GGNFS benchmarks wanted
 

I am slightly surprised that there isn't a gnfs-lattice-sieving benchmark thread already. So ...

Put the following into a file
[CODE]
name: 3-2.451
n: 3331540433209471147641567437194357362373622006000773687374639172326065960029337949460127653443576565403802071544024765946721314281537931
skew: 150691.94
# norm 3.15e+18
c5: 49980
c4: -511662381902
c3: -13290347110466912
c2: 9133385823294241706943
c1: -18916405907713323753737324
c0: -21335066943024777785076328167089
# alpha -5.66
Y1: 615222815884693
Y0: -146141483215623979683925156
# Murphy_E 3.29e-11
# M 2147849860332370447219772565321341023202516478418307635505723421624114672958437230909094686728473287311105869332501291519990254052678052
type: gnfs
rlim: 6000000
alim: 6000000
lpbr: 26
lpba: 26
mfbr: 49
mfba: 49
rlambda: 2.5
alambda: 2.5
q0: 40000000
qintsize: 5000
[/CODE]

(experts will point out that these are far from optimal parameters for a C136; indeed they're not, I have a perl script looking all over parameter space for better ones)

and time how long 'gnfs-lasieve12e -a file' and 'gnfs-lasieve14e -a file' take to execute. 12e should produce 397 relations in between three and five minutes, 14e should produce 1624 in between fifteen minutes and half an hour.

Results on machines in the office:

2.8GHz Prescott 12e: 272.3 seconds 14e: 1831s
2.2GHz K8 12e: 184.6 seconds 14e: 1154s
2.4GHz Core2 12e: 156.2 seconds 14e: 974s

I would particularly be interested in results for faster Preslers and slower Core2s, since the price-optimal compute-farm-node seems at the moment to be 3200MHz P4 / 2400MHz AthlonX2 / 1866MHz Core2 (in increasing order of price)