I don't know where you get your PICs from lijoppans but all the ones I've ever seen are extremely accurate and exactly predicatable. If you use timers and interrupts, at least 99% of the processing time will be available for displaying or other tasks, the RTC part of the code will be insignificant and run in the background. If you are using software timing loops to get delays you WILL have problems but thats the reason Microchip built the timers in silicon to take the problem away.
I am using a PIC to display the time and date inset in a TV picture, exerything is done in software, including all the RTC functions. It shows the time as HH:MM:SS DD-MM-YY as well as other text in the picture, handles a 38.4Kbaud serial link and a vision multiplexer at the same time. It's accurate to about 1 second per month but the time can be set via the serial link from a PC. It even works out leap years and daylight savings start/finish dates. All this is in a PIC16F628A clocked at 19.6608MHz.
Granted, you can use external segnment decoders like the 4511 but why bother when it only takes a couple of instructions in software to do almost the same thing.
Hemnath, the calculation for the timer values is simple but there are several calculators you can use if you want them. The best is probably here: **broken link removed**
Remember that timers count UP and generate an interrupt when they go beyond their maximum value and roll over to zero again so the value you use is the remaining counts before roll over, not the number of counts itself.
With a 12MHz crystal, even at the largest timer value in TMR1 you can't make a delay as long as one second so I suggest using a shorter inteval, say 10mS and counting 100 occurences to make up 1 second (100 x 10mS = 1 second). Bear in mind that you can do that in one instruction!
Brian.