running Sequential Multi Processor (Super Computer)software on home machines. Not to mention the ground breaking utilization of Graphics Processing Units unique talents.

I have always believed that the purpose of this project is to do work (science),
not just to see how many CPUs we can all have running.

For every job that I have ever had, I was paid according to the work that I accomplished.
The more work I work or tasks that were accomplished, the greater my pay or reward.
This seems fair to me.

The rule is pay for performance.

I have several uni-processor clients running and I do not feel disowned.

If you have the time, look up the definition of altruism.
This is, after all, an altruistic endeavor.

I think the solution is actually simple enough: Get rid of points system and display three things: WU count, total processor time, and WU/processor time (efficiency). That gives you competitive types three things to compare on: 1) time--the great equalizer (user contributions), 2) actual productivity (hardware contributions), 3) hardware performance over user contributed time. Because it does change the emphasis to efficiency, the most expensive hardware may not be the best for folding. It is a way to make Folding @ Home green for those that do build computers specifically for folding.

I think I would be happy if total processor time spent was displayed. I know that information is collected because it is used to calculate points. It shouldn't be too much of an overhaul to implement.

Do we know that processor time is collected? We do WUs that give us points, and those points are based off of a benchmark machine. But that benchmark is a P4. I believe with hyperthreading turned off. What if hyperthreading is turned on? The processor may be slightly more efficient. Or what if it's a different microarchitecture, like Core or Phenom? Those processors aren't necessarily going to take exactly twice as long as the benchmark machine to crunch out a 2x points WU. They have different clocks per instructions.

Or what happens if I do fold on a GPU, but not a CPU because my CPU overhead is so high? How does partial utilization factor in? Maybe I'm playing a 3D game, but folding with my spare stream processors. Am I to be penalized for playing games? If that's the case, I'd be encouraged to turn off FAH entirely while not willing to dedicate my hardware to it, and that defeats the purpose!

In the end, effort is laudable. But people's efforts are simply not equal. FAH basically says, "Whoa, that's a fast processor. Hey... If I give you these worthless points, will you donate some spare clocks for my paper?" And we say, "Well... Okay. I guess I wasn't going to do anything with the clocks anyway."

But the honest truth is, more performance *should* get more points. And a Q6600 *should* get many more points than a P4 2.8Ghz. It does more work. And if it happens to be that a brand new video card happens to be *so good* at folding than the Q6600, then so be it.

Pandegroup is smart enough to know that they'll need to adjust the CPU points higher (or the GPU points lower) if they can't get certain WUs done in a timely fashion. Much more chilling is the notion that GPGPU may simply be the way of the future. From what I've been reading lately, for many large-scale projects like this, that's just the way it is. The hardware is practically built for these kinds of problems. It's like having a supercomputer on a card, which is what the nVidia Tesla is trying to be, which in turn pretty much is an 8800.

Foxery wrote:You haven't said anything that makes sense, haven't accepted our explanations, and your suggestions only serve to hurt the project.

Explanations are not solutions.

Foxery wrote:You appear to have signed up for the sole purpose of whining, and are abusing Pande Group's good nature by wasting everyone's time with this nonsense. They won't say it, so I will: If you hate the project and can't be reasoned with, then take your attitude problem somewhere else.

Holy flamebait. I have already offered three potential solutions. This is a problem that doesn't plague just F@H but all distributed computing projects. A solution needs to be found and the only way to do that is compromise between the contributors as individuals and the hardware they possess. Some, like I, am involved to see just how much my spare clocks can help. Others are in it because they competitively fight for the highest points (which makes it all about the hardware). A solution needs to be found that satisfies both parties.

JBurton57 wrote:Do we know that processor time is collected? We do WUs that give us points, and those points are based off of a benchmark machine. But that benchmark is a P4. I believe with hyperthreading turned off. What if hyperthreading is turned on? The processor may be slightly more efficient. Or what if it's a different microarchitecture, like Core or Phenom? Those processors aren't necessarily going to take exactly twice as long as the benchmark machine to crunch out a 2x points WU. They have different clocks per instructions.

