Change in BA requirements

ChristianVirtual

[H]ard DCOTM x3
Joined
Feb 23, 2013
Messages
2,561
Some changes in BA requirements in 2014 coming ...

We have a policy of periodically re-evaluating the bigadv program, including the threshold required to run bigadv projects.
It is the intent of bigadv to match large and resource-intensive work units with some of the most powerful machines used by FAH donors. This "most powerful" line naturally advances with computing power. To date, bigadv has been a CPU-based program, and with increasing numbers of CPU cores and power of those cores, we have decided to lay out a roadmap of bigadv threshold changes for the next several months.

Feb 17 (two months from today): bigadv threshold will become 24 cores
Apr 17 (four months from today): bigadv threshold will become 32 cores

We want to give advance notice of these changes, and we recognize that change is not always welcome or comfortable. We should also emphasize that the science performed by donor machines is valuable in all parts of the FAH project, and 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.

We also recognize that core count is not the most robust metric of machine capability, but given our current infrastructure it is the most straightforward surrogate to evaluate.

https://foldingforum.org/viewtopic.php?nomobile=1&f=24&t=25410

Some of you guys will not be disturbed; but the low-end BA folder might get hit ...
 
My plan of a 2P 2680v2 seems still valid (or need to try harder to get those 4650; difficult where I am).

Really hope PG get QRB done for regular CPU folding and keep it interesting to run 16 and 24 core systems.
 
The only practical performance requirement is preferred deadline time, not core count :)

Core count should only be assignment *hint*.

If deadlines aren't changed, I will be sure to accommodate all BA-capable machines that do
meet preferred deadlines regardless of core count.
 
The only practical performance requirement is preferred deadline time, not core count :)

Core count should only be assignment *hint*.

If deadlines aren't changed, I will be sure to accommodate all BA-capable machines that do
meet preferred deadlines regardless of core count.

+1 for sure!

they didn't say anything about tightening up the deadline times (not to say that this isn't part of their master plan), so I do wonder what is driving the increased core-count criterion...
 
+1 for sure!

they didn't say anything about tightening up the deadline times (not to say that this isn't part of their master plan), so I do wonder what is driving the increased core-count criterion...

...maybe bigger bigadv... time will tell.

EDIT:

Some of you guys will not be disturbed; but the low-end BA folder might get hit ...

Define low-end BA folder.
 
It seems kind of counter-productive since the BigAdv community is small compared to the rest of the F@H community. We're going to nix the niche community further, which is doing projects that others can't do so we bring in new people? Wat? I mean those of us with HD7000/GTX 600 Series cards on up on the GPU side of things are doing twice the work of anyone with a GPU prior. Except there's only one project a year later with Core_17. Or is that just a reference toward CPU folding?

Want more people? Up the point count because technologically speaking it makes very little since to run CPU folding these days with powerful SMP rigs and GPU's for the point payout.

I'm angry for you BigAdv guys who are well invested just like how the last update from 16 screwed a lot of people with Intel rigs.
 
Few more observations:

There aren't many 16-thread (as that's what their requirement really is) BA-capable machines right now.

P8101 deadline is tighter than most 16t machines (think dual intel quads w/HT) so this change (without
pulling anything from the sleeve) will only affect high end AMD 4P quads.

But then, OTOH, low-end AMD 4P hexen perform about the same.

So either:
a) they keep preferred deadlines and AMD 4P quads are made to work *wink* *wink*
b) they shorten preferred deadlines and cut all AMD 4P quads, low-end (and possibly mid-range)
  AMD 4P hexen and most (if not all) 2P 1366 setups

While (b) may not happen in Feb, I will be surprised if it doesn't happen in Apr (24->32).

So, what kasson should be asked to provide is: preferred deadlines effective Feb 17 and April 17
Core count information is useless to BA folders.
 
Last edited:
It seems kind of counter-productive since the BigAdv community is small compared to the rest of the F@H community. We're going to nix the niche community further, which is doing projects that others can't do so we bring in new people? Wat? I mean those of us with HD7000/GTX 600 Series cards on up on the GPU side of things are doing twice the work of anyone with a GPU prior. Except there's only one project a year later with Core_17. Or is that just a reference toward CPU folding?

Want more people? Up the point count because technologically speaking it makes very little since to run CPU folding these days with powerful SMP rigs and GPU's for the point payout.

I'm angry for you BigAdv guys who are well invested just like how the last update from 16 screwed a lot of people with Intel rigs.

The project just seems incapable of learning from its past.
They need more people folding SMP. So they are making it harder to get into Bigadv rather than incentivizing non-bigadv smp. They did it not to long ago with the push to 16 cores. They also reduced deadlines. Although they were targeting those who were folding bigadv wu's on 4c chips cranked to the wall. I think they will find there are not many slow machines folding bigadv anymore. This push seems rather pointless.
 
What will happen to the SR-2 folders?

Exactly. We can't extrapolate anything from that announcement.

I've sent a PM to Peter Kasson in the FF. Content follows:
Dear Peter,

Thank you for the announcement.

Given significant performance variations between same-core-count-hardware,
what bigadv community would appreciate even more is information about projects'
preferred deadlines effective Feb 17 and Apr 17.

