Thanks. As I understand it, due diligence for a C180 poly search is ~1 month of GPU search time, though a new record e-score would seem to be “good enough”. I’ll expand my search space up to 1M. Will take at least a week.

Abandoning C180_132_95 - it’s already in 15e of NFS@Home. I grabbed the wrong composite when I started this effort. Taking a break from suggesting any more work proposals for a bit of self-reflection/deep embarrassment.

Sean-
You organize much of the work for XYYX, NFS@home, and some others. You personally do nearly half the LA work to support NFS@home. Such mistakes are an "error rate", and yours is quite low.

If it makes you feel better, I've run entire factorization jobs on personal projects twice; I didn't notice I'd run it the first time without reporting factors, so a week or two later I ran it again. Something like GNFS-140, not a huge waste, but still silly-feeling.

[QUOTE=VBCurtis;480933]Sean-
You organize much of the work for XYYX, NFS@home, and some others. You personally do nearly half the LA work to support NFS@home. Such mistakes are an "error rate", and yours is quite low.

If it makes you feel better, I've run entire factorization jobs on personal projects twice; I didn't notice I'd run it the first time without reporting factors, so a week or two later I ran it again. Something like GNFS-140, not a huge waste, but still silly-feeling.[/QUOTE]

Thanks. Still silly-feeling, most especially when my mistakes cost others their time/electricity.

C163
 

Can either of these be improved?
[CODE]N: 5291610598339803917854125624126199550883921985264025408871351973087265336861502115187345391316093888171298195520496677064033950538617659817036392737180252641770449
# 2878319791561117685582532984481924989856693782709674033037185004142735435821967737094116024989323^3-1
# expecting poly E from 8.58e-13 to > 9.86e-13
R0: -21320358869947392671304530479533
R1: 6671826766064153
A0: 125826081901408251812550173468825887640
A1: 634273237252226932148271560786862
A2: -174251494415887651731430603
A3: -407441597860652112408
A4: 10659191447702
A5: 1201200
skew 4242403.59, size 6.573e-16, alpha -7.364, combined = 8.646e-13 rroots = 5[/CODE]
[CODE]R0: -26390248613942286953255905942564
R1: 3393833987825897
A0: -17738613766983291273874478684163681784971
A1: 4974530346897943021359644171482785
A2: 1274931911768843829093573653
A3: -226879686479200332613
A4: -19308765389110
A5: 413400
skew 9513082.26, size 6.486e-16, alpha -7.678, combined = 8.642e-13 rroots = 5[/CODE]

Do you mean a tweak from Max or a further search for an e-score > 8.6e-13?

The [url=http://www.mersenneforum.org/showpost.php?p=478855&postcount=86]high water mark[/url] for a C163 is 1.096e-12, so there’s some room for improvement.

Hoping for a little tweaking. It is an easy 31-bit job but with a good poly it can fit as a 30-bit job.

C163 poly
 

CADO output for the second one is below. It's slightly lower in the E score but may sieve faster. Please test-sieve.
[code]
R0: -26390248615952502337921014298222
R1: 3393833987825897
A0: -20193065923418146666342028352047917688361
A1: 3241721956776338193620593742878017
A2: 1636579644193690708337989739
A3: -179681922722879266453
A4: -20533078427110
A5: 413400
skew 9550000.00, size 6.486e-16, alpha -7.678, combined = 8.640e-13 rroots = 5
[/code]

CADO output for the first one is below. It has the same E score but may sieve faster. Please test-sieve.
[code]
R0: -21320358870655493661467217229882
R1: 6671826766064153
A0: 57034389014970204975174936752613271965
A1: 657442166892199977313148822853228
A2: -43816453040895573316523743
A3: -411831459989495669872
A4: 10021756649702
A5: 1201200
skew 4250000.00, size 6.573e-16, alpha -7.364, combined = 8.646e-13 rroots = 5[/code]

C159
 

From the OPN 11232875364041^17-1 number.
This is pretty good but an optimize might be better.
[CODE]N: 567033079351417599792287481827798838189390134708539997609886372751620595472206543240914774142736016505565453857390328265649123632896794230620769751076259286187
# expecting poly E from 1.49e-12 to > 1.72e-12
R0: -4838152153707467647999587052081
R1: 1979898218066797
A0: 2525402780109740104271968199475880800
A1: 9596414219759589072414289455240
A2: -19454805664515688837147990
A3: -6920023955982287039
A4: -574862570116
A5: 213900
skew 2793565.99, size 1.855e-15, alpha -6.134, combined = 1.650e-12 rroots = 3[/CODE]

C159
 

@RichD
After CADO root-opt. Almost the same score, may sieve better, please test-sieve.
[code]
Y0: -4838152154122121691605752481385
Y1: 1979898218066797
c0: -275339914149872251015276659986247904
c1: 16857939037966920610872863799144
c2: -15277918181835796114510750
c3: -6344625174591414591
c4: -798850094116
c5: 213900
skew: 2805251.73042
# lognorm 48.56, E 42.42, alpha -6.13 (proj -1.57), 3 real roots
# MurphyE = 1.64933066e-12
[/code]

A C159 isn't worth the time to test-sieve. It doesn't make sense to spend 1-2 hrs of human time to save 2-4% of time on a 200-CPU-hour project. Human time is worth way more than 10x computer time!

Just grab the highest score and fire up the sievers. I'd probably compare the top two at C165, and any polys that score within 3-4% of the best at C170. Below 160, things are fire-and-forget unless one is consciously on one's learning curve.

This is why we ask for your help on nearly every C180+, but rarely for below C170. At C180, 2% could be 500 thread-hours.