Folding Forum

Decided to play around with MS Excel on the PPD and TPF of a particular CPU_SMP work unit (6994). I also used the Folding@home Bonus Point Calculator at http://www.linuxforge.net/bonuscalc2.php. These were the graphical results.

For WU completed successfully and sent after preferred deadline but before final deadline, PPD appears to increase linearly as TPF decreases.

There is a large jump in PPD (because of bonus points) for submission at the preferred deadline boundary. After that, PPD increases not linearly but with a 'power curve' as TPF decreases.

http://s14.postimage.org/ncp6i2t1t/PPD_vs_TPF.jpg
EDIT: OOPS! The labels for before and after need to be switched. Sorry about that error.

Try this at the front end of the curve

6994 - tpf 1:43, PPD 48k
6994 - tpf 3:54, PPD 14.2k

From the graph, it looks like if person A has 10x the PPD of person B, it doesn't mean person A has done 10x as much work as person B.

You didn't take the graph very far to the left with shorter time frames.

7im wrote:You didn't take the graph very far to the left with shorter time frames.

He may not have the machine horsepower to go shorter.

Thats why i posted a couple, unfortunately i do not have anything in between

user123 wrote:From the graph, it looks like if person A has 10x the PPD of person B, it doesn't mean person A has done 10x as much work as person B.

I do not understand person A vs. person B. Your graph of PPD vs. frame times and should be person independent.

The discontinuity is what happens at the first deadline -- You no longer get any bonus points but you will get the base points till the final deadline. Thus PPD goes linear and much smaller on the right hand side of the curve.

The left side of the curve reflects the value of the bonus points. Those bonus points are designed to encourage the quick return of the WU and that portion is exponential so if your machine has enough horsepower to return a WU much quicker, you get a lot more points and PPD. PG has stated that those bonus points correspond to the value of the scientific work being done. Faster is much better ...

P5-133XL wrote:

7im wrote:You didn't take the graph very far to the left with shorter time frames.

He may not have the machine horsepower to go shorter.

It takes very little computer power to use the bonus calculator at http://www.linuxforge.net/bonuscalc2.php and then plot it in excel

For the record, the formula of the curve is y = 109590x^-1.5

I did something similar in a excel spreadsheet a few months ago for the 8101 WU. I didn't use the points calculator, just the equation in the FAQ for QRB; then you can plug in any TPF's or WU completion time. What the graph emphasizes, and some people might not realize, is that the points were not designed to be linear. The Quicker you can return the WU ( TPF ) the bigger the bonus. That's why I wonder about posts that complain about low points or high points but don't mention the TPF.

As you can see on the graph, 28 hrs WU completion and above is mostly linear. Lower than 28 hrs WU completion and there are bigger bonuses per hour reduced.

This graph is of an 8101 mapping WU completion hrs to QRB:

[Edit]
I had used an equation that works off of WU completion time.

k1wi wrote:

P5-133XL wrote:

7im wrote:You didn't take the graph very far to the left with shorter time frames.

He may not have the machine horsepower to go shorter.

It takes very little computer power to use the bonus calculator at http://www.linuxforge.net/bonuscalc2.php and then plot it in excel

For the record, the formula of the curve is y = 109590x^-1.5

So true!

hi to all,

it possible to generate a reports in to excel. .?? how??


thanks for advance.

Some time ago I did a bit of (somewhat clumsy) spreadsheet plotting myself, see QRB_proposal.xls. IIRC, I had a little misunderstanding about the QRB formula, and the spreadsheet was intended as an example on how to cap the QRB curve for extremely low TPF. Feel free to fix and use if you like.

EDIT: my original topics were
viewtopic.php?f=16&t=19062&start=255#p190178
viewtopic.php?f=16&t=18797&p=190640#p190640