How about the performance difference between an in-process cache and a memcached instance on localhost? Surely anyone who knows what they're doing will run the cache on the same machine as the web server.
(I guess that makes a good argument for not using a PaaS that gives you too little control over locality.)
This is a lazy answer, but it needs to be said: it depends on the context and the number of roundtrips to the cache.
If we're talking about a vanilla object cache, I think that it probably only costs 50% or so to make the extra copies and system calls. However, an in-memory "cache" can be something more than just a key-value store. E.g., for the thing I'm building right now, there are some more structurally complex graph structures that need to be traversed with every request to the "cache", and of course making a localhost RPC call for each entry/exit to the cache is really problematic. Though I admit that most caching is unsophisticated stuff, and in those cases it's just a ~2x difference.
(I guess that makes a good argument for not using a PaaS that gives you too little control over locality.)