Different leagues for F@H?

Associate
Joined
29 May 2005
Posts
586
Location
Lazenby near Redcar
What does Team 10 think of having different leagues for F@H?
Each league would be for each different platform:
SMP
GPU
PS3
Standard
There would also be a total league which is basically the present league.
I think if it wasnt too difficult for Stanford and the 3rd party stats sites to set up it would be a good idea. Basically it would stop any cross-platform squabbling about the credit value of WUs. It would be easier for Stanford to benchmark the WUs and it keeps the leagues competitive nature.
What do you think?

sculptor.:)
 
It would be nice for some categorisation, but then after all we are doing this for the science, not the points.

The lead would have to come from Stanford. They are the ones who collect the stats and third party sites can only analise what they are given.
 
I know the science is the reason we choose to fold and we all want Stanford to succeed. But the credit system has always caused the occasional problem, just trying to think of a fair way for it to work.

Hope youre sleeping ok now mate, Emily has my 260GTX so there is no point in buying Blackshark for a while, havent played a good flight sim for years.

sculptor.:)
 
Lets face it, if takes any meaningful sort of effort for no advantage to the science, it ain't gonna happen. This really has no advantage for Stanford.
 
Basically it would stop any cross-platform squabbling about the credit value of WUs. It would be easier for Stanford to benchmark the WUs and it keeps the leagues competitive nature.
What do you think?

sculptor.:)

I think that. Its up to Stanford to work out the credit value of WU's. I don't think it would make it any easier.

At the moment I think that the hardware that does more work or science gets more points as per Stanfords allocation of points. I don't see how that should change. If you want to get more points then get equipment that does more work :)
 
I do have equipment which would score more points. If I wanted more points I would run F@H on that equipment.
This was just a thought of how to stop cross platform bickering about the credit values of WUs.

sculptor.:)
 
I don't think it's a stupid idea at all - isn't it what the passkey was kind of for...

Anyway the league tables are purely stats number magic from a simple
ID:Team No:Total - issued every 3hr (or does the data file have more to it that that)
 
so for this idea would each team have 4 different stats, or would a team be restricted to a single platform.

its a good idea in theory but i don't think they could come up with anything that wouldn't cause more complaints
 
My thoughts were if there were different stats tables for each client then each person depending on the hardware they were running would rise or fall in the tables accordingly.

The SMP client running on Linux whether through native install or VMs in Windows takes a lot of setting up and can be a pain to sort out if things go wrong. Now if you were in the SMP table you are putting in the same effort as others in that table, so the credit values of the WUs are fair.

To run the GPU client ranges from being simple to set-up to complicated depending on the ammount of GPUs and choice of OS. In its simplest form one GPU in Windows it is very easy infact ATI include the client in their driver suite. However the credit values for WU are fair accross the board within the GPU table.

The same would go for each of the other clients.

I feel it would cause less cross platform bickering because if a platform went through a period of low value WUs it would effect everyone in that platform so there wouldnt be the sudden advantage to using another platform.
However the whole thing may be too complicated to set-up and run, and it would only appease a minority of people.

sculptor.:)
 
it might adversly affect the science though, if someone has a PS3 they might add it to F@H if they also had a decent cpu and a gpu or 2, with different leagues someone might not bother with a PS3 and focus on GPUs so the project as a whole would lose out
 
It might be practical to implement this solely within ocuk instead of aiming globally. The single cruncher league has colour coding depending on what they're folding on.

Anything that leads to more folding is to be encouraged I think
 
I quite like the idea of the separate and global stats tables, although im still not sure it would fix the points bickering.

Im not sure the points values really are that far out of line. For example, if you were to run a single GTX280/285, you could realistically expect between 7-9000 ppd from the card for an investment of £250 (for example). Similarly, if you spent that £250 on a core i7 processor and overclocked and run the linux smp client, i think you could realistically hit around 8,000 or more. In this respect, i think hardware of a similar age is actually quite comparable. Obviously, a core i7 is more expensive, as you need the motherboard/memory etc, but im not really sure stanford should have to take the 'costs' of folding into account [otherwise, (arbitrary arguement here!) is it not equally fair to award people in England more points than those in the USA because power costs a lot more per unit?]. From a pure hardware perspective, they earn approximately the same number of points. Obviously it is much easier to add extra gfx cards, and i dont think there is really an easy way to get around this.

Until gfx cards move towards a MIMD architecture (or is it SIMD... err...) and can process data in the same way as a CPU, it would be very difficult to design a becnmarking system that would allow for a fair comparison between the two technolgies. I guess its similar to the time predating the GPU client. This was before my time, but when quads were introduced, weren't people able to get loads of points really quickly, leaving others with duals/single cores in the dust? I think everyone has to accept to a certain extent that there will always be people who will spend more money/have more kit available etc, and im sure the whole thing comes around in cycles.

I also think that having high points for the gfx card clients is quite a good idea for the project as a whole. I would imagine if you were to take a sample of people on this forum, quite a lot of people would be using dual cores with a pretty decent gfx card, rather than a quad. Whilst it would be great to get everyone folding on their dualcore, it would be almost impossible for the person to compete in a global way with others with i7s etc, and so probably wouldnt bother (assuming all anyone cares about is the points here). However, of they can run a mid level gfx card and get pretty good points, it helps bring people to the project who otherwise would not have bothered (who would hopefully bring their dualcore along to the party as well!).

Not to knock the idea at all - genuinely, i would like to have a better breakdown of my points/scores, but im not sure it would really stop the arguements. I hope these points make sense... its quite a long post and its a Friday... :p
 
Stanford's priority needs to be sorting their CPU clients out. The perpetual beta status of SMP and the crashes that come with it are annoying people. The standard client isn't an alternative for anyone who cares about points because the output sucks. Anyone who's building a new Folding rig is gonna cram it with as many GPUs as possible because it's no hassle and it's loads of points. But GPUs can't do all the types of work, so I can see the project going down the pan if SMP doesn't get more sensible soon. It's been out more than two years, and people are still expected to manually replace client binaries and troubleshoot all manner of crashes. The 'but it's beta' argument is all very well, but people do this for points, and the only way to get decent CPU points involves hassle.

Same problem as most DC projects. Scientific geniuses, not a clue about public relations. They need to remember that the contributors with the greatest resources are the same people who don't have time to run around babysitting them.
 
Last edited:
Mattus, you seem like a decent bloke to me, you fold to help Stanford as all of us do. The points are what makes the forums tick as far as I can see. Everyone posts about ppd, hardware, operating systems and different ways to get the most out of our hardware as well as Parping (on this forum atleast).
Anyone of us that have run the SMP client know just how much time and effort it takes. Up untill the Nvidia GPU WUs started thought we were getting bonus points because of this effort as well as the extra work being done.
However the GPU client has shown that this isnt the case. Points are based purely on performance any extra effort you make is worth nothing.
I think we are seen purely as a cog in their super computer (whether that is right or wrong is debateable) and not as caring human beings.
Ive given up on the SMP client, time, costs and my own future plans had dictated this a few moths ago. I can see that others will do the same and I do wonder if this is what Stanford wants.

sculptor.:(
 
Anyone of us that have run the SMP client know just how much time and effort it takes.
Err, no, it needs very little time and effort here. The only thing I can't get it to do reliably is to run multiple native clients under Windows. The single clients are no trouble at all these days, even co-existing with two GPU clients. Yes, of course I'd like to see it finished and properly released but I find it perfectly usable as it is.
 
Back
Top Bottom