PG announcement about Big Advanced Projects

[tank1023]

November 14, 2011


Planned changes to "Big Advanced" (BA) projects, effective January 16, 2012



Big Advanced (BA) is an experimental type of Folding@home WUs intended for the most powerful machines in FAH. However, as time goes on, technology advances, and the characteristics associated with the most powerful machines changes. Due to these advances in hardware capabilities, we will need to periodically change the BA minimum requirements. Thus, we are shortening the deadlines of the BA projects. As a result, assignments will have a 16 core minimum. To give donors some advance warning, we are announcing this now, but the change will take place in 2 months: no earlier than on Monday January 16, 2012.

We understand that any changes to how FAH works is a disruption for donors, and we have been trying to minimize such changes. For that reason, we are not changing the points system at this time.

However, we want to emphasize that the BA program is experimental and that donors should expect changes in the future, potentially without a lot of notice (although we will try our best to give as much notice as we can). In particular, as hardware evolves, it is expected that we will need to change the nature of the BA WUs again in the future.


Well it was fun but I'm tired of spending thousands of dollars just to have the Pande group change things around. If it is the intention of this group to lose folders then your on the right track.
How fast do you really need these Wu's to come in, cause I didn't realize the time sensitive nature of this project guess well se some huge results soon.

[Kamicrit]

So who has a 16 core cpu here?

[jimerickson]

some guys run 48 physical cores. not me i am just a smp & gpu folder on 8 cores. i am officially obsolete . no but seriously there is still valid science to be done with smp & gpu. fold on!!

[tanner2]

I hear you tank, I don't make much money so a multi thousand dollar outlay is out of the question to stay in the ba race.

[tank1023]

The thing I don't understand is why. My 980x was just fine finishing the work units in 3-5 days, so your telling me that the Pande group is utilizing this information at such a fast rate they need to change the requirements.
I would expect this type of action from a manufacturing company that want you to continue buying their product.
I gave three years 47million points and approx 38k work units to this project but Jan 16th I'm done.

I'm sorry Grandmom, I really wanted to find a cure for your Alzheimers.

[Jesse_V]

Tank, that really is a very remarkable contribution. I am completely impressed. I myself just crossed over a million points, but it really amazes me that people like you can be that productive!

The rapid completion of Work Units is of serious scientific benefit. The point simply reward users for this, and they really begin to mean something. I know how hard it was for me to contribute a million points, so it must have taken you a lot of effor to get your 47 million points. Yes, the Pande Group is utilizing the information this fast, and they wish it could go faster. Take a moment and think about it in their perspective. I can see myself launching a simulation on an Alzheimer's protein, creating hundreds of Runs and Clones, and having F@h run it. The simulation will be instrumental in disease research and protein folding theory, so I wish my results would be done tomorrow. But it takes a long time instead. So I'm trying to find a way to make it finish as quickly as quickly as possible. So I turn up the hardware requirements, and lower the deadline, effectively ensuring that my complex trajectories get completed on the fastest hardware possible.

All types of clients make very valuable contributions to the project. PS3, uniprocessor, SMP (regular and bigadv), and GPU all are very useful. It's one of the things that distinguishes F@h from other distributed computing projects. They are simply trying to set up a customized-to-the-hardware configuration to achieve overall maximum performance. I really understand where you are coming from, I really do. Points are valuable, and from the looks of it so many people pour so much time and effort into maximizing their PPD, as they should, since that generally equates to scientific production. Your 980x can still make great contributions on -smp 8, its not like that is no longer an option. Its just that the PG has some simulations that are very large and have a great priority, so they want to make sure that they are folded as fast as possible, by using the most powerful hardware F@h has.

I'm very sure that there has been a great deal of internal discussion about this, and I suspect that it may have something to do with many other folders barely completing the bigadv WUs on time, but I don't know since that wasn't mentioned in the post. Points measure science, but in my mind one the most important things is to just do the best that you can and stay folding.

[Horvat]

My first question is; Are you refering to 16 physical cores or 8 physical cores with H/T?

[ChasR]

As I understand the change, it's logical cores.

[Horvat]

As I understand the change, it's logical cores.

So I understand you correctly, logical cores as in "logical cores = physical cores + virtual cores". So a dual quad core intel server would still meet the requirements.

[Punchy]

As I understand the change, it's logical cores.

Reference, please?

[jimerickson]

wouldn't it depend on how much they tighten up the deadlines?

[Simba123]

ok, fine, I understand to a degree that Stanford wants BA results in quicker and more people folding SMP.

With this change they really, really need to up the Kfactor or maybe include an 'additional' bonus if an SMP unit is returned with more than 95% completion time (my 2600k can do that, so I reckon most of the 9xx can too).

If they do that, they will keep the competition going, and maybe not lose too many folders.
This has been very poorly implemented to say the least.
The Advance notification of the time frame for the change is nice, the lack of notification for what is/was under discussion is NOT.
They (Pande and Stanford), need to remember that without us their work would grind to a halt, and they would all be out of a job.
They need to treat us with a LOT more respect,

There are millions of donors world-wide who collectively would have donated BILLIONS in hardware and electricity to support their project over the years.
Angering (I'll put it that way, saying what many people really think of this decision would get this post deleted) people off like this is not going to do them any good.

