Change in BA requirements

[Bill1024]

They just cut out ALL 4P 24 real core socket F servers, all of them.
So now people are going to spoof the core count so that PG sees xx amount of cores so they can still do bigadv. I will not
I will use my rigs where they are wanted.
If PG really needed us to do more SMP because it is not getting done; all they had to do is ask.
Say "Hey donors for the next couple months would you please fold SMP. We are going to increase the value of the QRB as a reward"
I can bet most all of us would do it. Someone just last week built 2 socket F servers to donate to PG. 4P 24 core servers that beat the 8101 deadline no problem at all.
I my self just spent hundreds of dollars last week. Did you ask me to, no. But I tell you as of right now I am done.
I had the wife talked into buying me a GTX780 for Christmas, just saved 500$.
Leave core count alone or make it 24, tighten the deadline or just ask us for help with the backlog of SMP.
The points from bionic WCG are worth the SAME as FAH, 1,000,000,000,000 points and a dollar fifty will buy you a cup of coffee.

[7im]

I think the simple thing is that in future I won't be spending more money on building a server to fold bigadv unfortunately, I'd feel a lot more comfortable building a multi-GPU system going forward.

I'm sure that people said the same thing after the last time BA points changed. People who then built multi-GPU systems using ATI 4xxx series cards later questioned their choice.

Please remember that when purchasing hardware specifically for fah, that hardware will always and eventually go end of life. So the smart buyer spends the money on something not soon to go EOL. Or they spend a little more to get a slightly newer generation or model (to postpone obsolescense), or they choose an option that gets slightly less PPD, but that will last much longer, and earn more PPD/$ in the long run.

PG announced their plan (November 2011) to make occasional updates to BA, and have done so. This change should be no surprise to anyone.

And if you also remember the last change (BA-8 to BA-16), the new core requirements were only on new projects after the change date. Existing projects (BA-8) continued for *many* months after. I don't expect it to be any different this time. The hardware you have now still has lots of life left in it. And anyone looking to make a hardware purchase in the future is now well informed also.

[Bill1024]

I think the simple thing is that in future I won't be spending more money on building a server to fold bigadv unfortunately, I'd feel a lot more comfortable building a multi-GPU system going forward.

And if you also remember the last change (BA-8 to BA-16), the new core requirements were only on new projects after the change date. Existing projects (BA-8) continued for *many* months after. I don't expect it to be any different this time. The hardware you have now still has lots of life left in it. And anyone looking to make a hardware purchase in the future is now well informed also.

So what you are saying is 24 core will still get work for a long time and only new bigadv like say 8106 8107 when they come out will be 32 core or better?

[Rattledagger]

Since points have no intrinsic value other than competition within F@H, then I question the value of PPD/$ as a scale to compare different DC projects. What other project can give you a better PPD/$ scale (snip)

Directly comparing credit/day between projects isn't possible, but still some users (and teams) crunches whatever project is paying most credit/day for the moment.

Just to clear something up here those of you talking about changing to Boinc projects are in for a bit of a disappointment if your choice is biological (Protein Folding) I have always ran some Boinc from time to time and fah smp pays more ppd than boinc does. WCG which has it's own point system does a little better than the rest of the projects whereas it pays 7 WCG points to 1 Boinc point.

A quick look reveals the current top BOINC-computer produces 4.9 million/day (running DiRT) while no such info is available for FAH. The largest FAH-contributor produces roughly 12 million/day and would be really surprised if he's only got 3 computers running FAH...

WCG still shows the dog-points internally on their own servers, but this is expected to sooner or later be fixed so only the BOINC-credit will be shown. Even if compares with the dog-points, WCG isn't the highest-paying biological/medical/environmental project. Both in Poem@home and GPUGRID the top computer produces over 1 million/day while WCG's top is only 32 k/day (220 k/day in dogpoints).

