18/19M DC Done....full bore <10M Digit LL!!!!
 

[QUOTE=petrw1;164359]4 were mine in February and like Garo I am only a couple weeks away from completing my last dozen 18M DC's at which time I will have 9 CPUs in this range ... from a PIV 2.4 to 4 cores of a Q9550.[/QUOTE]

As of 22:18 server time today, I have completed the last of my share of the 18/19M DC assignments (64 in total).

:fusion:
I now have 10 (not 9) Cores (PIV 2.4 to Q9550) dedicated to cleaning up my remaining 64 (from a total of 87) <10M digit LL tests.

:spot:

I still have a bunch of 18Ms that are distributed among 4 CPUs. I realized I had a little too many so I unreserved some last week. I'll unreserve some more if it seems that my assigned exponents are hindering progress.

We now have 222 18M and 723 19M tests left. The Feb expiry and the fact that these exponents are only given to reliable computers that have returned a minimum number of LL tests means that this number is going down fast. We should confirm M[SIZE=2]20996011 as M(40) before the end of the year, unless....
[/SIZE]

[QUOTE=garo;165980]I still have a bunch of 18Ms that are distributed among 4 CPUs. I realized I had a little too many so I unreserved some last week. I'll unreserve some more if it seems that my assigned exponents are hindering progress.

We now have 222 18M and 723 19M tests left. The Feb expiry and the fact that these exponents are only given to reliable computers that have returned a minimum number of LL tests means that this number is going down fast. We should confirm M[SIZE=2]20996011 as M(40) before the end of the year, unless....
[/SIZE][/QUOTE]

I don't think you would be hindering progress ... the number is starting to drop slower and there are always some stragglers that, for whatever reason, will take a few months yet.

[quote=garo;165980]... and the fact that these exponents are only given to reliable computers that have returned a minimum number of LL tests means that [SIZE=2]...
[/SIZE][/quote]

I didn't know that the server will only give the those exponents to reliable computers. How does it know a computer is "reliable"? And what's the minimum number of LL tests? I'm curious to know what type of measure is being used.


P.S. I just realised that I got assigned an avatar :lol:

I think the number of stragglers will be a lot less this time round due to the server assignment policy.

To check assignment rules see [url]http://www.mersenne.org/thresholds/[/url] . You can find this link in the submenu under Primenet summary. To check your computers' reliability and confidence, look under the CPUs in your account summary and click on individual computers. The maximum reliability, confidence I have seen on my computers is 0.98, 10.0. To get a preferred exponent you need 0.9, 2.0. I don't know the exact formula used to compute these numbers but the more verified LL tests you complete successfully, the higher the numbers are.

[QUOTE=hockmeng;166073]I didn't know that the server will only give the those exponents to reliable computers. How does it know a computer is "reliable"? And what's the minimum number of LL tests? I'm curious to know what type of measure is being used.[/QUOTE]

A little insight here: [url]http://www.mersenneforum.org/showthread.php?t=11569[/url]

On Feb 23th there was 1 LL and 4 suspect LL missing below M(24036583), and 29 LL + 19 suspect LL below M(25964951):

[URL="http://www.mersenneforum.org/showpost.php?p=163681&postcount=91"]http://www.mersenneforum.org/showpost.php?p=163681&postcount=91[/URL]



Now there is 0 LL and 4 suspect LL left below M(24036583), and 5 LL + 6 suspect LL below M(25964951):

[B]18817921[/B] Suspect LL 58C0769123318B__ by "twinsun"
Assigned LL testing to "ANONYMOUS" on 2008-12-07
[B]21403913[/B] Suspect LL 90E42CAE7F1FF2__ by "Ars Technica Team Prime Rib" on 2008-07-25
Assigned LL testing to "starrynte" on 2008-12-07
[B]23422657[/B] Suspect LL 005C8E41DEAD07__ by "Charles B. Cameron"
Assigned LL testing to "Carsten Kossendey" on 2008-12-07
[B]23492927[/B] Suspect LL 4B2C1613F15C8B__ by "Steven H. Norton" on 2008-03-22
Assigned LL testing to "ANONYMOUS" on 2008-12-05


[B]24935683[/B] Suspect LL 7B258E954F3C83__ by "rudimeier" on 2008-11-30
Assigned LL testing to "curtisc" on 2009-03-06
[B]25419799[/B] Assigned LL testing to "James Hintz" on 2008-10-28
[B]25539649[/B] Assigned LL testing to "rudimeier" on 2009-02-05
[B]25556353[/B] Assigned LL testing to "rudimeier" on 2009-02-05
[B]25813453[/B] Assigned LL testing to "rudimeier" on 2009-02-05
[B]25838773[/B] Assigned LL testing to "ANONYMOUS" on 2008-12-06
[B]25945919[/B] Suspect LL A0EA024ECBD3D9__ by "John Meyers"
Assigned LL testing to "rudimeier" on 2009-02-05

[B]21403913[/B] Suspect LL 90E42CAE7F1FF2__ by "Ars Technica Team Prime Rib" on 2008-07-25

I did this one and got a repeatable error - so my guess is that this one is ok.

[QUOTE=garo;165980]I still have a bunch of 18Ms that are distributed among 4 CPUs. I realized I had a little too many so I unreserved some last week. I'll unreserve some more if it seems that my assigned exponents are hindering progress[/QUOTE]

Is there a way to make sure they go to fast testers? E.g. could you transfer them directly to me or any other one in this forum that is commited to finish them off first priority? Otherwise you might be unreserving them only to hand them over to somebody slower then yourself.

[QUOTE=lycorn;166685]Is there a way to make sure they go to fast testers? E.g. could you transfer them directly to me or any other one in this forum that is commited to finish them off first priority? Otherwise you might be unreserving them only to hand them over to somebody slower then yourself.[/QUOTE]

These assignments would be in the "preferred" range and only be assignmed to "reliable/confidence" CPUs.

See... [url]http://www.mersenne.org/thresholds/[/url]

[quote=garo;166359][B]21403913[/B] Suspect LL 90E42CAE7F1FF2__ by "Ars Technica Team Prime Rib" on 2008-07-25

I did this one and got a repeatable error - so my guess is that this one is ok.[/quote]
I am testing this right now (obviously), it should be done within 2 weeks