I tried n=81 yesterday for 2<k<2^32. As far as I know, there is no upperbound, since NewPGen splits this large range into smaller pieces (dependent on the max. RAM assigned?) and sieves these ranges until 1 billion.
As I said, I tested n=81 yesterday with OpenPFGW and there was no complete set in the log-file. I still have the sieve file, so anyone who wishes to redo it can ask me for the sieve-file. I uploaded the sieve file on
http://www.diamondvalley.nl/octoproth
I'm taking 82+....sieving until k=2^32 takes about 1,5 or 2 hours for me. Testing these numbers for PRP is done in less than a minute (small n)