PPD Amount

[peter24601]

Not sure where to put this thread so it's here.

PPD. Just wondering why you don't' get the full PPD amount stated.
On one GPU my PPD was just over 80,000 when I checked http://kakaostats.com/ my score was just over 14,000.

What I'm I missing or have not read?

Thanks

[kiore]

Not sure where to put this thread so it's here.

PPD. Just wondering why you don't' get the full PPD amount stated.
On one GPU my PPD was just over 80,000 when I checked http://kakaostats.com/ my score was just over 14,000.

What I'm I missing or have not read?

Thanks

To get that sort of output on a single GPU you must be doing one that has a quick return bonus (QRB), a project 8900 or similar and be using a flag to request advanced or beta test units.. For quick return bonus you need to have a passkey and have completed 10 units with more than 80% success. If you do not have a qualified passkey this may be why the reported and actual PPD are different.

[GreyWhiskers]

For those who are easily amused, the following formula will let you do a back-of-the-envelope conversion of TPF in minutes (should be average over whole work unit) to ppd:

ppd = (Points * 14.4) / TPF in Minutes

If you are wondering where the 14.4 comes from, the actual formula is:

ppd = (points * 14,400) / (TPF * 100).

In simple algebra, the 100 can be removed from both the numerator and denominator.

14,400 is the number of minutes in a day (24 hours * 60 minutes in an hour), and 100 is the number of frames in a work unit.

[folding_hoomer]

For those who are easily amused, the following formula will let you do a back-of-the-envelope conversion of TPF in minutes (should be average over whole work unit) to ppd:

ppd = (Points * 14.4) / TPF in Minutes

If you are wondering where the 14.4 comes from, the actual formula is:

ppd = (points * 14,400) / (TPF * 100).

In simple algebra, the 100 can be removed from both the numerator and denominator.

14,400 is the number of minutes in a day (24 hours * 60 minutes in an hour), and 100 is the number of frames in a work unit.

I think both times you mean 1440 (minutes)

[Napoleon]

For those who are easily amused, the following formula will let you do a back-of-the-envelope conversion of TPF in minutes (should be average over whole work unit) to ppd:

ppd = (Points * 14.4) / TPF in Minutes

I'm not that easily amused.
According to http://folding.stanford.edu/English/FAQ ... sNew#ntoc9 the formula for "PPD" with a QRB is

PPD = base_points * max(1, sqrt( k * deadline_length / elapsed_time))

which is misleading because max( ... ) is just a QRB scaling factor for base_points. It should say:

Points = base_points * max(1, sqrt( k * deadline_length / elapsed_time))

Therefore,

PPD = 14.4 * base_points * max(1, sqrt( k * Expiration / elapsed_time)) / TPF (in minutes)

provided that you return the WU within Timeout (Preferred Deadline in v6 terminology) and meet the rest of the QRB criteria. Also note that I replaced the somewhat ambiguous deadline_length with Expiration (Final Deadline in v6 terminology) for clarification. Converting elapsed_time to days from TPF in minutes yields

elapsed_time = TPF [minutes] * 100 / 1440 [minutes/day] == x [day]

So the PPD formula becomes a bit of a monster when you use just the QRB parameters as they are presented in Psummaries, and TPF in minutes:

PPD = 14.4 * base_points * max(1, sqrt( k * Expiration / (TPF*100/1440) )) / TPF
PPD = 14.4 * base_points * max(1, sqrt( 14.4 * k * Expiration / TPF)) / TPF

Let's quickly check that against http://www.linuxforge.net/bonuscalc2.php:
P8585 (Core_a3)
base_points = 1668
TPF = 15.25 minutes (15:15 min:ss)
k = 3.00
Expiration = 13.4 days
Timeout = 8.0 days
elapsed_time = 1.06 days
PPD = 9703.94

And my PPD formula would yield

14.4 * 1668 * max(1, sqrt(14.4 * 3 * 13.4 / 15.25)) / 15.25 =
14.4 * 1668 * 6.161115505 / 15.25 = 9703.938723 =
about 9703.94

Time to set the FAQ straight, perhaps? Using only TPF [minutes] and base_points, k & Expiration as they are represented in Psummaries would seem more "reader friendly" than using a vague term like elapsed_time and deadline_length - and still get it wrong. Of course, my formula doesn't take download/upload times and other factors like that into account, whereas elapsed_time is more generic term. In any case...

Now I'm amused.

[GreyWhiskers]

I think both times you mean 1440 (minutes)

Snore.

Right you are - it's 1440 - the number of minutes in a day.

[GreyWhiskers]

PPD = 14.4 * base_points * max(1, sqrt( k * Expiration / elapsed_time)) / TPF (in minutes)

I was just trying to show that however you come up with the number of points for a work unit (for example, looking at the credit line in the folding log after you have completed your upload), you can easily convert that to ppd if you have an average TPF.

Interestingly, the formula in your quote is essentially the one in my post (14.4 * points / TPF) with the QRB calculation for points thrown in so you have a single formula. That can be quite useful.

[TheWolf]

x

[Napoleon]

Or, one could calculate average PPD over several days. See http://folding.extremeoverclocking.com/ ... =&t=191980 for example. EOC calculates "Points 24hr Avg" as a sliding average from "Points Last 7days".

That is the result from credited points after WUs have been finished and uploaded. What GW and I (among others) were discussing here is PPD estimate, extrapolated from the TPF (Time Per Frame) of a currently running WU.

The estimates tend to assume that you are already qualified for QRB (Quick Return Bonus) points.