It could be that the old server had a different number of system buses (NUMA nodes) and this is affecting PLE. I assume you're looking at the Buffer Manager: PLE counter? Given the specs you've said above, you must have multiple NUMA nodes and this counter is the average across all of these. Buffer Node: PLE gives the stats for individual pools.

Threads can access foreign memory to their local NUMA node, but there's a preference for locality, which I'm assuming would end up recycling individual (smaller) pools faster than one big one, therefore have a lower average PLE. In some situations, that could be worse for overall system performance, but generally not.

What are you seeing in terms of overall system performance, e.g. wait stats, average transaction times etc.?