If you are thinking you will be more completive with boinc I can tell you that it will most likely be the opposite there are allot of really big producers, I am currently helping my team out in the Christmas Challenge and With 7 OCed top end 4P's, 3 - 980X and 3 - GTX680's (GPUGRID) I cannot make the top 20 in WCG. So I would not count on that as far as greater ppd rewards go.
(snip) Anyway I just wanted to clear up a little misconception about other projects.

WCG is the 2nd. largest DC-project as far as #active users concerned and with many of them having 8 years head-start (9 if includes the UD-agent-year but these points isn't shown on external stats-sites) so obviously it will be hard catching-up in total credit. Dominating in some of the smaller projects will be easier.

(snip) especially considering whatever project you change to you'll be starting your points at zero?

If PG looses production they will reevaluate the SMP points scheme if they don't they will lose market share and I doubt they want to have that happen.

Giving more points for regular SMP lwill not cost PG a dime, and it may make some folders happy and they may stay with FAH.

FAH's point-system has been FUBAR for many years, the various bonus-points and QRB has already basically zeroed-out everyones older FAH-credit so really starting at zero again with "Folding@home 3" or another project wouldn't really be a big issue.

As for market share, FAH has already lost 2nd. place to WCG and if Kakao is anything to go by FAH has also dropped below Einstein@home when it comes to #active users.

[Grandpa_01]

Just to clear something up here those of you talking about changing to Boinc projects are in for a bit of a disappointment if your choice is biological (Protein Folding) I have always ran some Boinc from time to time and fah smp pays more ppd than boinc does. WCG which has it's own point system does a little better than the rest of the projects whereas it pays 7 WCG points to 1 Boinc point.

A quick look reveals the current top BOINC-computer produces 4.9 million/day (running DiRT) while no such info is available for FAH. The largest FAH-contributor produces roughly 12 million/day and would be really surprised if he's only got 3 computers running FAH...

WCG still shows the dog-points internally on their own servers, but this is expected to sooner or later be fixed so only the BOINC-credit will be shown. Even if compares with the dog-points, WCG isn't the highest-paying biological/medical/environmental project. Both in Poem@home and GPUGRID the top computer produces over 1 million/day while WCG's top is only 32 k/day (220 k/day in dogpoints).

If you are thinking you will be more completive with boinc I can tell you that it will most likely be the opposite there are allot of really big producers, I am currently helping my team out in the Christmas Challenge and With 7 OCed top end 4P's, 3 - 980X and 3 - GTX680's (GPUGRID) I cannot make the top 20 in WCG. So I would not count on that as far as greater ppd rewards go.
(snip) Anyway I just wanted to clear up a little misconception about other projects.

WCG is the 2nd. largest DC-project as far as #active users concerned and with many of them having 8 years head-start (9 if includes the UD-agent-year but these points isn't shown on external stats-sites) so obviously it will be hard catching-up in total credit. Dominating in some of the smaller projects will be easier.

Ummm Dirt is not a biological (Protein folding) which myself and many others that are here to find cures could care less about the stars and prime # I myself would not waste the energy on many of the boinc projects. Poem@Home or Roseta does not compare to fah as far as PPD goes I crunch them also GPUGRID is exactly what it says it is GPU folding not smp thus cannot be compared to smp (CPU) PPD.

[agrikk]

That's a shame. A ton of scientific work for Folding@home gets done with extremely high-end CPUs such as yours. Dr. Kasson wants to focus the bigadv program, that's his call. If I owned a machine such as yours, I would keep running SMP work. There's plenty of F@h scientific work that has yet to be completed and SMP is a big part of accomplishing that science, and that's the bigger picture. PPD is just a by-product of that, calculated from a formula. Personally I care more about getting science done efficiently however I can for Folding@home, and I'm happy with whatever tier I land in. PPD is not that big of a deal for me, it's just a number and is almost never mentioned in any of the resulting scientific papers. The science is.

