Folding Forum

It is my understanding that originally there was a 10x cap but that it was later removed.

10x caps didn't work when they started using K values of 20+ instead of just 2.

And it's not like anyone was near that number until just recently.

Another glitch with the original concept.

All this does is prove the incredible power of exponential functions.
(something most of us are well aware of)

Nothing is being demonstrated about the benefit to solving the underlying science. Neither pro nor con.

I will leave you to debating the most ascetically pleasing curve and equation.

Anyway... if people can't see the numbers and come up with their own solution, I'll lay this out in simple terms.

We can't have a BigWU bonus system for any length of time that is anchored to a 4 core i5 chip. The right side of the graph approaches infinity too quickly. IMO, that should be obvious.

Leave everything exactly the same (bonus forumula), but use an 8 or 16 core chip to do the benchmark with the exact same numbers (WU data) from that FAQ I linked. PG needs to put some real world numbers down instead of using some 3 year old estimate on 3 year old hardware.

Then let the PPD fall where it may. Poll that!

7im wrote:Why triple the points when the power is only doubled?

Why is power related to points? Isn't points related to time related to power?

And time is related to the amount of work being done, the measurement points per day indicates this clearly.

I would like to know if my meager contributions are actually worthwhile. It seems that quickly returned work units are several magnitudes more important to the project than the number of work units. Has the project advanced to a point where my stock clocked Q6600, for example, is more of a hindrance to the project than a help due to how long it takes to complete a work unit?

No, that is not the case at all, at least not intentionally, or at least not until lately. <caugh>

But when PG caters to the top 1% of folders and leaves the rest of us to the scraps, one does begin to ask questions just like this.

Points appear to be devalued, and thus peoples' contributions appear to be devalued.


FlipBack looks to be +1 for a change.

Rattledagger is another +1.

http://foldingforum.org/viewtopic.php?p=189567#p189567

But, atleast my opinion is the "pay" is too low on the FAH-single-cpu-client compared to the SMP-client to be worth it, assuming fast enough computer to run SMP-client that is...

One might conclude that SMP pay is too low as compared to -bigadv, and people should stop running smp.

FlipBack wrote:I would like to know if my meager contributions are actually worthwhile. It seems that quickly returned work units are several magnitudes more important to the project than the number of work units. Has the project advanced to a point where my stock clocked Q6600, for example, is more of a hindrance to the project than a help due to how long it takes to complete a work unit?

I've been following this thread, and I agree with this sentiment.

My old GPU/Quad CPU farm sucks (and so does Atlas Folders) judging from the the fact that the same work can be done by a single SR-2. Although I'm not sure where the above quoted SR-2 figures of 250-300k PPD come from, as most of the stats I can find seem to indicate 110-140k PPD.

Since my old gear isn't generating much value relatively speaking, I may as well save the power until I can get an SR-2 online, or just wait until the next X79 comes out.

- JP

To make it clear for anyone, you're complaining about the gpu work unit's not yet getting a QRB. That does not imply the need to change the qrb for work units configured for it, it implies the roll out of qrb needs to be accelerated.

7im you include smp and compare it with bigadv but you don't take into account that the qrb isn't the only factor influencing the bigadv ppd. Change the benchmark system to reflect the high core systems running bigadv and if needed make it hard coded requirements before a unit is requested ( based on a performance indication not a count of something like compute engines without taking into account how efficient they are ). The benefit of that is that the performance data could be used to make a final revision to the qrb after the roll out has been completed, not before though as that is premature.

If you are suggesting a change, as you seem to be doing, why only go half way? Why only a bandaid? A benchmark computer change only delays the inevitable. Why not fix the whole problem?

JPinTO wrote:

FlipBack wrote:I would like to know if my meager contributions are actually worthwhile. It seems that quickly returned work units are several magnitudes more important to the project than the number of work units. Has the project advanced to a point where my stock clocked Q6600, for example, is more of a hindrance to the project than a help due to how long it takes to complete a work unit?

I've been following this thread, and I agree with this sentiment.

My old GPU/Quad CPU farm sucks (and so does Atlas Folders) judging from the the fact that the same work can be done by a single SR-2. Although I'm not sure where the above quoted SR-2 figures of 250-300k PPD come from, as most of the stats I can find seem to indicate 110-140k PPD.

Since my old gear isn't generating much value relatively speaking, I may as well save the power until I can get an SR-2 online, or just wait until the next X79 comes out.

- JP

We need to back this turnip truck up a little bit, that has to do with the cpu technology not the QRB my 970's and 980's can do around 10 smp or a4 WU's a day for around 55,000 PPD a sr2 would be almost double that, and there is very little risk involved. If I lose a smp or a4 WU it is only 2.5 hrs work if I lose a 9300 it could be 72 hours work. Should a cpu that can do 2 a day receive the same value per WU as one that can do 10. The QRB is just that Quick Return Bonus and was designed to give a bonus for faster returns. Thus giving an incentive to those that wanted to upgrade their hardware to do so. As far as GPU's go they are very limited in what they can do a few years ago they were the kings of folding but technology has passed them by in another year or so they could be the Kings again (who knows).

