Well, what do you mean by 48GB load? Do you mean an application that allocates 48GB of data?I knew someone was going to chime in about workload. Obviously. The question is, at what point is there a performance fall-off? Test all three machines with a 48-gb load.
In that case, it's honestly (and I know this isn't a simple answer) still going to depend, it won't really be entirely the same on each application. Some applications will allocate a lot of memory and will be actively hitting all of it frequently. These kinds of applications will see a fairly steep performance cutoff if the amount of RAM on the system is less than you have.
There are, however, a surprisingly large number of applications that might not take that big of a hit. Take virtualization, for example. Suppose you have 48GB worth of virtual machines on a 32GB Mac. The chances are slim that all of these virtual simultaneous machines will be frequently accessing 32GB worth of data constantly, so in all liklihood, you wouldn't have that much of a performance fall-off on just a 32GB Mac.
Max Tech (I know that name isn't especially well like around here) actually attempted to test something like this last year. They took 8GB, 16GB, and 32GB Macs and tried a whole series of different photo, video, audio, and productivity apps as well as various multitasking scenarios. They found that with with their specific tests, the difference between 8GB and 16GB was pretty noticeable, but the differences between 16GB and 32GB were much smaller (in the realm of 3-5%) even though swap was fairly heavily used on both. This was just on their specific tests, more scientific workflows that need giant amounts of RAM obviously would have had a very different result.