I agree, it's all about that resource allocation model again.
And just a few things that haven't quite already been said.
Factoring will only find 'smooth' factors. Yes I know that by adjusting B1 and B2 it can test for progressively less smooth ones, but it won't find a lot of the sieved factors.
Sieving does factoring a little favour. It allows the 'already checked to...' to be increased. Currently we're at 2^47, soon that'll be 2^48. Not sure how much of a difference this makes, but it must be something.
Sieving and P-1 live in harmony. I agree that P-1 needs more resources. If I had more suitable resources, I would do more P-1.
Looked at another way, I have a dusty old P2 that spends most of it's life acting as a debug console with hyperterminal, and it found
Code:
255.576T 67607 5937147 28153.104 Sat 20-Mar-2004
It has 128K of RAM, and if it wasn't sieving it would be doing nothing.