Or what happens if I do fold on a GPU, but not a CPU because my CPU overhead is so high? How does partial utilization factor in? Maybe I'm playing a 3D game, but folding with my spare stream processors. Am I to be penalized for playing games? If that's the case, I'd be encouraged to turn off FAH entirely while not willing to dedicate my hardware to it, and that defeats the purpose!

Those are issues that obviously have to be looked at. I don't think they can be reasonably answered without knowing the priorities of Pande Group. Is their objective just work (more processors the merrier)? Is it to promote the health of the environment (fewer processors but efficient ones)? Is it to cherish every contribution, big and small? A new scoring system cannot be built without knowing the underlying priorities.

Most of you lost the initial vision of the project, which is to use the idle cycles to do something useful. With this in mind, any workunit returned, big or small, is valuable for the science.

If you are complaining about the value of certain clients vs others, you are obviously blind about the main goal. And the answer is actually very simple in that case : Upgrade or live with this until you can afford this. Myself, I have a big farm that I bought solely for folding but I know what I'm doing because I believe in the contribution to F@H and the associated costs of making this contribution.

To answer the hardware vs user base, you are having a narrow view of the whole picture. High end hardware make up for under 5% of all the computer park in the world while 95% of the computers are not high end (single or dual core, integrated or mid-range graphics, less than 2 Gb of memory) so it would be more reasonable to gather to the majority by offering a large array of clients. For the big majority, just crunching units and contributing to the science is all that matters. Some will use the points only as a benchmark to see if the production is up to the task and any dips often mean a problem. Those who fold competitively often do it while being aware of the costs and drawbacks of staying competitve.

EDIT : With the points system, they are already green since it weed out slower systems in favor of faster ones. By weeding old systems, we reduce power consumption.

FordGT90Concept wrote:Funny you mention that because x64 support is still non-existant (10 of the 11 processors run on XP/Server 2003 x64 Edition). I have an 8800 GT and 8500 GT but I will not run beta software on them (especially my server). As far as I am concerned, a lot of "sitting back" is occuring.

Having an x64 bit client does NOT make the client go faster, so there is NO point in developing one yet. Why waste time developing and support both a 32 bit and 64 bit client when the 32 bit client works just fine?

Also, the GPU client was just released OUT of beta in to production, so Pande Groups is NOT sitting back on their heels, IMO.

FordGT90Concept wrote:My problem is that now, great monetary contributions in terms of CPU expense have become small in comparison to relatively minor GPU expenses. There is no longer any incentive for me to remain folding on my 11 processors...

Like I said, if competition is your ONLY incentive to fold, then this may no longer be the project for you if you do not what to compete. However, there are many incentives to fold other than putting your name in the top 100. You still get points, so that is some incentive, and you still help the project just as much as before. You are still helping to find a treatment or cures to several diseases that you or members of your family are likely to get. If that's not enough, well...? You have not lost anything, other than a rank on a chart somewhere.

FordGT90Concept wrote:The messages F@H is sending me is that my contributions are not worth the personal cost of keeping all 8-cores running at full bore 24/7.

That's not the message they are sending, that is your interpretation. CPUs have not been disowned or penalized! They have not reduced the points you get, nor have they reduced the amount of science your systems can contribute. Only the competative field has changed. But again, folding is a personal choice, and if you deem your personal costs are not worth it, then we are sorry that you have to leave. We hope you can come back when your costs are less.

Best of luck to you.

Xilikon, I pretty much agree with what you said. I know I have overly simplified things. My only real concern here is how CPU contributors are belittled by GPUs.

7im wrote:Having an x64 bit client does NOT make the client go faster, so there is NO point in developing one yet. Why waste time developing and support both a 32 bit and 64 bit client when the 32 bit client works just fine?

Because 64-bit is more precise. Obviously, if all the clients went 64-bit, the performance would be less on 32-bit machines. Since the extra precision isn't that vital, transition to all 64-bit would be counter-productive at this time. However, there has to be some WUs out there that could benefit from double precision floats rather than single precision. Why not exploit that?

