Folding Forum

I understand exactly what you're talking about, but folks are saying that FAH has changed, and that's not true. Hardware has changed; the value of the $ has changed; people's perceptions have changed, but one point is still worth one point according to the most recent definition that I can find.

bruce wrote:I understand exactly what you're talking about, but folks are saying that FAH has changed, and that's not true. Hardware has changed; the value of the $ has changed; people's perceptions have changed, but one point is still worth one point according to the most recent definition that I can find.

Agreed,
The thing that threw "the spanner in the works" were bonus's for steep hardware requirements, fast returns etc,
if that is what makes these new Bigadv Wu's "unfair" there were a lot more "elegant" solutions that were discussed in those 18 pages,
instead it was an "across the board" change to all Bigadv Wu's,
As I've stated earlier, this should have been foreseen long before they were released from closed beta.

HaloJones wrote:I'll wait to see what difference this makes before making sweeping statements.

I can give you mine as an example of what to expect:
1 x SR-2
3 x X58 i7 970
All running under Win7/64
previous 24 hr average production: 269,000 ppd
projected 24hr average production after downloading updated values into HFM: 215,000 ppd
In other words it's wiped the output from 1 970 rig,
And that's still with a huge power bill.
Thanks guys.

Firstly I should say I understand why Stanford may seek to change the current points structure. The bonus for bigadv work units has created too much demand for this work type and is therefore reducing demand for other projects. In addition, there has been unease from a number of participants that the high rewards undermine their current and past contributions.

Having said that I’m dreadfully disappointed in the decision and the way it has been handled. It was only a month or so ago that the uber bigadv wus were released with very large bonuses for quick completion. I along with many others took this as confirmation that quick completion of bigadv units was extremely important for the science, at least in the short to medium term. I’m sure many investment decisions were made based on this very recent affirmation by the Project. After all, although we are advised against doing this we’ve got to buy hardware based on something, haven’t we?

Then there is the matter of the discussion and consultation. The thread discussing bigadv points and then the poll on whether the QRB should be changed did receive an awful lot of input. Many of us commented on the voting thread so as to avoid the drama and less than friendly posts from a certain individual within the debate thread. The poll posts were later deleted although the poll did come down in favour of no change by a majority of 3 to 1. How this discussion then leads Stanford to believe that a substantial change is warranted defeats me.

Moving beyond the decision to make a change and onto the way it’s been handled. After the official post announcing a review I honestly expected some sort of a change but not a 20% ppd reduction overnight. I really thought there would first be some proposals put forward for debate. Something along the lines of:

An increase in reward for projects where supply is high but demand is low e.g. the introduction of a QRB bonus on GPU projects perhaps. As tear stated on the beta thread, it is probably more palatable to see a colleague get a pay rise than to receive a pay cut.

or

A phased approach that would see the bigadv premium reduced from 50% to 40% followed by a review some time later to see how this has impacted demand and a warning that a further reduction may follow if demand still outstrips supply.

or

A cap at the top end eg getting a bigadv work unit back within 30 mins rather than 1 hour may not make a lot of difference to the science and therefore may not need further incentive.

The impact of all this has led me lose confidence that anybody really knows what is most important for the science and that the goal posts won’t again be moved substantially. I’ve already moved a couple of clients over from bigadv to smp (so the whole demand/supply thing is working already) as there’s less risk in the event of a wu failure and less time needs to elapse to close a client after a oneunit flag. In addition any further investment has been put on hold until there is a greater sense of direction to the steering of the ship.

I don’t expect any change of heart by anyone and have enough belief in cause to carry on folding but I’m afraid an awful lot of the passion and fun, and therefore money, associated with it has evaporated for the time being. End of bleat.

What this change does in my case is it lets me run my computer less. If I have no reason to leave my computer running for 45 hours strait then ultimately less work is being done. I think this change will have greater ramifications for productivity than they are thinking. If Pande wants me to run less units, I'm fine with that.

When a reply like that comes from phoenicis, a "top 20" Folder it becomes even more unclear where the whole
"extensive discussion" process took place ?

HaloJones wrote:I followed the whole of this 18-page discussion as it happened but felt no particular desire to comment. Does the silence mean I agreed with the need to change or that I didn't understand the need for the discussion in the first place? In fact it was the latter. If someone builds a machine that can get 1000000 ppd, why should I feel angry? Jealous, of course but angry?

I don't see why this needs to change. If folders with single-core machines are feeling devalued, why? Their contribution has never changed. They're doing the same science and getting the same points. If that science has become less useful because all the kewl stuff is being done with bigadv, they should be grateful the science they do hasn't been terminated.

I couldn't agree more,
The project is Folding At Home, but how many true "home computer's" using "spare resources" can return a Bigadv Wu in a day ?
Does having a few with the resources who can return these Bigadv Wu's quickly in some way devalue other contributions or enhance the project as a whole,
making every contribution more valuable to the science,
How about we totally ditch the Bigadv project, then when for whatever reason something is needed very quickly we can all throw Stanford
the dollars needed to hire time on a supercomputer,
Say a hundered bucks a week each, ok, who's first ?

Didn't think so,
Just whine about those already doing that and more.

Jester wrote:When a reply like that comes from phoenicis, a "top 20" Folder it becomes even more unclear where the whole
"extensive discussion" process took place ?

That's the point, it wasn't.

They should have left 6903 and 04 in Beta for long enough for the beta team to flag the points as a potential issue and adjust before being released to the whole project. This is part of the beta team brief and would have caused most of the current venting/frustration etc to not happen.