While I suspect that most of us are here for the science and fold because it does some good, the points are what keeps us feeling like we are contributing something each month when the power bill comes in. To take points to the logical extreme: If points were removed altogether I suspect people wouldn't be nearly into folding, simply because one can't see the contribution being worth anything anymore. Do any of us really know what each project is? Sure we know we fold to help find a cure for Alzheimer's and "certain types of cancer" and that probably breakthroughs are being made based on the processed WUs we return to some lab somewhere. But we don't really see or hear about any of that.

It is the points that make us feel like progress is being made and the competitions that develop over points make it fun and add a little zest to our computers turning electricity into heat 24/7/365.

So now, with a binary change in BA points incoming (you either have 32 cores or you don't), having a 4P machine that costs $100/month to run suddenly becomes less fun to run when you once were pulling in BA points and now you are merely pulling in SMP points.

Now that there is such a huge spread in processor core counts, I agree with what someone else posted about maybe expanding the tiers available.

Why not break it up into:
<=8 cores (single processor types)
<=16 cores (for 2P rigs)
<=24 cores (older 4P rigs)
> 24 cores (4P monsters)



Right now you can either fold BA or not, and soon a 24-core beast will be only incrementally better than a quad-core i5 instead of orders of magnitude better. I expect that most of the hard core folding enthusiasts (who ultimately perform most of the work for F@H in terms of points) will retire their older 4P rigs in favor of building new 4P rigs to qualify for BigAdv instead of keeping the rig for SMP projects. Which flies in the face of why the core limit was increased in the first place (getting more rigs to fold SMP).

[Rattledagger]

Ummm Dirt is not a biological (Protein folding) which myself and many others that are here to find cures could care less about the stars and prime # I myself would not waste the energy on many of the boinc projects. Poem@Home or Roseta does not compare to fah as far as PPD goes I crunch them also GPUGRID is exactly what it says it is GPU folding not smp thus cannot be compared to smp (CPU) PPD.

Was trying to combine my comments about PPD for both your post and P5-133XL by both including the currently highest-paying project and the two highest-paying bio/med-projects.

As for being "disappointed", everyone knows, well nearly everyone knows you can't directly compare crediting across projects, but some users (and teams) still jumps to whatever is the current highest-paying project. So getting 500k/day in project X and 10k/day in project Y on it's own doesn't tell anything.

Comparing within the same project over time on the other hand is normally possible.

[bruce]

As for being "disappointed", everyone knows, well nearly everyone knows you can't directly compare crediting across projects, but some users (and teams) still jumps to whatever is the current highest-paying project.

So they're saying that if FAH just changed 1 points to 10 points (and 10K to 100K) for all projects and for all points earned in the past, people would leave other DC projects and flock to FAH?

I don't think so.

[Viper97]

As for being "disappointed", everyone knows, well nearly everyone knows you can't directly compare crediting across projects, but some users (and teams) still jumps to whatever is the current highest-paying project.

So they're saying that if FAH just changed 1 points to 10 points (and 10K to 100K) for all projects and for all points earned in the past, people would leave other DC projects and flock to FAH?

I don't think so.

No what we are saying is the disparity in points between SMP, GPU and bigadv needs to be addressed.

What we are saying is if it isn't addressed most likely you will lose the heavy hitters.

[bruce]

So you're comparing three parts of FAH: SMP, GPU and bigadv. You're not comparing FAH to BOINC (or some other DC project) [It sounded like somebody suggested the heavy hitters would flock from FAH to some project X that awards more PPD.

How you suggest Stanford resolve the disparity between those three parts of FAH? Arbitrarily inflating the points for one part is probably not the answer (FAH has a pretty high inflation rate rather than maintaining the former value of a point.) An anti-inflationary action like reducing the pay of the nation's CEOs is not going to be popular either.

[Note: Don't read into this discussion that I have any power to implement your suggestions. I don't.

