We have had much better success using sntp.
We have determined that we have better time control if we use one server to use ntp to contact a ntp server and all other network machines use sntp and cron to update their local clock. We do this by adding a cron command similar to:
*/5 * * * * root /usr/sbin/sntp -P no -r 192.168.1.1 >/dev/null 2>&1for all the other servers not running ntp.
The "-P no" prevents prompting if the user for time changes outside of the maximum value (default 30 seconds.)
This solution sets the local clock on the machine to the correct time from the locally available ntp server which typically gets its time from a known, low stratum ntp server. We also often stop and disable the ntp damon on the server.
Let us know what you have done or run into with ntp.