Live info + questions about doublecheck

[VJS]

Oh and 0.5M n in less than a month is a good increase.

Try to stay with it for a while longer if you can. The k we are testing is very light compared to the average so we should advance fairly quickly.

It's quite possible that we will catch firstpass but we will need a little more help to do that. It's my feeling that we will lag behind by around 1M with our current effort, I havn't put any numbers to it just a rough guess.

[ltd]

For the jump from n=3.3M to n=3.5M there is a simple reason. I have two manual reservations for my offline PCs.
One is from n=3.3M to 3.5M and the other is from 3.6M to 3.9M.
Above 4M there will be another big gap where primegrid had its range with every result already beeing double checked.

Depending on the availability of my machines I will reserve another range for me late in january.

[opyrt]

Originally Posted by VJS

Oh and 0.5M n in less than a month is a good increase.

That was kind of what I was trying to write in my post...
And it was also a comment to the fact that 17 tested n's every day was too good to last.

Try to stay with it for a while longer if you can.

Sure thing! I don't want my quad workstation at work to be bored during my x-mas holiday!

[opyrt]

We're closing up on 4M now!

If I understand correctly, the ranges 3M3-3M5 and 3M6-3M9 will be finished by ltd late january? And from 4M, primegrid has done a huge chunk?

That should mean that the first half of 2009 will probably bring the DC to 5M? Or is this just wishful thinking?

[VJS]

Thats the way understand it as well.

Getting to 6M before the end of next year should be very very easy in double check. The question is are we going to close in on the firstpass level by next year??

------------------

Lets do some rough Math to put things in perspective... and estimate our success chances.

B/c of our one k approach and the fact we are using a light k; double check really only needs to do a portion a small portion of the worktest that first pass does to maintain its n level. (about 6%)

This is based on test denisty per 1M of n.

Assuming we don't find a prime (which we will soon IMHO)

That means secondpass needs to do one test for every 16 tests firstpass does. (after a prime maybe 1 for every 10-14...)

Firstpass has been averaging a litle more than 80 tests per day.

80 / 16 = 5

So Joe and I by ourselves were barely holding.

Now that opyrt has come on our tests have increased to about 12 per day.

So we are gaining fast on the firstpass level, those huge chucks help as well.

Problem will be our tests are easier than firstpass currently. Perhaps firstpass tests are twice as hard???

In either case we are gaining but will we reach firstpass level with our current effort? Probably not since more people are joining. Will we come close to firstpass???? We should be able to bring in to within 1M.

^
|

Does any of this make sence???

[opyrt]

Originally Posted by VJS

That means secondpass needs to do one test for every 16 tests firstpass does. (after a prime maybe 1 for every 10-14...)

Firstpass has been averaging a litle more than 80 tests per day.

80 / 16 = 5

I think this makes sence. And as secondpass tests take about half as much time as firstpass at the time writing, and we are doing approx. 12 a day, I am pretty sure we should be able to catch up with firstpass. Atleast in my mind it seems correct that if we were doing 10 secondpass tests a day, we would be at the same speed as firstpass according to your calculations.

And this is when we don't think of ltd's huge dumps of results from his offline computers, that arrive every now and then. :)