Stats Milestone - NEW

[Joe O]

3. Sieve to 750T

Less than 25T reserved but not complete below 750T.
8T not yet reserved below 750T.
Looks doable!
AND
only 13T reserved but not complete between 750T and 800T! (and I know that a lot of it has been done and submitted)
0T not yet reserved between 750T and 800T!

[VJS]

Might also want to look at this one...

4. Double check to 1M

[mdettweiler]

Originally Posted by VJS

Might also want to look at this one...

4. Double check to 1M

Maybe a second LLRNet server (you could do it on the same physical server, but on a different port, like SR5 does) could be set up for doublechecking work.

Or, when PrimeGrid gets on board with BOINC LLR for PSP, they could do doublechecking--either they could do just doublechecking, or they could let you choose which one to do, doublecheck or first-pass. (They already have a nice mechanism set up so you can choose exactly what blend of their handful of different types of available work you want, so this probably wouldn't be too hard.)

[Joe O]

Originally Posted by VJS

Might also want to look at this one...

4. Double check to 1M

And just how far have you advanced the double check effort

1.2M

Congratulations

[ltd]

For the low ranges a llrnet server is not possible. The reason for that is that all these results were calculated using either "PRP.exe" or an older version of "LLR.exe". The residues are not compatible to the residues created using
the llrnet software.

To get usefull DC residues for the low ranges the only way is to handle them manual. The use of the new llr software started somewhere around 1.2M and above 1.5M most of the results are llrnet compatible.

But looking at the quality of results created i would not recomment to make double checks above 1.2M before we did not reach n=5M or maybe even 6M.

[VJS]

One of the major reason why I'm doing double checks now is to verify the stability of my highly overclocked q6600.

In other words, there is no benifit to the project to do double checks currently.

Once I have 100% stability I will switch over to firstpass.

Perhaps the only advantage to double checks currently... I suppose Lars only has to deal with one person and the results can be mailed back and forth.

By the time I'm finished with my stability testing I would think the double check effort won't be need until we reach at least 6M, perhaps that may change if we eliminate two primes before then.

----------

If you want to reset the benchmark to all prp.exe residues doublechecked. That might be a better project goal, I have a feeling that is somewhere <1.5M.

----------

In ortherwords if your stable run firstpass, if your not stable I'd suggest sob-secondpass.

[Joe O]

Originally Posted by Joe O

Less than 25T reserved but not complete.
8T not yet reserved.
Looks doable!

3. Sieve to 750T

Everything has been reserved, and only 9T reserved but not yet complete.

[VJS]

4. Double check to 1M

This one is close to done.

Also we should change all references from prp to llr

[Joe O]

Originally Posted by Chino112

1. 12 thousand PRPs to go until we reach this. only 274 until done all <3.5M

2. total untested at 1.21M

3. since reservations have just past 550T, a much higher target needs to be set. 750T?

4. no double checking at present

5. a constant, no?


so what about

1. PRP all candidates below 4M
2. Total untested <1M
3. Sieve to 750 trillion
4. Find a prime

Originally Posted by Citrix

Let's propose new milestones:

1. PRP all candidates below 5M
2. Total untested <1.2M
3. Sieve to 750 trillion
4. Double check to 1M
5. Find one more prime if possible (I think there is one under 5M)

I think these milestones are more realistic. Though, I don't know if we need to double check to 1M, right now.

Originally Posted by Citrix

Approaching untested tests <1.2 M left. (Should reach there in 1-2 days)

Congratulations and good work guys. Keep it up.

(Waiting for Geoff to implement the improved sieve, so we can sieve faster and reach the goals before the end of the year)

Originally Posted by ltd

New small milestone.

Today the first test with n>=4Mil was handed out.

And the winner is SB2.

1. We have gone from 12000 candidates untested with n below 4M to 85 candidates untested.
   There are 19900 candidates untested with n below 5M.
   Our goal for 2008 remains at 5M?
2. We have gone from 1210000 candidates untested with n below 50M to 1153921 candidates.
   Our goal for 2008 remains at 1000000?
3. We have completely sieved to 885T, well beyond our goal of 750T. We have also sieved, with only a few gaps, to 1300T non-BOINC and 2000T BOINC.
   Our goal for 2008 should be 1500T for non-BOINC and 5000T BOINC?
4. We have doublechecked almost everything with n below 1.5M exceeding our goal of 1M.
   Our goal for 2008 should be 4M?
5. We have not found a prime
   Our goal for 2008 should be to find two primes?

I have put suggestions for new goals after each point. They are followed by ?, because they are only suggestions. Let the discussion begin!

[ltd]

Originally Posted by Joe O

1. We have gone from 12000 candidates untested with n below 4M to 85 candidates untested.
   There are 19900 candidates untested with n below 5M.
   Our goal for 2008 remains at 5M?
2. We have gone from 1210000 candidates untested with n below 50M to 1153921 candidates.
   Our goal for 2008 remains at 1000000?
3. We have completely sieved to 885T, well beyond our goal of 750T. We have also sieved, with only a few gaps, to 1300T non-BOINC and 2000T BOINC.
   Our goal for 2008 should be 1500T for non-BOINC and 5000T BOINC?
4. We have doublechecked almost everything with n below 1.5M exceeding our goal of 1M.
   Our goal for 2008 should be 4M?
5. We have not found a prime
   Our goal for 2008 should be to find two primes?

I have put suggestions for new goals after each point. They are followed by ?, because they are only suggestions. Let the discussion begin!

1. I think we should stay with that goal. The runtimes of the llr tests have increased that much that it will not be easy to reach that limit.
2. Lets stay with it but only reachable if we find a prime.
3. 5000T is hard to reach. But lets go for it. For the non BOINC part I have no clue about the progress so I don't know if it is an easy goal.
4. With the amount of factors found and the error rate we have at the moment I think we should not push that high.
5. That is OK. But to bad that we have no real influence here.

[VJS]

1. I think 5M is a reasonable goal will probably get further.

2. I never really understood this goal since it relys on other goals such as sieve and testing.

3. Yup, I'll agree that finishing all manual sieve to the Bionic level is a must for 08'. Pushing Bionic to 5P not sure about that I think 4P may be a realistic strech.

4. 4M is a little high, I think making the double check effort more public should be a goal. We will certainly get double check to ~2M even at the current rate. Perhaps getting secondpass to greater than 50% of firstpass might be a good goal. I'd also propose if the error rate is not that high, continue with a select k approach. For example by the time n reachs 5M in firstpass double check should be around 2M, with 1 or 2 k's around 4.5M. This would help reduce the possibilty of an n-range with alot of errors in it. It would also help identify users producing bad data.

5. Yup this is a guess, but 2 to 3 primes by years end doesn't sound unreasonable. We have gone through a prime lull lately.