FAHMON ppd reading is wrong

LOUISSSSS

Diamond Member
Dec 5, 2005
8,771
58
91
FAHMON has been telling me that i should be getting 4573 ppd since i've been using it. but whenever i check the F@H site and the stats on these forums they say i only get about 3520 ppd. whats with the 1000 point difference?
 

biodoc

Diamond Member
Dec 29, 2005
6,338
2,243
136
Fahmon is usually pretty accurate. Go to the preferences, click on the monitoring tab and then go down to the "calculate PPD by" and choose "last three frames" or "last frame" in the drop down menu. Then wait 4+ frames for each of your clients to see if it is more accurate. These choices will give you the theoretical maximum rate (24/7 dedicated cruncher with no down time) but if you use the computer for games, etc the real PPD will be lower.

Say you are running 2xSMP on a Q6600 and each take 18 h to finish (1760 pts each) & they both are synchronized.

Most days will be 3520 pts but there will be days you will have 7040 pts so you really have to go a couple of weeks to get the real PPD.

Hope this helps!

PS: there are some WU's for the console client (-advmethods switch) where fahmon thinks the WUs are 50 steps but really they are 100 steps thus fahmon shows 2x the actual PPD (a couple of the double gromacs B WUs fall into that category).

Cheers!:beer:
 

Insidious

Diamond Member
Oct 25, 2001
7,649
0
0
If you are using the Extreme Overclockers site for your comparison, it might explain some of it. That site's "24 hour average" is in DESPERATE need of a new algorithm. It takes several days of constant production for it to be even close to correct. :thumbsdown:

I don't know how they got "24 hour" for the name of that column. :roll:

-Sid

PS: the estimates from either fahmon or EOC will not affect the number of points you actually get..... I prefer to treat them as guidance rather than an absolute measure.
 

biodoc

Diamond Member
Dec 29, 2005
6,338
2,243
136
Originally posted by: Insidious

PS: the estimates from either fahmon or EOC will not affect the number of points you actually get..... I prefer to treat them as guidance rather than an absolute measure.

good point Sid!!;) You are absolutely correct!
 

Diogenes2

Platinum Member
Jul 26, 2001
2,151
0
0
Originally posted by: biodoc


Say you are running 2xSMP on a Q6600 and each take 18 h to finish (1760 pts each) & they both are synchronized.



Cheers!:beer:
Biodoc,
Could you tell me how to set this up, or point to a link with instructions ..

I know how to set up two installs in separate folders with different machine I.D.'s,
but when I ran them, it didn't seem to increase my PPD ..

Maybe I need to know how to ' synchronize ' ..

Thanks.

 

Insidious

Diamond Member
Oct 25, 2001
7,649
0
0
Originally posted by: Diogenes2
Originally posted by: biodoc


Say you are running 2xSMP on a Q6600 and each take 18 h to finish (1760 pts each) & they both are synchronized.



Cheers!:beer:
Biodoc,
Could you tell me how to set this up, or point to a link with instructions ..

I know how to set up two installs in separate folders with different machine I.D.'s,
but when I ran them, it didn't seem to increase my PPD ..

Maybe I need to know how to ' synchronize ' ..

Thanks.

the synchronize was just to set an example.... you don't have to do anything to make that happen.

Most of us are using a utility called SMP Affinity Changer. It really ups the PPD (~25%) for two instances on a Q6600. (edit: IF you are running windowsXP)

you install it once and then it takes care of itself after that.

Good luck!

-Sid

 

biodoc

Diamond Member
Dec 29, 2005
6,338
2,243
136
Sid is THE expert when running 2xSMP on WindowsXP with affinity changer!!:)

:beer:
 

Insidious

Diamond Member
Oct 25, 2001
7,649
0
0
Originally posted by: biodoc
Sid is THE expert when running 2xSMP on WindowsXP with affinity changer!!:)

:beer:

Not really, but thanks! :cool:

:beer: to ya Doc

-Sid

 

uncle fungus

Junior Member
Dec 30, 2007
9
0
0
If you're wanting FahMon to display close to the real PPD you will get form your machine choose the "Effective Rate" option as that takes into account computer down-time and projects it forward. Current frame and last 3 frames are useful if you want to know the instantaneous speed your client is running at, and the average PPD uses data from the first 255 frames processed of a particular project (which isn't actually that useful but I've left it as an option anyway).

FahMon 2.3.2 which is currently being tested fixes the issues with non-standard frame counts.