Page 5 of 47
Re: Change in BA requirements
Posted: Thu Dec 19, 2013 8:03 pm
by Rattledagger
sbinh wrote:With 100k ACTIVE donors now, if 50K stop folding, PG would say: "Screw them, we still have 50k others. We rule because they run our program. They voluntarily run F@h, we didn't ask them to .. lol......" --- That's how I look that what 7im said.
If Kakao is anything to go by, FAH doesn't have 100k active users but 33k. Folding@home had according to Kakao a 30% drop from 01.12.2012 to 30.11.2013.
Re: Change in BA requirements
Posted: Thu Dec 19, 2013 8:29 pm
by Grandpa_01
7im wrote:I agree there also. But only if everyone was a FAH enthusiast like you, who buys hardware specifically to maximize FAH production, and to do nothing else. Don't forget there are large numbers of home and corporate donors that fold part time, or fold on a single computer. SMP is a good choice for those donors. Points are not the only incentive for them to fold, nor their biggest incentive. Clearly that doesn't apply to everyone.
When AVX comes to SMP, the whole landscape changes again.
This is a question / statement that I have often pondered what percentage of fah science is done by enthusiast, I do know several of the current WU's cannot be run reliably on occasional running machines and lower end office type machines due to qrb time restraints. SMP appears to be the bread and butter of FAH due to the shear amount of them and evidently there are not enough of the non enthusiast to run them.
I have often wondered if it would not be better for Stanford to go to a true = pay for = work point system and allow Stanford to control the flow of WU's. Meaning Machine x at x cores gets y points for 24hrs work and Stanford determines what that machine runs if it is a bigadv capable machine and bigadv is needed it gets bigadv, if smp is needed it gets smp but it still gets the same PPD as folding bigadv (=work=pay)
Re: Change in BA requirements
Posted: Thu Dec 19, 2013 8:55 pm
by Zagen30
Grandpa_01 wrote:I have often wondered if it would not be better for Stanford to go to a true = pay for = work point system and allow Stanford to control the flow of WU's. Meaning Machine x at x cores gets y points for 24hrs work and Stanford determines what that machine runs if it is a bigadv capable machine and bigadv is needed it gets bigadv, if smp is needed it gets smp but it still gets the same PPD as folding bigadv (=work=pay)
The problem I see is that, if the current regular SMP/core 17 point balance is already equalized, that would make bigadv no longer an attractive option. A 4P 4650 rig gets around 200k on regular SMP WUs (as you probably know, given you have 3 of them), and up until a couple of months ago that was literally the fastest thing you could build. I'm sure that, if bigadv didn't exist, no one would build one of those when, for $700, you can get a 780 Ti that uses 30% of the power (not counting the rest of the rig's draw) and gets better PPD with a decent overclock. Maybe they'd have to increase the k-factor on all the SMP work, so that the increasingly fast return times of a bigadv-capable server would add up to a much higher PPD.
I know I wouldn't mind if your plan were implemented and the PPD of a super-expensive machine was still well out of reach of even the top current GPUs (for the time being; obviously if I keep my current Intel 4P the same for another 2 years, I should expect new hardware to close the gap). I didn't buy my 4P because I was particularly excited about the work being done with the bigadv projects, but because the PPD was so much higher. If I were getting similar PPD on regular SMP work, I'd would not be disappointed at all in doing it.
Re: Change in BA requirements
Posted: Thu Dec 19, 2013 9:12 pm
by powerarmour
7im wrote:
Sentiment and supposition are nice, but let's try to stick with facts.
Fact: You can define equal work for equal pay any way you want, but the only one that counts in the one defined by Pande Group. It has never factored watts, electricity cost, or hardware cost in their definition of equal (scientific) work for equal pay (points). Optimizing PPD/$ has always been up to the donor.
Which is exactly why it's not always easy being a donor, we are enthusiasts, but when bigadv is involved it's generally a big money decision in terms of hardware and power usage. I'm not surprised that many of us are going to complain about the changes when they do happen.