Grandpa_01 wrote: We need to back this turnip truck up a little bit, that has to do with the cpu technology not the QRB my 970's and 980's can do around 10 smp or a4 WU's a day for around 55,000 PPD

Right, thus my question. If quick returns are as important as the exponential function makes them out to be, is my Q6600 actually worth running when there's plenty of other CPUs out there that could fold the same WU much much faster?

I'm not here to complain about not getting enough points, I don't care. I just want to know if my contribution actually has worth. Right now I'm inclined to think the scientific returns probably aren't enough to outweigh the power draw anymore. Unfortunately, I cannot afford to upgrade to a more powerful/efficient machine right now.

To make it clear for anyone, you're complaining about the gpu work unit's not yet getting a QRB.

I have read (and contributed) concerns, discussion, and suggestions. I've seen little in the way of complaining.

FlipBack wrote:

Grandpa_01 wrote: We need to back this turnip truck up a little bit, that has to do with the cpu technology not the QRB my 970's and 980's can do around 10 smp or a4 WU's a day for around 55,000 PPD

Right, thus my question. If quick returns are as important as the exponential function makes them out to be, is my Q6600 actually worth running when there's plenty of other CPUs out there that could fold the same WU much much faster?

I'm not here to complain about not getting enough points, I don't care. I just want to know if my contribution actually has worth. Right now I'm inclined to think the scientific returns probably aren't enough to outweigh the power draw anymore. Unfortunately, I cannot afford to upgrade to a more powerful/efficient machine right now.

I look at it this way: I to have a q6600 and it doesn't even run the 3,4 mentioned in the benchmarking faq.

I run it because I know there are allot of work unit's waiting to be folded -> http://fah-web.stanford.edu/serverstat.html ( lucky me there isn't a wu shortage now! )

Since there are work units for my hardware, my contribution will be valued, maybe not with the amount of points I would like to get, but I will get points. The qrb is there to incite you to upgrade, just as you I can't afford a decent 2p system. Points per watt are very important to me but then again I can't afford to upgrade to recent more effective hardware. That doesn't stop me from folding wu's on what I do have, as long as there are work units being given out for them and I'm able to run them.

There are long running projects which rely on smp donors to finish their research subjects/reach their research goals. That's enough worth to me alongside with not being botherd with the higher ppd from modern high investment / capacity machines.

If the science benefits from it that much, let's reward them equally.

@Leonardo bad choice of words

@7im because you need to see the whole picture before making an adjustment? Wait untill qrb is applied to gpu and let's see how things look then.

Also, I want to concentrate on bigadv as it's a special case. I don't expect you will ever see it happen on the gpu side for instance.

It might be a big difference between a gt240 and a gtx580, maybe even big enough to say the gt240 is obsolute and delays the project. But f@h benefits from wide as much as fast, so I don't think we should ditch the slower hardware, just make sure projects are concentrated on likely capable hardware so it's not slowed down by the snails in the pack, that way you're validating speed is essential twice instead of once. Will that make the q6600 get more points? Probably not but it will ensure it keeps folding as long as there are work unit's for them. Edit: in it's core -bigadv is the indication that diversification based on actual performance is going to be a bigger part of folding.

FlipBack wrote:I'm not here to complain about not getting enough points, I don't care. I just want to know if my contribution actually has worth. Right now I'm inclined to think the scientific returns probably aren't enough to outweigh the power draw anymore. Unfortunately, I cannot afford to upgrade to a more powerful/efficient machine right now.

Yes, your contribution has worth. (Actually, the worth goes down quite radically if you exceed the Preferred Deadline, but I don't suppose that's a problem for you.)

Start with a simple uniprocessor that can meet the Preferred Deadline. Your quad can complete perhaps 10x as much work, and complete it in less time (and especially if the assignments require significantly more RAM than the WUs for the uniprocessor) so it's worth maybe 100x as much to the project. The uniprocessor can't compete with your quad, but it's contribution still counts, especially since there are a lot of them and there's plenty of work that is less critical than the work that's being assigned to you.

Repeat the same argument where your mid-range machine is being compared to a high-end machine. They're doing work that have even tighter deadline requirements and can only be done by bigger/faster machines.

Some work fits the Uniprocessor; some work fits SMP; some work fits bigadv-8, some work fits bigadv-12, some work fits a GPU, some work even fits a PS3. It all has value, but there's still plenty of room for growth. There are still proteins that need to be studied that are bigger and need more {RAM, GLFOPS, etc.} than is available today so we'll have to see what the next generation hardware brings and what improvements in software might bring, too.