problem with Unicast and non-Unicast graphs/counters

[iDen]

Default cacti template for unicast PPS and non unicast PPS drawing graphs with number of packets they are getting at polling time.
How it's possible make to cacti to graphs with diffrence data between old and new values it gets from polling ?
For example in mrtg case it makes simple function and shows difference between polls:
like in/out 2053/2856 packets per second
while cacti showing 481.66k/481.66k PPS
And by default Data sources are indicated only by "IN" data source. Have put OUT source manually.
Also Unicast and non-Unicast graphs are absolutely identical, but snmpwalk and mrtg showing different proper data.
Cacti screen

Mrtg screen

[gandalf]

Using a data source type of COUNTER would allow for such a graph
R.

[iDen]

It is COUNTER right now.
I got idea.
graphics showing not PPS Packet Per Second, but PPM Packet Per MInute.
Sorry, even PP5M Packet per 5 Minutes

It's wrong.
So to get PPS out from PP5M i have to devide some 600M value by 300s and will get packets value per one second.
I don't want to change polling interval to less then 5 mins.
I've tried to add to data templates, associated RRAs - "Hourly 1 minute avarage" but no luck.
I'm not PRO so can't find solution in such case where to put function so PPS graphs would be showing always real PPS and not PPM or even PP5M. Even weekly and monthly graph should show exactly PPS just grid scale should be changed on graphs.

And yes i know that it won't be exact values of PPS because of simple "devide function" but SNMP not giving this query every second.
What ever it would be better than PPM PP5M .... 2.6 billion packets (LOL) when it should be 2k-5k per second.

[iDen]

oops sorry for broken screens. fixing it

Other question was unicast and non-unicast graph were same

[gandalf]

Please see those graphs at Graph management and use DEBUG to find out, whether they are using the same Data Sources. Next, See "System Utilities -> View Poller Cache" and find out, if they are polling the same OID which would be wrong as well
R.