SIP phones should be doing their own echo cancellation; they are in a much better position to do it as they have access to the linear version of the signal generating the echo and they only need to store a few milliseconds worth of speech, and for only one line, rather than the up to about a second, for every line, that the PBX would need to do.
Unless the softphone is more broken than we are assuming, the echo will result from audio coupling between the speaker (earphone) and microphone attached to the softphone.
However, the point is that the soft phone should be doing any cancellation, as it knows exactly what it is sending to the speaker, and the delay time between speaker and microphone is relatively limited.
Which is this softphone that we should be avoiding (assuming that this isn’t just a case of failing to turn on its echo cancellation)?
Analogue phones can suffer echoes due to transmission line effects, but that isn’t a problem for digital to the phone connections.