Things that make you go Hmmmm????

Sunin

[H]ard|DCer of the Month - August 2008
Joined
Dec 27, 2005
Messages
3,421
http://folding.typepad.com/

Per the new articles it mentions that the quality of the science output is being examined... to see if it is good.

I had to sit there and think for a bit on this and then started getting rather angry!

Ok so F@H decides to release a GPU client for Nvidia... which is producing MASSIVE Points 4-6k PPD +++ and they don't even know if it is producing good science?

Two things concern me here...

1. Are the points overly inflated to get people to hard core beta test it and thus the future PPD are going to be reduced?

2. Why is F@H allowing such points for something that may not be producing good science...


I'm really scratching my head here... Are they just moron's at wording or is it really this out of hand?



 
I think you'll find that the results from from the original GPU client while good as a proof of concept where not that accurate in the science department.
I think it's one of the reasons it got dropped so quick once the GPU2 client came out.

I dont think the GPU/SMP points will be reduced but the single clients points may well be raised.
If you remember that far back, thats what Stanford did when they swopped from mainly Tinkers to Gromacs as the benchmark protien.
All protiens got a 2x PpD bonus so that Intels crunched Gromacs at the same PpDpGhz, which AMD's used to crunch Tinkers.
People object when the top PpD goes down but are happy when the lowest PpD goes up.

Luck ............ :D
 
Two things concern me here...

1. Are the points overly inflated to get people to hard core beta test it and thus the future PPD are going to be reduced?
It's possible. Look what happened to SMP with P2665s and apparently that trend is going to continue according to Stanford.

2. Why is F@H allowing such points for something that may not be producing good science...
Once more, I think that's not unprecedented. When F@H started it was intended as a trial run. Stanford relaunched the project after the initial project ended back in the early days. People were wondering pretty much the same thing concerning the quality of the science produced with the inaugural clients. Stanford replied that the results produced during the first project was used to establish the second run, which was to be permanent. This isn't even the first F@H DC project we're working on.

I'm really scratching my head here... Are they just moron's at wording or is it really this out of hand?
The longer you become involved with the F@H project, you soon realize the more things change, the more things stay the same. This isn't an exception but seems rather the rule.

I dont think the GPU/SMP points will be reduced but the single clients points may well be raised. If you remember that far back, thats what Stanford did when they swopped from mainly Tinkers to Gromacs as the benchmark protien.
I'd gladly see the standard client's WUs values increase. They are undervalued IMO, and I've been stating that for quite a while.

All protiens got a 2x PpD bonus so that Intels crunched Gromacs at the same PpDpGhz, which AMD's used to crunch Tinkers. People object when the top PpD goes down but are happy when the lowest PpD goes up.
I think the top WUs are overvalued and way out of balance with the rest of the DC project. Stanford and anyone else can proselytize the virtues of the points allocation system and how it yields greater science for the project, but to me it seems to promote elitism as much as anything else. Whether that's better for the science in the end, I can't say, but once you're bitten by the vagaries of the F@H PTB, you become twice shy about hardware reinvestment and the total commitment it demands.

 
This just reinforces the fast that you don't buy hardware that you can't use some place else for folding.

and yes the dog needed a computer :p

 
Unlike all of you, I understand Vijay perfectly so what he really meant is that since the GPU2 client is new and the NVIDIA core is new, they need to first validate if the results generated by the client is valid. In other words, they are using the results of a previous project which is already crunched and validated to rerun with the new client to make sure the end result is the same. This is a normal beta testing procedure before it's put in official production but the points value is still there no matter how good is the results. I'm actually surprised to see that they wanted to put the GPU2 out of beta to be in the same rank as the console client or ps3 client very soon but I guess the ease of use is good, which is a criteria (SMP is rather difficult to maintain due to the stupid MPI requirements).

I don't think they will ever slash points anytime soon. As soon as a benchmark value is set, it will stay like that unless it's proven it introduce too much inconsistencies. If you are worried because the ATI client is not producing as much as the NVIDIA client, it's just because there is tweaks to do and ATI is crippled (Mike Houston think it's not even utilizing all the shaders available while NVIDIA already do it efficienly, hence the production difference for cards which should be similar in gaming). There is a possibility that the NVIDIA production will be reduced by 5-10% as they tweak for stability and to make computer use smooth (watching video, gaming, flash running, etc...).

I hope I vulgarized what Vijay said correctly and I'll say to not worry too much.

 
http://folding.typepad.com/

Per the new articles it mentions that the quality of the science output is being examined... to see if it is good.

I had to sit there and think for a bit on this and then started getting rather angry!

Ok so F@H decides to release a GPU client for Nvidia... which is producing MASSIVE Points 4-6k PPD +++ and they don't even know if it is producing good science?

Two things concern me here...

1. Are the points overly inflated to get people to hard core beta test it and thus the future PPD are going to be reduced?

2. Why is F@H allowing such points for something that may not be producing good science...


I'm really scratching my head here... Are they just moron's at wording or is it really this out of hand?




I'm glad someone else is also thinking:

http://www.hardforum.com/showpost.php?p=1032680714&postcount=16





 
Unlike all of you, I understand Vijay perfectly so what he really meant is that since the GPU2 client is new and the NVIDIA core is new, they need to first validate if the results generated by the client is valid. In other words, they are using the results of a previous project which is already crunched and validated to rerun with the new client to make sure the end result is the same. This is a normal beta testing procedure before it's put in official production but the points value is still there no matter how good is the results. I'm actually surprised to see that they wanted to put the GPU2 out of beta to be in the same rank as the console client or ps3 client very soon but I guess the ease of use is good, which is a criteria (SMP is rather difficult to maintain due to the stupid MPI requirements).

I don't think they will ever slash points anytime soon. As soon as a benchmark value is set, it will stay like that unless it's proven it introduce too much inconsistencies. If you are worried because the ATI client is not producing as much as the NVIDIA client, it's just because there is tweaks to do and ATI is crippled (Mike Houston think it's not even utilizing all the shaders available while NVIDIA already do it efficienly, hence the production difference for cards which should be similar in gaming). There is a possibility that the NVIDIA production will be reduced by 5-10% as they tweak for stability and to make computer use smooth (watching video, gaming, flash running, etc...).

I hope I vulgarized what Vijay said correctly and I'll say to not worry too much.


LOL had he said that I might have understood it :)
 
Back
Top