Replacing a GPU when it runs EOL is a lot cheaper than replacing a pair of Xeon E5's
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 8:47 am
by bowlinra
7im wrote:Fact: You can define equal work for equal pay any way you want, but the only one that counts in the one defined by Pande Group. It has never factored watts, electricity cost, or hardware cost in their definition of equal (scientific) work for equal pay (points). Optimizing PPD/$ has always been up to the donor.
I think this is the EXACT problem. Value as defined by the Pande Group / F@H Project is based on the scientific work. Where we the donor can't see, touch or feel this on any regular bases. What the donors see, feel and touch is the hourly point production totals and the monthly utility bills.
As the scoring/reward system is setup by the F@H, you are providing incentive in the way of more PPD / Watts spent:
- BigAdv (~1,000 points per watt)
- GPU (latest gen ~650 points per watt)
- SMP (~125 points per watt)
If you feel the current point system is setup for "equal work for equal pay", why are the results in low demand for SMP WUs surprising?
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 9:17 am
by ChristianVirtual
7im wrote:
Fact: You can define equal work for equal pay any way you want, but the only one that counts in the one defined by Pande Group. It has never factored watts, electricity cost, or hardware cost in their definition of equal (scientific) work for equal pay (points). Optimizing PPD/$ has always been up to the donor.
Sure, donors can't change the point system; sure, PG could. Don't need to be radical, just a bit tweaking on k-factor to give preference/points to SMP. That's why I would like to hear their comments too.
Finally, The best thing we as donors can do is raising our voice.
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 12:34 pm
by Viper97
If the voices are raised but not heeded what then? An all out strike? I'm curious. FAH generally doesn't take the donors inputs and act on them. The DAB is neutered. About the only way I see to get their attention is already happening, a slow exodus away from folding and into other projects that are DC related.
Folding isn't what it once was and I came in late in the game. Even I've noticed the drop in active machines/accounts and folders. Is this born of a frustration with events? Possibly and I think to a good degree but there are other games now and ones that don't have the overhead requirements for hardware to either compete or have fun with.
It's a new day and if things don't change people will wake and realize they are getting nowhere and will move on. It's our nature to find greener pastures no matter how good are intentions are to contribute. We want some sort of reward/feel good thing. We need to know we are being listened to.
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 2:00 pm
by kasson
We do pay attention to feedback. We may not respond, but we do listen. We appreciate that this is a complicated problem. And we appreciate the participation of all of our FAH donors.
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 2:02 pm
by 7im
ChristianVirtual wrote:7im wrote:
Fact: You can define equal work for equal pay any way you want, but the only one that counts in the one defined by Pande Group. It has never factored watts, electricity cost, or hardware cost in their definition of equal (scientific) work for equal pay (points). Optimizing PPD/$ has always been up to the donor.
Sure, donors can't change the point system; sure, PG could. Don't need to be radical, just a bit tweaking on k-factor to give preference/points to SMP. That's why I would like to hear their comments too.
Finally, The best thing we as donors can do is raising our voice.
What if all the raised voices want a simple quick fix which is not good for the long term health of the project?
Should PG make every change at the whim of the horde?
If you threw points at every problem, there would be so much points inflation that every WU would be a billion points. That also devalues the contributions of long time loyal donors. Are those really the people you want to punish for a half assed bandaid?
Do you really want to decouple the science from the points and make the points completely meaningless?
And some of the alternate equal work for equal pay suggestions make it sound like you deserve more points if you pay a high electric bill. Do we really want a points for cash reward system?
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 2:35 pm
by EXT64
I agree 7im, we (the donors) do not know the long term plans for f@h so our suggestions may seem foolish to those who know. Of course PG is hesitant to say too much about those plans as nothing is set in stone (particularly in regards to the schedule).
