7+ table

[garo]

Size	Base	Index	+	Diff	Ratio
245	7	386	+	326.2	0.751
314	7	388	+	327.8	0.957
221	7	395	+	267.0	0.827	/5q
300	7	397	+	335.5	0.894
332	7	398	+	336.3	0.987
260	7	401	+	338.8	0.765	
316	7	412	+	348.1	0.906	
256	7	416	+	324.5	0.787	/13
331	7	422	+	356.6	0.926	
231	7	425	+	287.3	0.802	/5q
224	7	430	+	290.7	0.76	/5q
351	7	431	+	364.2	0.962	
301	7	436	+	368.4	0.815	
313	7	439	+	370.9	0.842	
261	7	442	+	344.8	0.755	/13
365	7	443	+	374.3	0.973	
366	7	446	+	376.9	0.969	
303	7	448	+	324.5	0.932	/7

[sean]

7,350+ C139
5168898551438758822224323913308984024753373866126288201 (p55) *
577933906565255372277248693235754607933851459064795784559428005583229072181753383001 (p84)
by GNFS, 11 days

[akruppa]

2300 curves at B1=11M on 7,277+ 7,308+ 7,340+ 7,361+ 7,369+ 7,374+ 7,377+ 7,380+ 7,386+ 7,393+. Adds 0.50108 to p45 and 0.06662 to p50 for each.

Alex

[R.D. Silverman]

Originally Posted by akruppa

2300 curves at B1=11M on 7,277+ 7,308+ 7,340+ 7,361+ 7,369+ 7,374+ 7,377+ 7,380+ 7,386+ 7,393+. Adds 0.50108 to p45 and 0.06662 for each.

Alex

If I interpret a recent email from Bruce Dodson correctly, Bruce has run
sufficient curves to add (at least) 1.0 to p45 for all Cunningham numbers
below 260 digits. He is working to achieve the same result for those over
260 digits.

Bruce suggested that in another year, running ECM on the current
Cunningham numbers will not longer be worthwhile. I would like to see
ECM run to at least level 1.0 at p50 before succumbing to this point of view.

[philmoore]

Originally Posted by R.D. Silverman

Bruce suggested that in another year, running ECM on the current Cunningham numbers will not longer be worthwhile. I would like to see
ECM run to at least level 1.0 at p50 before succumbing to this point of view.

This seems surprising, as I would expect that there are quite a few Cunningham numbers on which one would want to do ECM factoring to the 55 or 60 digit level before proceeding with SNFS.

[R.D. Silverman]

Originally Posted by philmoore

This seems surprising, as I would expect that there are quite a few Cunningham numbers on which one would want to do ECM factoring to the 55 or 60 digit level before proceeding with SNFS.

One must consider the possibility that for the larger composites, finding a
factor will still leave a composite cofactor that must be finished with NFS
anyway.....

[bdodson]

That's p49 = 5980904229992785230777515196566061969934200643121,
found with b1=260M (and ecm-6.1). It's my 2nd July factor, the first
from the Opterons, for 17-days. In the last 600 curves needed for a
complete t50. I've spent something like 38K curves since my last Opteron
factor, a p47. If condor were able to support sieving, the time might well be
better spent on snfs? I'll likely be running ecm through Nov 2006, with
good progress on 1.0 t50 (cf. elsewhere).

Bruce Dodson

[geoff]

7,539L C142 = P63 * P80
P63 = 628632369546652104427939137048349725113498032710782245605608731

This was by SNFS (difficulty 195.2) with GGNFS (CVS 20060513) using 28 bit large primes and factor base limits of 20 million each side.

Sieving took 104 GHz days on a mix of P2 and P3 CPUs, linear algebra (on the
second attempt) took 24 GHz days on a P4. Peak RAM usage was 978 MB.

[bdodson]

Originally Posted by akruppa

2300 curves at B1=11M on 7,277+ ... 7,377+ ...
Adds 0.50108 to p45 and 0.06662 to p50 for each.

Alex

The Lehigh count on c234-c250 is at 56.38% t50, which was enough
to finish 7,377+ C248 with

p50 = 46215747540095507833650423122055432531249537597319
and a p198 cofactor. An xp-condor factor for ATH's "Prescott without"
binary of 6.1.1, the 3rd so far. This is the first of the low-memory run,
on public pcs unlikely to switch to sieving, so other things being equal,
another 3500/7830 curves to go for t50.

The Opterons are on the last run of 525 curves with b1=260M to
finish t50 on c211-c233, running the last 75 numbers. -Bruce

[bdodson]

Originally Posted by garo

Base    Index    Size    11M(45digits)  43M(50digits)    110M(55digits)    260M(60digits)    Decimal

...
7    391+    C281    0(0.267423)    0(0.0522979)    165(0.00921839)    0(0.00148122)    15447648579494257995639523563041579963781419175650953782737517531404675841338563961311901583097189888459303423561315822968403211737689491218288706539727488463400652825916891808614621701770654797080449720857175315477849499313649816507699060789062875463578838378145683464685573644031
....

Another small xp factor (to go with the p50 Opteron factor, as my first
from 2007, in the non-sticky 2+ discussion).

p46= 4429642801163795117773992875901047668871730239

As reported in the title (especially for people that don't read titles!),
the cofactor is a 235-digit composite. Quite some ways to go on
this one. An earlier report in this thread mentioned an additional
3500 xp curves (that's b1=43M, gmp-ecm611 ATH's prescott binary;
b2 so that 7830 curves are needed to test for p50) being needed
to complete p50 on c234-c250. Looks that would have been with
800 curves finished; while the current count is 2400 finished, so
3500-1600 = 1900 left to go?

This p46 is from the c251-c299 range, which had an initial t45, and
is now at new 1200 curves, b1=43M. The 2+ and 2- in c251-c366
have had 2000/7830, enough for a bit more than a 2nd test to p45,
so I'm working on bringing the rest of c251-c299 up to 2000. The
least tested range of Cunningham numbers is the part of c3xx
(i.e., 300-c366) that's not on the 2- or 2+ list (for n<1200, so
c3xx's from 2LM is under-tested, relative to the rest) --- these ones
have only had 1000 new b1=43M curves beyond the initial test to
p45. Hope this suffices for a current report-from-the-front, pending
a comprehensive update on my July post on the 2.0-thread.

In brief, the race between the AMD's, xp's and the old P3's here isn't
a very straight comparison, on several counts. Never-the-less,
1 for the Opterons, 1 for the xps, so far in 2006.

Regards, Bruce

[akruppa]

> (especially for people that don't read titles!)

Huh? Who, me??

Alex