BA changes for curent WUs

[ChasR]

Here's some news from the DAB. Please keep this confidential:

Re: Bigadv-16 first beta upcoming

Postby kasson » 10 Feb 2012, 14:42
Ok, based on preliminary data we have a tentative first-round adjustment to the bigadv points. (Notice all the qualifiers here--this is a work in progress.)
Over the bigadv-capable systems we have sampled (reporting 24-128 cores), we're seeing 8101/6904 PPD ratios at about 0.32, 8101/6903 at about 0.33, 8101/6901 at about 0.51, and 6903/6901 at about 1.55. This also confirms what we've known for a while, that 6903 & 6904 were over-valued with respect to 6901 and friends.

WIthout trying to fix everything at once, we're considering the following changes to base points:
8101 -> 2.2x
6903 & 6904 -> 0.8x
6901 & 6900 -> 1.15x

This would take us to 8101/6904 at 0.89, 8101/8903 at 0.91, 8101/6901 at 1.02, 6903/6901 at 1.08.
This is in part correcting for the historical issue of high values for 6903 & 6904, where we reduced points for 6901 as part of an overall bigadv markdown.
Of course, here we're not looking at the slope of the bonus values. We could do a more sophisticated set of fits that consider the k-factors as well, but let's start with this and see how we do...

 

[ChasR]

Well, since nobody gave any feedback, I'll tell you what I think.

Nuts!

In June, p6900 and p6901 values were cut because they were too high, so on a 2600K @ 4.8 they went from 64,000 ppd to 50,000. Now they've been put back up to 60,000 ppd. What the hell are they doing? Maybe better put is what the hell is Peter Kasson doing?

According to what I've read, the value of a BA WU was originally intended to produce about 20% more than regular smp. The BA12 Work is supposed to provide a small performance boost over that. With a chance to fix things, they have instead chosen to keep BA work completely out of line with all the other forms of folding.

That doesn't bother me so much as what comes next. With BA ppd providing an incentive to abandon all other forms of folding, the values of all other forms will HAVE to be raised. Lets face it, a 4P M-C @ 2.475 GHz is 3.4x faster than a 2600K @ 4.8 GHz, but it is getting awarded 16 x more ppd. If someone like me turns off all their existing rigs and builds a 4P rig, a LOT less science gets done, but I'd make 100,000+ more ppd. On the other hand a LOT more of Peter's project gets done. It makes me wonder if he's intentionally bollixing the point system for personal gain.

Anyway there will have to be an incentive to keep guys like me form abandoning my desktop rigs and gpus. That means yet another round of serious points inflation, which you guys know I really hate that.

In the end, they are cutting BA16 work by 33%, so now it will only be 100% higher than regular SMP work (using my 2600K to measure).

 

['Cuda340]

Sorry ChasR, didn't notice this up here.....

I admit, i haven't paid much attention since i found about their 16 core limit if ya want to go to the BA party........I'm more likely to stop folding before i go all server grade equipment though.

My folding future is very much undecided at this point.

Thanks for keeping us updated.

 

[dfonda]

I thought I posted in this thread, but I see I just thanked you ...for the info. My thoughts are contrary to yours, I'm gonna go where the electricity savings are and therefore money.

If I am going to keep contributing its gotta fit in to my budget and right now that means 4P. (If and when I ever get a bo'ed Mobo).

I can run 1 4P all summer, not so with my GPU's. So although I am clearly chasing points. The real reason is I love to build new PC's, the science comes 3rd for me.

I don't have as much invested in this as you Charles, and certainly respect all the knowledge you have on the working's of the Pande Group. Its just a hobby for me, and its just going to remain so no matter what they do to the points system. In my case ignorance is still bliss.

 

[harlam357]

I wish PG would just get their formula(s) straight and stick with them for a considerable length of time... now, I won't say that I haven't enjoyed my new found PPD but I certainly see your point Charles; and it sure does hit dome a lot more for guys like you and Hayes who have many, many, many individual machines tied up in it. For small farmers like me it's just a lot easier to flip machines.

For what it's worth, I support the leveling of points for BA work.

 

[ChasR]

Sorry ChasR, didn't notice this up here.....

I admit, i haven't paid much attention since i found about their 16 core limit if ya want to go to the BA party........I'm more likely to stop folding before i go all server grade equipment though.

My folding future is very much undecided at this point.

Thanks for keeping us updated.