However I feel having SMP/Core 15 getting much lower points than everything else already decoupled from the science. The fact that PG is trying (at least how we interpret it) to get more to run SMP seems to indicate it has a higher priority than bigadv. The fact that (currently) there is only one public core 17 project and many core 15 and smp projects indicates that (currently) more researchers are getting valuable science done on smp/core 15 than core 17. Perhaps we need an uber complicated point system, like a currency exchange, where the points of the projects is based off of demand by the scientists
(not really suggesting that
).
(It is also possible that some bigadv projects are nearing completion and there are too many rigs for the future runs*clones, but that is pure speculation).
So there is the crux of the problem: if you couple PPD to ns/day (equal
computational work for equal points) SMP will be crushed compared to GPUs (which is its true competitor, bigadv seems a smaller niche that is harder to compare). However the backlog of SMP units (and core 15) seem to indicate they have a higher priority to the scientists (equal
scientists need done now work for equal points).
Perhaps all of this will change early in the new year. Perhaps a combined CPU/GPU core will come out. Perhaps many projects will switch over to Core 17 or many new Core 17 projects will be released and SMP (and perhaps bigadv) will completely fade from everyone's radar and all this discussion will be pointless. But if we just extrapolate the current situation (a dangerous thing to do, I know), things seem a bit off.
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 3:22 pm
by HaloJones
Arguably BigAdv and now Core17 are actually the cause of much of the problem. I bought a bunch of GTX670s to run core17 but now that they are in scarce supply and my points production has halved I feel cheated. Also, because of their output under core17 I all but stopped my CPU folding. If work is backing up on older projects, irrespective of whether it devalues previous work, the points need to be raised to parity with other work else the units will either be dumped or folders will move on.
We already have points inflation, the question is only whether to apply it to new processing of old units.
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 4:48 pm
by ChristianVirtual
7im wrote:
If you threw points at every problem, there would be so much points inflation that every WU would be a billion points. That also devalues the contributions of long time loyal donors. Are those really the people you want to punish for a half assed bandaid?
Do you really want to decouple the science from the points and make the points completely meaningless?
And some of the alternate equal work for equal pay suggestions make it sound like you deserve more points if you pay a high electric bill.
This imbalance of points to long term donors we have already today. I'm a "perfect" example: joined in May 2013, folded ca 3000 WU and already nearly in top 1000 with 42m Points. I run over so many donors with 10'000th of finished WU on their belt and much less points. Sometimes I feel really sorry for them. But on the other side the science argument let me think: lucky me I'm late joiner and have the more efficient technology/GPU at my disposal.
To the other point: not want to make the points meaningless, just react flexible. If SMP is what is need right now more because there is science stuck in the folding pipeline it should receive a higher value to get done.
Maybe the suggestion from EXT64 is a good one and we should look into how markets are working, maybe a "spot market" mechanism to automatically lower some points for high supply/high volume WU and higher points for high demand /low volume WU.
Joe Donors like me don't really understand lots about the science in each WU, but we understand the utility bill and cost for hardware. We are obvious willing to spend that. No question. But that also define/impact our value system much more compared to ns/day..
I wouldn't link the points to the power bill directly. Of course there are huge differences in energy prices which should not influence the points. That's the job for each donor to find the right balance.
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 5:08 pm
by 7im
ChristianVirtual wrote:...
To the other point: not want to make the points meaningless, just react flexible. If SMP is what is need right now more because there is science stuck in the folding pipeline it should receive a higher value to get done.
If you consider raising the SMP points a valid solution, then you must also consider lowering the BA points a valid solution. 2 sides of the same coin. And we know it's a valid solution since they've already done that once before.
Be careful what you wish for. They may not rebalance the points the way you want them to... (in your favor).
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 5:18 pm
by HaloJones
Maybe that's already being resolved by the current scarcity of core17! It suddenly makes the output from my SMP-CPU look pretty good!
Re: Change in BA requirements
Posted: Fri Dec 20, 2013 5:20 pm
by EXT64
Conspiracy!