[mdk777]

The point everyone is making:

part of the change in bigadv threshold is because we would like to encourage moderately powerful machines to help boost the capabilities of non-bigadv SMP projects where we do a lot of this science.

is a very flippant statement.

Those of us who have debated the merits of changes in the point system in great detail, know that it is complex and difficult.

However, simply ignoring that complexity, simply taking a que sera sera attitude seems rather lax.

There will be obvious negative ramification to donor morale and subsequently to donor participation.

Given this self-evident fact, it seems a little more thought and attention to detail would be shown.

A disincentive is not the same thing as an incentive.

Yes, removing the BA bonus for a class of machines will certainly be a disincentive to their utilization.

However, thinking that this will translate into an increase in smp with those same machines is pretty absurd.

[cokeman54]

An anti-inflationary action like reducing the pay of the nation's CEOs is not going to be popular either.

[Note: Don't read into this discussion that I have any power to implement your suggestions. I don't.

Yes it would be very unpopular to the CEO's (very few of them) but very popular to the workers (donors, thousands of us)
If you can not implement these suggestions, can you at least pass them onto the people that can. Thank you

[PantherX]

Assuming that the Equal Pay For Equal Work is implemented in a manner that unifies SMP, GPU, bigadv based on the fact that now CPUs and GPUs (FahCore_17) can run the same calculations (implicit and explicit), it isn't a pretty picture for CPUs at all (viewtopic.php?f=38&t=24225). Moreover, the PPD of GPUs depends more on the Drivers provided by the Hardware Vendors. There are recent cases where the latest WHQL drivers perform poorly on a large number of GPUs. So you end up with varying PPD on a single GPU depending on what Driver version you are using. This might not be an issue for dedicated folding systems but can be if the system is multipurpose. Moreover, would anyone be willing to fold on high-end CPUs to get significantly lower points than a mid-rage GPU? I highly doubt it unless you don't have access to GPUs or are living in an area with dirt cheap electricity. If you include a temporary bonus for the CPUs, then it doesn't follow the rule of Equal Pay For Equal Work.

The way I see it, it is a very complex issue to manage PPD given that the Hardware advancements outpace the Software advancements.

BTW, I did expect the number of bigadv Cores to increase from 16 to a higher value once Intel/AMD announces the desktop CPUs featuring 8 Cores with 16 Threads. I guess the announcement is sooner than I expected.

[Haitch]

Equal pay for Equal work is one thing, asking BigAdv users to do some SMP work because it is important for the project is another thing; attempting to force BigAdv machines to do SMP work, for 1/3rd of the "reward" that we get doing BigAdv is something all completely different. My 12 core/12 thread box got pushed out of BigAdv last time around, I bought three 24 core/24 thread machines that churned through BigAdv work, but now they're no longer needed/wanted. If the BigAdv work my boxes were doing is no longer valuable to the project, then I'll reconsider what I do with my boxes - it just won't be SMP WU's.

H.

[Grandpa_01]

I am not at all sure GPU's can do bigadv so far the largest is quite a bit shy of a bigadv WU and takes several hrs to complete. There have not been many core 17 projects released and there have been several smp projects released since the core 17 WU's were released, so that in itself makes me question the validity of GPU being capable of doing the same work at this point in time.

Allot of people are choosing either to do GPU work at this time or to build a bigadv capable machine. The reason is simple ( cost vs reward ) I tried to explain (= pay for equal work) before and it got turned into (= sience done for = pay)

Well that does not work it needs to be an = pay for = work system, say it takes (x amount of energy to do x task = y value) if a cpu takes 200 watts and cost x to run 24 hrs and get's 30K points and a GPU takes 200 watts and cost x to run 24 hrs and get's 80k which one are you going to run.

Now enplane to Joe Cpu why you want him to work for 24 hrs and receive 50k less that Joe Gpu for working the same amount of time and using the same amount of energy.

That is = pay for = work

Many have chose not to run cpu because there is no good or logical answer to that question.