While our forum probably isn't a reflection of the real world, what I see happening every day is a loss of interest in regular SMP and GPU folding. The ppd awarded for BA work is obviously at least as powerful a disincentive to those without a 4P rig as it is an incentive to those with one. Most who build one, use it to replace all their other folding machines. I don't see that as good for the project.

 

[harlam357]

Most who build one, use it to replace all their other folding machines. I don't see that as good for the project.

Quite true. I'm not shutting down all my other machines but I am slimming down to my two 2600k machines and 'no GPU Folding'. Of course, I have to keep at least a GPU around to test with HFM, but it won't be a 365 operation moving forward.

Power draw is my main concern. When I get everything moved around I suspect my overall draw to be ~1kw. This will be a 30~40% drop in my overall power use with less heat generation and thus less cooling necessary in the summer.

 

[ChasR]

As it is now, I would be better off to stop folding altogether, pocket my savings for later use on a new rig. PGs approach is fresh meat is better than old meat, so if I ditch the old hardware, I can become fresh meat again, with new stuff, and rise to the top again in less than a year. All the 4P donors should keep that in mind, what has happened to my C2Q, GPU and 2600K fleet, WILL happen to you in a few years, when you aren't fresh meat anymore (PG's points inflation, Moore's Law and the QRB guarantee it).

The points aren't really worth anything except as a measure of one's contribution. If the contribution isn't going to be fairly recognized, a lot of folks will quit or simply never start. 4.7 x the ppd of regular SMP for BA work on A 4P is far from fair as I see it. Kind of hard to talk your friends into Folding, telling them the competition is so much fun, and BTW, you'll have to have a 48 core rig to be competitive.

I may have a bit of sour grapes going on ATM, but I do think that things that make me (and cuda and others) lose interest or want to quit, aren't good for the project. I long for the day when, if you were twice as fast, you got twice as many ppd.

 

[dfonda]

How do you like those 8031's Chasr? 20kppd on a 560Ti!! That's a bit of a change, is that good or bad?

Do you have a choice of breaking down any of your rigs and selling them off for one 4P or do they all do something else besides folding?

 

['Cuda340]

I long for the day when, if you were twice as fast, you got twice as many ppd.

That would be to straightforward.......

 

[ChasR]

Dave,
20K ppd is irrelevant now. Its about the same as 20 ppd was in 2005. Is it good? Depends. I haven't invested in a GPU in a long time and 20K ppd won't change my mind.

 

[harlam357]

I long for the day when, if you were twice as fast, you got twice as many ppd.

Thanks QRB!!!

Maybe PG should just adjust the formula... flatten the exponential curve as it were.

I have already come to terms with the fact that my investment in 4P will likely be obsolete in a couple years. Hopefully AMD will come with another new cpu for G34 though to extend its life for a while. However, once it loses its Folding swagger I plan to repartition it into a bare metal hypervisor. So my plans extend beyond its life as a Folding power house.

 

[orion456]

I have already diverted several machines to other projects, or shut them off, and when the 12-core BAs end, I will be turning off almost all of FAH for the summer at least.

I'm annoyed at the constant point juggling, the ridiculous bonus points and the attitude shown by the Pande Group which shows little respect for donors. I have other things to do with my electricity bill cash and I will vote with my feet.

 

[harlam357]

I hate to see long timers feeling this way. But it's understandable for sure.

Reminds me of how things went down when GPU2 and SMP came on the scene and guys like nik, wedo, and mud decided to move on from folding... feeling their previous contributions devalued.

It's PG's fault for sure... but, as ChasR has said on multiple occasions, PG seems to be catering to new blood and not the old guard... which I think is a mistake. Most new blood is a flash in the pan and it's the old guard who have done, and continue to do, the majority of the work.

I think the level of displeasure with how things are currently going is also proportional to the time and money invested in folding hardware. Take someone like AtlasFolder for example, he put a ton of money into all those GPUs when GPU2 was all the rage... and even till a few weeks ago he was still making some good PPD for the project. However, he's also paying a hefty Edison bill for what (PPD wise) one 4P AMD can do in a 600w envelope. Now for him it's either try to sell off all that older hardware or just eat the loss... and that has to have a bitter taste.

 

[orion456]

Have you see how many zeros there are in the first two pages? On page one, 36% are zeros and on page two its 50%. That's a lot of people with substantial point totals who have totally stopped folding recently. Why is that happening in such large numbers? Why do they quit completely? Are other teams seeing a similar drop in donors?