Definition of WU

B

berky

2[H]4U
Joined
Aug 28, 2001
Messages
2,233
i know it's a work unit, but how is it defined? I've had a single WU be the work of 500 somethings, and i've also had other single WU's be the work of 250 somethings. shouldn't those 'somethings' be the WU's? i've completed like 4 or 5 WU's, yet have scores close to others with like over 100 (iirc). what gives? maybe its just me, but it just seems like a dumb way to do it.
 
Every WU is made up to find the data that Stanford needs. They pick a convenient spot to send it back and get a new one.

These points are often in different spots. Each WU is then assigned a point value using a formula. The timing is done using a standard C500 that Stanford has. This makes things fair.

Currently many of the WU's are large, however this was not always the case.

Hope this helps.
JTV
 
Originally posted by James_TheVirus
Every WU is made up to find the data that Stanford needs. They pick a convenient spot to send it back and get a new one.

These points are often in different spots. Each WU is then assigned a point value using a formula. The timing is done using a standard C500 that Stanford has. This makes things fair.

Currently many of the WU's are large, however this was not always the case.

Hope this helps.
JTV
Click to expand...
Yeah, take me back to the days of the small WU anytime.:D
 
You mean you don't like these forever chewy 70.9 point "work units" that take only 35 to 40 hours or so? ;)

Some of these make me really wonder if Stanford really does to testing on a 500 meg celery:rolleyes:

Fold on

BillR
 
What your probably seeing is a symptom of how when Gromacs core went into full use Stanford started assigning really large point values to the units compared to their previous point system. So people who had already been working on the project for years and completed lots of units are now being passed in the rankings by others who are relatively new simply because they are focusing the ranking on point scores vs the number of actual units completed. We used to be doing REALLY good to have work units worth 9 points, although most got you about 5 points. Now they are worth a lot more as others have pointed out and you have observed causing the skew in the standings. Some users have completed two or three times the number of proteins of others who may be several thousand points ahead of them. I don't really agree with it, but that's just the way Stanfod has let things develop. They'll probably have to wait until they start a another version of the project before altering it again because of how lopsided things are now and will continue to progress. So maybe in Folding 4 :)
 
Definition of a WU -
To the Google masses - a protein which returns credit and helps science.
To the addict - A complex interaction of Stan4rd and my computer, yielding credit which should represent the value to science.

A large part of the problem is that Genome vs Tinker vs Gromacs is platform and OS dependent, PLUS the natural variation within proteins.

IF Stanford actually does some serious testing to baseline and assign value and time to the WU I would be amazed.
IF they could determine how many cycles at what speed each unit consumed it might give a "kinda fair" point system.

Remember just before the "Genome Wars" when many peeps here, especially relic, Gaiden, and PageFault, put a lot of effort into coming up with a "fair" point value ?

I do not understand why Stan4rd is "testing" so many darn proteins. I would expect the science to fully use a few, get them right, and go for another few.

We have to remember that numbers tell the truth.
How many active users are there ?
How many active users give an excrement ?

A WU is What U want.

This paradigm of DC was not designed for the addict that goes to work an hour early to seed, and an hour late to harvest, lugging her gleanings home on a fully laden USB stick.
 
Originally posted by CIWS
They'll probably have to wait until they start a another version of the project before altering it again because of how lopsided things are now and will continue to progress. So maybe in Folding 4 :)
Click to expand...

What exactly do you mean by Folding 4? they have version 4 of their program out now. or does that refer to something completely different?
 
Originally posted by berky
What exactly do you mean by Folding 4? they have version 4 of their program out now. or does that refer to something completely different?
Click to expand...
Newest beta
 
Originally posted by berky
What exactly do you mean by Folding 4? they have version 4 of their program out now. or does that refer to something completely different?
Click to expand...

There was Folding 1 - Then they started completely anew with Folding 2, then with Folding 3 we have the Gromacs. Even though there is a ver 4 client out now nothing really new or different has come about with it. At whatever point they carry the project to the next step or introduce new ways of working on the Proteins where a change in the point system would take place is what I referenced as Folding 4. :)
 
CIWS:

At whatever point they carry the project to the next step or introduce new ways of working on the Proteins where a change in the point system would take place is what I referenced as Folding 4.
Click to expand...

I believe we are seeing those changes take place now. The point system sill makes no sense. I'll foldem, that part is ok, but I would think breaking the actualy Work Units into smaller parts would yield them (stanford) more information faster, espcially those folding on stock home or office boxen. I only mention this as I see my Work Unit count go down by the day although my farm remains the same. No parity in the calculations so to speak.

BillR
 
For unit count priority vs point value Stanford would first have to start new like they did with Folding 2 so everyone was at parity. Establish a new "test box" to rate units with. A 500mhz Celeron is starting to show it's age quite honestly. Take the test box and have it run their smallest / quickest protein (used to be a hairpin) and measure the amount of time it takes the test box to complete it. Let's say 5 hours. That becomes the base unit for reference and will be equal to 1 point. Now if they throw their current largest unit on the test box and it takes 30 hours to complete, then the user would only get 6 points for that one unit, not freakin 70+ points like now. Then Stanford would have the system monitor unit counts not point values for determining rank. Finally it would be up to them to also control which proteins they release for processing and not throw out this mis-mosh of various sizes on multiple servers. Decide upon about three-five proteins to look at that are of the same appox size and only release those at one time. That way some users don't get stuck with units that take 30 hours to do while others get a bunch of hairpins that are finished in only 5 hours.
 
Originally posted by AtomicMoose
Newest beta
Click to expand...
Moose - you've been out in the cold to long :)
v4 is the latest full official release.
 
14 months ago when I started Tinkers on my systems where worth around 0.7 points per hour per Ghz.
Today Tinkers and unboosted Gromacs are still worth around 0.7 points per hour per Ghz.
With SSE boosting Gromacs now worth aound 1.5 +/-0.1 points per hour per Ghz.

So the basic points system has not changed.
Yes the work units have got bigger but cpu have got faster.
What has changed is that people are running a lot more Ghz.

Luck.........:D
 
You must log in or register to reply here.
Back
Top