Only that will allow us to plan/coordinate future bigadv donations.

Thank you,
Kris
 
Remember that a 7970 running Core 17 gets more science done then most current big adv rigs according to Proteener.
 
"More science" is a relative term. Performance is a factor, sure. But so is importance/priority and that has
nothing to do with performance.
 
Thanks for that Nathan, I wanted to make sure my topic got through the filters (was my first post at FF).
 
From reading Kassons statement and reading between the lines, there are more people running bigadv than needed at this time, they know exactly how much they will lose by implementing the limits, they are not stupid. He pretty much said they need more SMP and less bigadv so I would guess one way or the other they will achieve their goal.

It really does not matter what I or anyone else thinks about it or whether we like it or not, in the end it will be what it will be. And we will adjust over time. :rolleyes:
 
From reading Kassons statement and reading between the lines, there are more people running bigadv than needed at this time, they know exactly how much they will lose by implementing the limits, they are not stupid. He pretty much said they need more SMP and less bigadv so I would guess one way or the other they will achieve their goal.

It really does not matter what I or anyone else thinks about it or whether we like it or not, in the end it will be what it will be. And we will adjust over time. :rolleyes:

In the end... they will find out that no one wins with this approach...

They won't get more SMP users this way... just less BigADV... :rolleyes:
 
In the end... they will find out that no one wins with this approach...

They won't get more SMP users this way... just less BigADV... :rolleyes:

This is very true. On the couple of occasions that my Intel 4P has picked up a regular SMP WU, it's done around 200k PPD. You can now do better than that with a 780 Ti or 290X, and those are far cheaper than an Intel 4P.
 
Aye. That's how you kill non-BA CPU folding, didn't we project it some time in the past? :)
 
That is true, what the move will do is make them reevaluate the smp points scheme or convert the smp over to GPU, which I have a feeling they are having some problems with since they have not done that already and they are still releasing smp projects on a regular basis and not releasing many or any GPU work.
 
Right.
Maybe next year will be time to start crunching some BOINC projects instead of wasting power on SMP folding ?
 
uh, I forget -- why did they do that anyway?
I was referring to some past discussions we had about PG making (intentionally or not)
BA less appealing (lowering rewards/increasing requirements etc.) and, at the same time
intensifying the use and promotion of GPUs.

At the time, general consensus was that it would make people shift away from SMP folding
either to GPUs or completely different DC projects.

Don't get me wrong, I'm not saying BA shouldn't accelerate. Nor I'm saying that GPUs
are given too much attention. None of that.

The point is that the upcoming changes are unlikely to have the effect PG expects :)
 
Maybe if everyone turned off all FAH for 48-72 hours call it being on strike.
They will get the hint.
My hexcore AMD-1045T was o/c to 3.6ghz and used to get 18k or so on regular smp.
Now I am lucky to get 10k ppd.
I would not mind doing smp, but why the big reduction if they need more to fold them?
 
At the time, general consensus was that it would make people shift away from SMP folding
either to GPUs or completely different DC projects.

The point is that the changes are unlikely to have the effect PG expects :)

Thanks for the history lesson tear. the GPU shift sure did occur.

and for what PG is expecting to happen, I certainly agree…

...if there's one thing I've learned from the forum here; it's that [H] team members are incredibly effective (and fast) at switching around their resources :D

regards,

Louis
 
Maybe if everyone turned off all FAH for 48-72 hours call it being on strike.
They will get the hint.
My hexcore AMD-1045T was o/c to 3.6ghz and used to get 18k or so on regular smp.
Now I am lucky to get 10k ppd.
I would not mind doing smp, but why the big reduction if they need more to fold them?

Our team has over 500 members. Our top 100-150 is fairly active but I don't think we are quite agile enough to pull something like that off without weeks of communication ahead of time.

We should let PG make his reply on deadlines changes if any.

Don't be Hasty ...
 
I switched off all my standard smp clients because they just weren't worth the power they used. Basic i5 desktops were getting like 15k ppd for me. Using over 100w of power. My dual xeon machine might use ~300watts but folds 500k ppd (3 times the power for 33 times the points) If they want more smp, make it worth the electricity I'm using. Don't just reduce the number of bigadv capable machines.

If they all of a sudden made my dual xeon worth 100k ppd. Guess what, I'd stop folding on it!
 
I switched off all my standard smp clients because they just weren't worth the power they used.
I only fold with my dedicated AMD 4P systems. Although my multipurpose box has a nice i7, SMP-capable processor, it's just not worth the power bill to keep that machine running 24/7. I may be fortunate to have a small garden 48-core crunchers, but I do have a limit over ratio of power consumption versus production. It may sound elitist, but everyone must draw the line somewhere.
 
You guys should have your DAP representative bring this up in the DAB forum.
 
DAB is dead as far as I know, and it didn't seem to accomplish anything when it was alive.
 
Wasn't Kendrak [H]'s rep? If so, when was the last meet?
 
Better question - when was the last time you saw Kendrak post in the DC section?
 
Kendrak is AWOL and..
DAB has been useless since its inception, thanks to bruce & company.

I'll be hitting people responsible directly (in a day or two).
 
Back
Top