I thought I had better write something about how things are going for the 321search project.
We have no numbers less than 1 million digits left for first time testing.
PrimeGrid have been on board for a couple of months and they have increased the search's throughput greatly.
We have started testing at 4 million bits. The new batch here for manual testing takes much longer, nearly 50% longer, to crunch compared to the previous batch because the new ones are much longer -- PrimeGrid and James Burt are doing the intermediate ones -- and use a bigger FFT window. The effect of worse caching and bus contention must kick in at some length -- whether we are there yet I do not know.