GPU to 72 status... - Page 258

kladner · 2014-04-28, 04:04

It is interesting, and somewhat clearer for me when I look at the Exact Graphs, in this case. The regularity of the spikes, with the consistent P-1 red cap is impressive.

TheJudger · 2014-04-28, 17:00

Quote:

Originally Posted by LaurV

[...]it could mean we can expect some heavy improvement of the speed for the new version of mfaktc?

false hope, sorry!

TheJudger · 2014-04-28, 18:59

Quote:

Originally Posted by kladner

It is interesting, and somewhat clearer for me when I look at the Exact Graphs, in this case. The regularity of the spikes[...]

Chris has asked me to report my results "at least once per week" so I dump my results each Sunday.

Oliver

LaurV · 2014-05-02, 06:02

Grrr... Hey Chris/[George], where those 15k come from? (it seems they are the same on PrimeNet, but why? were they forgotten, or some brain fart of the server when we moved to new web page format? Should we launch a fast TC?

henryzz · 2014-05-02, 13:32

It is always possible that this is deliberate to clear out the huge amount of extra errors in the 10M digit range.

markr · 2014-05-03, 00:06

How many verified & unverified did you see then in the primenet LL result report you linked to, LaurV? Now, it shows 112 distinct exponents with verified LL and 94 with unverified LL, for the same total as mersenne.info with its 110 2LL, 71 1LL and 25 0LL. The primenet summary does not show that bump either. I think it's a glitch in mersenne.info; it's happened before. Trust nothing!

LaurV · 2014-05-03, 03:22

Quote:

Originally Posted by markr

I think it's a glitch in mersenne.info; it's happened before. Trust nothing!

What glitch? There is no discrepancy between the numbers. What I linked on gimps page is just a different range, because I could not link all the range, too much time to display. But both PrimeNet and mersenne.info show the [B][U]same[/U][/B] data, i.e. the fact that there are [B][U]15k[/U][/B] exponents [B][U]between 33M and 34M[/U][/B] which are [B][U]not LL-tested, not even once[/U][/B], which is absurd. According with the milestones, everything below [B][U]51M[/U][/B] or so, is tested at least once. I think this was an error introduced once the PrimeNet was ported to the new format, or some chunk of the database is lost/destroyed. We may need, or not (George should know better where this problem comes from) to TC some of those exponents, just to be on the "safe" side. Do we? That was my question.

Ok, scrap that! I think this is intentional. Doing some totals on gpu72 candidates, I believe some exponents were just manually moved to "no LL test done" category, to increase their minimum TF bitlevel (from 70 to 71). If they "are DC-ed, i.e. done", then the GPU72 spider would not pick them for TF. This means we did (and still doing) TF on GPU72 for some 33M expos which were already DC-ed. Well, it would be nice to move everything to 71 there, but as we are low on resources, we better stop doing this, and concentrate on expos in the higher ranges, 36M, or so, which really didn't have DC done and need TF...

chalsall · 2014-05-03, 04:36

Quote:

Originally Posted by LaurV

Ok, scrap that! I think this is intentional. Doing some totals on gpu72 candidates, I believe some exponents were just manually moved to "no LL test done" category, to increase their minimum TF bitlevel (from 70 to 71).

Nope...

Keep in mind that Mersenne.info and GPU.com are completely independent. The database behind the former is actually based on a server here in Barbados which spiders Mersenne.org once a night, and uploads the summaries after the spidering is completed.

I'm not sure what happened here, but I'm going to wait until tonight's spidering has finished, and see if the problem corrects itself.

And, lastly, GPU72 is not assigning anything for DCTF'ing which is not still a real DC.

LaurV · 2014-05-03, 04:57

That's good to know!

chalsall · 2014-05-04, 15:46

Quote:

Originally Posted by TheJudger

Chris has asked me to report my results "at least once per week" so I dump my results each Sunday.

Bloody hell!!! A new GPU72 record: just short of 100,000 GHz Days done by one participant in seven days!

Thanks Oliver! We should now being able to start bringing back in candidates released at only 73 as well as those at 72 as they're recycled!

TheMawn · 2014-05-04, 18:38

Quote:

Originally Posted by chalsall

Bloody hell!!! A new GPU72 record: just short of 100,000 GHz Days done by one participant in seven days!

Jaysus... What is his hardware normally doing?

2014-05-02, 06:02	#2831
LaurV Romulan Interpreter Jun 2011 Thailand 7·1,373 Posts	Grrr... Hey Chris/[George], where those 15k come from? (it seems they are the same on PrimeNet, but why? were they forgotten, or some brain fart of the server when we moved to new web page format? Should we launch a fast TC? Last fiddled with by LaurV on 2014-05-02 at 06:08 Reason: added PrimeNet link and [George]

2014-05-03, 00:06	#2833
markr "Mark" Feb 2003 Sydney 3·191 Posts	How many verified & unverified did you see then in the primenet LL result report you linked to, LaurV? Now, it shows 112 distinct exponents with verified LL and 94 with unverified LL, for the same total as mersenne.info with its 110 2LL, 71 1LL and 25 0LL. The primenet summary does not show that bump either. I think it's a glitch in mersenne.info; it's happened before. Trust nothing! Last fiddled with by markr on 2014-05-03 at 00:08

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Status	Primeinator	Operation Billion Digits	5	2011-12-06 02:35
62 bit status	1997rj7	Lone Mersenne Hunters	27	2008-09-29 13:52
OBD Status	Uncwilly	Operation Billion Digits	22	2005-10-25 14:05
1-2M LLR status	paulunderwood	3*2^n-1 Search	2	2005-03-13 17:03
Status of 26.0M - 26.5M	1997rj7	Lone Mersenne Hunters	25	2004-06-18 16:46

2014-04-28, 04:04	#2828
kladner "Kieren" Jul 2011 In My Own Galaxy! 2·3·1,693 Posts	It is interesting, and somewhat clearer for me when I look at the Exact Graphs, in this case. The regularity of the spikes, with the consistent P-1 red cap is impressive.

2014-05-02, 13:32	#2832
henryzz Just call me Henry "David" Sep 2007 Cambridge (GMT/BST) 2³×3×5×7² Posts	It is always possible that this is deliberate to clear out the huge amount of extra errors in the 10M digit range.

2014-05-03, 04:57	#2836
LaurV Romulan Interpreter Jun 2011 Thailand 7×1,373 Posts	That's good to know!