7im wrote:Like I said, if competition is your ONLY incentive to fold, then this may no longer be the project for you if you do not what to compete. However, there are many incentives to fold other than putting your name in the top 100. You still get points, so that is some incentive, and you still help the project just as much as before. You are still helping to find a treatment or cures to several diseases that you or members of your family are likely to get. If that's not enough, well...? You have not lost anything, other than a rank on a chart somewhere.

The reason why I came here and made this thread is because GPUs have, in effect, made my past and future contributions appear meaningless. The point system, alone, conveys that message. I don't care where I am in ranking because I am not the competitive sort; however, I do want to know that my contributions are valued. The scoring system is so lopsided right now that it is sending the message to me that they are, in fact, not valued. So why should I help out if no one really cares? That is not unreasonable, is it?

For those that are saying that GPUs deserve to be awarded more points than processors, I call "shenanigans"... It's an absurd idea to give gamers an advantage or to even design the program to perform FPU calculations in the first place. Think about it.

When the PS3 gets turned on, is it being turned on for folding? Think of what the statistics must be for all PS3 users that use their console for folding instead of gaming... The same goes for a high-end gaming PC... Did you buy the expensive video card to play games or to fold? How many people do you think would rather fold than game?

The answer to finding more cures is to change the programming logic. Large floating point calculations can be done through simpler calculations by breaking them down into more simplistic parts. Designing the programs to use CPUs rather than GPUs would result in more work getting done. It's simple math, really. The number of people interested in folding that own gaming machines or consoles that will actually use them for folding is much smaller than the number of people that own a normal PC and are interested in folding.

Honestly, if a GPU is outperforming a CPU that runs at 3 times the speed, then there is something seriously wrong with the software. Looking at GT90's setup, am I the only one that thinks that 11 CPUs being outperformed by a single graphics card is ridiculous? Does it honestly take 22 processors to match what a video card can do? It's unacceptable. There's no way that a server processor is 1/20th the speed of a graphics card, or no one would use processors anymore... That's definitely a programming error...

My farm produce over 40,000 PPD and it consist of 5 Q6600 and 4 NVIDIA GPU. The mix will give me about 50%/50% between the SMP and GPU clients. this is not unbalanced and I like to spread my production so even if the GPU server is taking a dump (this happen pretty often), my SMP clients can still crunch units and vice versa. Don't forget the GPU cannot run by itself and it need a host to work. This is where a cpu is still useful.

I kinda understand the feeling of seeing a n00b with GPU cards outproducing a veteran which took years to get to the current points total. However, that's a fact of life and this will repeat itself in the future so we cannot do anything.

Video cards have been around for a long time. Suddenly, practically overnight, they are 22 times faster than a processor? Get real... <sarcasm>I'm sure it has nothing to do with how many video cards that you and other people are running out to purchase in order to increase your numbers or anything...</sarcasm>

I'm not going to debate bit width and precision. That is not a concern for folding, and you can research that on your own in the WIKIPedia if you want.

I will admit that having high scoring clients does have the appearance of devaluing the older clients, though not true because no points values have been reduced.

But none of the alternatives are any better. If your remove the link between points and scientific production to make the older and newer clients score more evenly, then the value of points becomes zero, arbitrary, pointless (pardon the pun). And then how does one compete at all when there is no measure of performance? That's not an improvement, IMO.

The issues of points vs. performance has been debated for as many years as the project has been running, and no one has yet to suggest a change that would be a win-win for all parties involved. I wish I could point you at the multipage rants from the old Folding Forum that went on for 10s of pages, and still didn't come to any workable improvements.

You can't add points to the CPU clients without devaluing the work us CPU folders done over the last 6 years, or without devaluing the much higher amounts of science the faster clients are producing. And if you reduced the points given to the GPU clients to make the CPU clients more competative, then all the GPU users complain. You can't steal points from one type of client without offending the other types. And because scientific production is why we are all here, I think it best to keep the points tied to that.

P.S. I'm not trying to disuade anyone from trying to suggest improvements, but please consider that the topic may have been beaten to death many times over already. Sorry.