Folding Forum

I am not interested in points or credit, just wanted to help out a little. I have run each client a little, I change from time to time just for fun.
Does the folding team publish any information that says where they need help, a particular client or project that they need more processing power or is running behind schedule.

Sorry, no.

While you don't care about the points, they have said that PPD (points per day) is designed to match their science needs. So clients that get the most PPD are generally the ones that have the most scientific value to them. Perhaps you can use that information to help you decide which client to run. Do note that regardless, any help is appreciated.

Also note that they assign the work units in the order that they need to be processed to best help the project move forward, so you don't have to worry about trying to pick the best work units. Run a GPU or SMP client (or both) for the most points. After that, all donations are welcomed.

AH, so they have already done this in there point system.
So, if they needed people to run the standard client they would up the points so people would migrate to that.
Currently they would like people to run the SMP version since it gives the highest PPD.
Did I get that correct ?

The -bigadv units give a ridiculous amount of points. That's where they are trying to encourage the points hounds to go. Takes some beefy dedicated hardware to run them though and return them under the deadlines. The other half of it is the bonus system. The faster you can return a -smp or -bigadv unit, the more bonus points you get.

The points serve the needs of the PG. It encourages donors to get a unit, crunch it as fast as possible, and return it in as little time as possible. Speed over quantity.

[email protected] wrote:AH, so they have already done this in there point system.
So, if they needed people to run the standard client they would up the points so people would migrate to that.
Currently they would like people to run the SMP version since it gives the highest PPD.
Did I get that correct ?

That would be the theory. As to SMP vs. other clients it would all depend on the HW, and how much processor (GPU/CPU) time allocated. The SMP client PPD is highly dependent upon how fast the WU's back to them. So, if you have lower-end HW or not a full time folder the SMP client may not actually get the highest PPD for your system. For many, it is actually better to run multiple uniprocessor clients (1 per core) than a single SMP client.

Generally, if you have Nvidia GPU's you can run those concurrently with a SMP client to maximize PPD, but that may not be true for ATI video cards for they generally use more CPU resources which damages your SMP productivity. Obviously adding a PS3 to the mix is totally independent to your CPU/GPU folding choice.

Sorry to add confusion to your choice, but you can't just make a blanket statement about which client is best. It all depends on what you have hardware wise and how much you are willing to donate.

Do note that while the PPD for the Classic Client is significantly less* than that of SMP2, it still does some very important scientific work that even the SMP2/GPU can't do. If you are interested in the type of F@h Clients available and their "target audience" I would suggest that you read this -> viewtopic.php?f=61&t=16207

BTW, welcome to the F@H Forum [email protected]

*There are plans to introduce a Bonus point system to the Classic Client but there isn't any ETA.

PantherX wrote:Do note that while the PPD for the Classic Client is significantly less* than that of SMP2, it still does some very important scientific work that even the SMP2/GPU can't do. If you are interested in the type of F@h Clients available and their "target audience" I would suggest that you read this -> viewtopic.php?f=61&t=16207

BTW, welcome to the F@H Forum [email protected]

*There are plans to introduce a Bonus point system to the Classic Client but there isn't any ETA.

I will second this suggestion of running a few classic clients, they are needed and with people going for the most paying clients (GPU and SMP), classic work need to be done by someone. Since you don't care about points, that's the best way to help.

The disparity in points between SMP and the Classic CPU client is also related to the speed vs. quantity issue that was mentioned earlier.

Let's say you have a Quad. The simplest choice that you have is between four copies of the Classic client or one copy of the SMP client running -smp 4. Although Stanford has to build the WUs for those configureations slightly differently, from their perspecive, you're making a choice between running four WUs at the speed of one core vs. running one WU at the speed of 4 cores. There's an inherent speed bonus built into the structure of FAH. If they award X PPD to each classic client for a total of 4*X, then the PPD for SMP should be greater than 4*X because you're returning the work on a tighter schedule -- and that was true before the advent of the bonus.

Similar logic can be applied to the difference between WUs that are assigned with the -smp flag and WUs which are assigned with the -bigadv flag.

Within either the -smp or the -bigadv category the bonus distinguishes between WUs that are returned slowly or quickly.