Forever tight spot

[Hua Luo Han]

Why phase 2 seems worst than phase 1

lot of time wasted on Tight Spot

[pointwood]

This new client generates much better structures - just check the current Top10 of best structures - we've already gotten pretty low

Besides that, it's equal to everyone and the stats have been zeroed, so no one gets an advantage or disadvantage

[djp]

Science
With my limited understanding of the science of the DF project, I'd say that time spent in a tight spot is not really time wasted. My technical knowledge is limited, so I'll try to use layman's English. I probably also have the order of magnitude wrong, but here's my understanding:

Out of the billions and billlions of impossible protein configurations, your computer is working to calculate the mere millions of configurations that are possible. Every step of the fold involves hundreds of calculations to see whether or not your computer is working on something possible. All that time ruling-out the impossible seems like a waste, but the current (Phase II) algorithm is much better at cutting great chunks out of the huge universe of impossible protein configurations. Compared to Phase I, this Phase II project returns a much better quality of data. It also returns a lesser quantity of data. This means that the servers in Toronto will not process as much "noise" from our work in order to find the pure "signal" of valuable information.

It looks like your computer is contributing less information to the project, but this is a good thing. The data you contribute in this phase are more valuable. Take a look at the RMS values. I threw quite a few GHz of Intel's best processors at the project and after the first week I never found a configuration with a better RMS value than 12.65, despite having contributed nearly 200 million structures. After only a few hours of crunching with slow machines, I've contributed 8,000 structures and my lowest RMS value is already down to 9.88 A.

Teamwork
If you're concerned about statistics, keep in mind that all of your DF team-mates or competitors are working with the same improved client that you are using. We can't compare fairly the count of structures returned from Phase I with the count in Phase II, but we're all on equal footing in Phase II, just as we were in Phase I. If you want to have better statistics, throw more high-end computing horsepower at the project.

Since "points" acumulate with the square root of the number of generations you've contributed, there's a definite stats advantage to staying with the project. As you contribute more work, a given parcel of work counts for more.

The overal purpose of the Distributed Folding project is science, and the new client generates better quality results.

[Hua Luo Han]

My rigs should be power enuff though

P4 3.0C
XEON 2.8
AXP3000
AXP2800
AXP2400
AXP1800



And thanks for the writeup ...........

[Ned]

I know its frustrating to be in a tight spot... Seems like your not getting your fair share of points....

However, my experience in the Beta shows that some of my best RSM of A values came from those tight spots....

YMMV

Ned

[FoBoT]

it is a feature, not a bug

[Grumpy]

We are searching for the Magical 1.00, the Stats are the fun we have along the way

[aptarasc]

Uhh djp, this is a different protien than the last P1 so you can not really compare the RMS differences you have seen like that.

This looks like the same protein that was used in the beta which we did get an overall better RMS than when the same protein was run on the the P1 client.

[raggie757]

if it spends a lot a time on tightspot does it mean my pc is unstable?brain tryied to explain it to me but im not to smart lol.my pc is a amd xp @200x11.5.and im just droping down the stats page

[peaz]

same here... been folding the whole day on a p4 2.4 but still only at gen #14... but best rms is already 8.2

but so little score

even my p3 is generating more scores than this p4 2.4GHz...

hmm shouldn't smaller rms factor into the score too???

[raggie757]

yeah its weird i know my pc aint slow i blow away a amd xp 3000 on sandra. super pi lowest is like 43 seconds.only thing i can think of is i only have 256 megs

[peaz]

doubt ram got anything to do with it all...

got a p3 1GHz folding and generating 3x more structures than this p4...

the only difference is the RMS.... the p4 got a low one right from the start

the p3 didn't... and was a high 13 something...

[Grumpy]

Gen #14 and a RMSD of 8.2 is good.....hopefuuly it will keep going down and will see you in the top 10 for RMSD

[Brian the Fist]

Just remember the law of averages. If yours is going slowly, surely everyone elses will too at some point or other. It will all balance out over time.

[raggie757]

rofl im folding so slow now.its like i can just as fast with pc off goes and crys in corner

[pointwood]

Maybe we can get Dyyryath to slightly modify his stats so our numbers will be "real points*100" or something