Adding an additional bonus would not be too difficult to add to the formula, and if they get on it and at least do something to alleviate the points nerf from not being able to fold BA
before the new requirements come into effect, they may keep a few more folders.

[Xavier Zepherious]

what gets me is that a 8 core- 16 thread xeon(or dual 4) at 2.6 ghz is in the guideline
while a 6core /12 thread sandy e at 5ghz is not - the sandy-e is twice as fast as the 8 core/16 thread sr2 (or server board)
be equivalent to 12 core 24thread in the 2.6Ghz range

If you are gonna have consistancy base it on core speed as well as cores
the 970-990x and sandy-e should be left out for a bit

otherwise start telling people with SR-2s with older chips to do SMP too

what Vijay has to do... is slowly implement changes and phase out older 8 core 16 threaded servers as well as the high end enthusiast...do it gradually...slowly squeezing point system down
keep the bottom fixed

based on time to get it in for bonus(preferred).... not on cores
when people stop making deadlines - then they will either adapt/improve hardware or migrate the old to SMP

[Simba123]

what gets me is that a 8 core- 16 thread xeon(or dual 4) at 2.6 ghz is in the guideline
while a 6core /12 thread sandy e at 5ghz is not - the sandy-e is twice as fast as the 8 core/16 thread sr2 (or server board)
be equivalent to 12 core 24thread in the 2.6Ghz range

If you are gonna have consistancy base it on core speed as well as cores
the 970-990x and sandy-e should be left out for a bit

otherwise start telling people with SR-2s with older chips to do SMP too

what Vijay has to do... is slowly implement changes and phase out older 8 core 16 threaded servers as well as the high end enthusiast...do it gradually...slowly squeezing point system down
keep the bottom fixed

based on time to get it in for bonus(preferred).... not on cores
when people stop making deadlines - then they will either adapt/improve hardware or migrate the old to SMP

See, this is what I mean by poor communication. If this had been discussed more openly, then this fairly obvious fault in just saying "16 core/threads is the new minimum"
would have been pointed out.

You have pointed out the best and easiest way to implement the change.

Just change (reduce) the preferred time. Systems that can complete a BA unit in the preferred time will continue to do so -who cares what the setup is -
and those that fail to complete within the preferred time will go back to SMP.

[Jesse_V]

ok, fine, I understand to a degree that Stanford wants BA results in quicker and more people folding SMP.

With this change they really, really need to up the Kfactor or maybe include an 'additional' bonus if an SMP unit is returned with more than 95% completion time (my 2600k can do that, so I reckon most of the 9xx can too).

If they do that, they will keep the competition going, and maybe not lose too many folders.
This has been very poorly implemented to say the least.
The Advance notification of the time frame for the change is nice, the lack of notification for what is/was under discussion is NOT.
They (Pande and Stanford), need to remember that without us their work would grind to a halt, and they would all be out of a job.
They need to treat us with a LOT more respect,

There are millions of donors world-wide who collectively would have donated BILLIONS in hardware and electricity to support their project over the years.
Angering (I'll put it that way, saying what many people really think of this decision would get this post deleted) people off like this is not going to do them any good.

Adding an additional bonus would not be too difficult to add to the formula, and if they get on it and at least do something to alleviate the points nerf from not being able to fold BA
before the new requirements come into effect, they may keep a few more folders.

If you've read Dr. Pande's previous post http://folding.typepad.com/news/2011/11 ... d-dab.html it describes a Donor Advisory Board which is designed to improve communication and transparency. You'll notice that its goals are "What can PG do to help improve communication?" and "How to make PPD more consistent". Understand, it is probably very difficult for them to make a WUs have consistent PPD. Each simulation has different numbers of atoms, uses the same architecture in varying ways, and performs differently on different hardware. The Kfactor would help with this, but I believe we need to get a lot more data on this. The change is probably based on a large sample of long-term patterns donors take. It's likely many of the admins that monitor and reply in these topic were noting what they saw. It's been my experience that when donors propose a change by using data, logic, and neutral tones, their chances of making an impact are much higher. Let's discuss your ideas as such, and perhaps they'll be heard somehow. We'll see.

They are very much aware of how valuable we are. Dr. Pande has mentioned it numerous times on his blog, and its pretty much a well known fact. What would you propose that they do? How would they notify us? What fraction of those that would care or be impacted would respond? Would the decision from such a survey positively affect science? And I'm not in any way trying to downplay your importance, but aren't the bigadv people a very small fraction of F@h? It's my understanding that each scientist who launches their projects sets the k-factor, so Dr. Pande is not directly responsible there. Adding an additional bonus is very unlikely on these grounds alone, as you would have to update all the WUs and the servers, and I don't know about the WUs but I know server updates are very difficult and cause downtime and other issues, which negatively affects science. There's a lot of factors here that come into play for any change.