I mean 250M nonunique.
At the start of December I did the C164 cofactor of 1920:2520 using 32bit large primes and 14e, sieved 25M75M for 247M relations (223M unique) in 3250 corehours, and got a perfectly usable 12.2M matrix at density 70. At the start of November I'd done the C163 cofactor of 5748:1533 using 31bit large primes and 14e, sieved 25M75M for 152M relations (135M unique) in 3200 corehours on the same cores and got a 9.8M matrix at density 70.
In my experience an SNFS251 took about as long to sieve as a GNFS172, and an SNFS280 about as long as a GNFS187; that suggests mid170digits is practical in _d and high180digits in _e.
Last fiddled with by fivemack on 20150126 at 23:59
