I recently managed to keep BOINC from prefetching GFN tasks which caused each task to have a 2 minutes longer send/receive duration. I didn't see any change in the "first ratio", though.

The reason being, nearly all tasks where I'm second the other host got the task 10-15 minutes earlier than me. With these short tasks that means, they are first. Of course, the same is true other way around, I often get 1st against 2080s just because I got the task 10 minutes before them.

Accordingly, 1st ratio is around 50%, since it's randomly decided who gets the task first. This might even be a good thing, but I'm not so sure. :) Is there a way to prevent this from happening or are the GFN tasks so rarely fetched that it cannot be helped? It's GFN17-mega.
____________
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 5,700,000

I was curious about how often not prefetching tasks helps to get a 1st, so I looked at the range including my most recent 100 validated workunits for which I got 1st (without prefetching) for GFN-16. They take me about 5:11 each.

216 total WUs
118 (54.6%) 1st (or still pending), and would have been 1st even if I had prefetched (that is, I returned task at least 2:30 before wingman)
89 (41.2%) 2nd
9 (4.2%) 1st, but would have been 2nd had I prefetched (that is, I returned task less than 2:30 before wingman)

Conclusion: for roughly 95% of workunits, the server decides who gets 1st.

I'm not certain how this scales to longer tasks.

Edit: I looked at BOINC manager as a task was completing, and it seemed to take roughly 8 seconds between completing one task and starting the next. This is about 2.6% of the time it takes me to complete one task, so if I had prefetched, I would have been able to do an extra 5 WUs in the time it took to do the 216 above. Maybe better to turn it back on...less selfish too...

Accordingly, 1st ratio is around 50%, since it's randomly decided who gets the task first. This might even be a good thing, but I'm not so sure. :) Is there a way to prevent this from happening or are the GFN tasks so rarely fetched that it cannot be helped? It's GFN17-mega.

Ok, good to know. I'm not sure if it's even a bad thing, it has a similar result as LLR2 had.

Even with a slow GPU you get a lot of 1st. But having a fast GPU still gives the advantage of more numbers tested.
____________
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 5,700,000

Please keep in mind that for most of our projects, the concept of "being first" in order to discover a prime either has already come to an end, or is likely to end sometime in the not too distant future. At some point, I expect all LLR and all Genefer projects to use fast double checking. Only the sieve-like projects AP27 and the upcoming WW project will still have the competition to be first.
____________
My lucky number is 75898⁵²⁴²⁸⁸+1

At some point, I expect all LLR and all Genefer projects to use fast double checking.