Anothr suggestion would have been to leave them as is and amend future projects in order to fix the problem

They didn't and now we have uproar from the uber folders - understandable given the rigs that they are currently running

It might also be an idea for Stanford to get a bigger, more powerful -bigadv machine, one that has twin 12core cpu's might be a good starting point

As judged from the high demand for bigadv work units, this has been very much a success, perhaps a little too much so. We would like to continue to offer a bonus for bigadv to offset the above factors, but we don't want demand for bigadv to overwhelm the rest of the project or imbalance the points system.

Here is the most telling part of the post.

The question remains why?

Why would you not feed the demand, provide bigadv WU to anyone who wanted them.

It is not a zero sum game.

Even if the demand does "overwhelm" that is how efficiency works.

The jet age "overwhelmed" demand for transatlantic ocean liners.

The speed and efficiency and low cost made the service obsolete in a very short number of years.

Should jet ticket prices been taxed at 10X rate to keep the transatlantic ocean liner business afloat?

Of course not.

You service demand, you CAPITALIZE on it to grow.

What kind of logic tells someone they should discourage exponential growth in processing power because it makes old antiquated systems look weak (only in comparison to the new)

This is exactly the complete opposite of what is needed to advance this project.

One can only surmise that real growth and advancement is not the primary objective.

Nathan_P wrote:

Jester wrote:When a reply like that comes from phoenicis, a "top 20" Folder it becomes even more unclear where the whole
"extensive discussion" process took place ?

That's the point, it wasn't.

They should have left 6903 and 04 in Beta for long enough for the beta team to flag the points as a potential issue and adjust before being released to the whole project. This is part of the beta team brief and would have caused most of the current venting/frustration etc to not happen.

Anothr suggestion would have been to leave them as is and amend future projects in order to fix the problem

They didn't and now we have uproar from the uber folders - understandable given the rigs that they are currently running

It might also be an idea for Stanford to get a bigger, more powerful -bigadv machine, one that has twin 12core cpu's might be a good starting point

Looking in the beta team forum it seems it caught one or two there by surprise too....

You can ship freight by air, by rail, by truck, by ship, etc. It's about having a balanced approach. Shipping iron-ore by plane makes no sense That doesn't mean air-freight is a bad idea but it also doesn't mean that it doesn't need to be moved from one place to another.

Some projects need really powerful hardware and really fast returns. The results from GPUs and from slower/smaller CPUs are important, too, but not for the same proteins. The Pande Group has always maintained that all contributions are important. You're forgetting that Kasson said

kasson wrote:We very much appreciate the donors who have volunteered to run bigadv work units; these projects add substantially to our scientific capabilities. We do important science with all classes of work units, however, and we want the points system to reflect that. Based on extensive feedback, we are considering renormalizing other parts of the system but have not finalized decisions in that regard.

kasson wrote:. . . We would like to continue to offer a bonus for bigadv to offset the above factors, but we don't want demand for bigadv to overwhelm the rest of the project or imbalance the points system. . . .

and that in his original announcements of the bonus plan, that it was an experimental program that might be adjusted later.

Nobody likes their contribution to earn less that it once did, but considering all of the factors, that seems to be the best way to put the system back in balance.

Shipping iron-ore by plane makes no sense That doesn't mean air-freight is a bad idea but it also doesn't mean that it doesn't need to be moved from one place to another.

Exactly

However, you allow the two form to co-exist and seek their own economic scale and niche.

You don't "renormalize" one system in regard to another.

I wish I could get ocean rates for my over the road shipments. I wish I could get full truckload rates for my LTL. I wish my rush overnight air rates were the same as my LTL rates.

If I asked my freight expediter to "normalize" all of these rates...I wouldn't even hear laughter, just the click of the phone...

mdk777 wrote:

Shipping iron-ore by plane makes no sense That doesn't mean air-freight is a bad idea but it also doesn't mean that it doesn't need to be moved from one place to another.

Exactly

However, you allow the two form to co-exist and seek their own economic scale and niche.

You don't "renormalize" one system in regard to another.

I wish I could get ocean rates for my over the road shipments. I wish I could get full truckload rates for my LTL. I wish my rush overnight air rates were the same as my LTL rates.

If I asked my freight expediter to "normalize" all of these rates...I wouldn't even hear laughter, just the click of the phone...

and why would that be ?
because they want to stay in business ?
because they are actually accountable to someone ?

I just wanted to thank the whole F@H team. It proved that one can talk to pandegroup and they are listening to their community. Even though it was a only a small change, I hope it can motivate people to put their GPUs and their smaller CPUs back to work and help process science. Meanwhile I will put my sandy back to work on some bigadvs.

Thanks guys.

bruce wrote:You can ship freight by air, by rail, by truck, by ship, etc. It's about having a balanced approach. Shipping iron-ore by plane makes no sense That doesn't mean air-freight is a bad idea but it also doesn't mean that it doesn't need to be moved from one place to another.

Some projects need really powerful hardware and really fast returns. The results from GPUs and from slower/smaller CPUs are important, too, but not for the same proteins. The Pande Group has always maintained that all contributions are important. You're forgetting that Kasson said

kasson wrote:We very much appreciate the donors who have volunteered to run bigadv work units; these projects add substantially to our scientific capabilities. We do important science with all classes of work units, however, and we want the points system to reflect that. Based on extensive feedback, we are considering renormalizing other parts of the system but have not finalized decisions in that regard.

kasson wrote:. . . We would like to continue to offer a bonus for bigadv to offset the above factors, but we don't want demand for bigadv to overwhelm the rest of the project or imbalance the points system. . . .

and that in his original announcements of the bonus plan, that it was an experimental program that might be adjusted later.

Nobody likes their contribution to earn less that it once did, but considering all of the factors, that seems to be the best way to put the system back in balance.

It probably would have gone down better if the new points structure had been put in place for new projects ratehr than modifying existing ones. At the moment there are some mightly upset -bigadv folders who may end up pulling their hardware