Status reports in Prime95

[ThomRuley]

I am currently setting things up to start trial-factoring 56.5-57M to 60 bits. I set the FactorOveride to 60 and gave the computer a list of exponents in that range to trial-factor (it seemed more efficient to give the computer a bunch of prime number exponents to test than to tell it to trial-factor every exponent in that range). However, I noticed something very interesting. I used a previous trial-factoring to estimate that it would take about 1h10m to do each exponent to 60 bits (assuming no factor is found). However, even after setting the FactorOverride to 60 bits, the computer still tells me that each factoring assignment will take about 2 weeks. Does Prime95 not take altered FactorOverride settings into account when making its estimates?

[garo]

Have you set the UsePrimenet switch to 0 - if Prime95 is set to use primenet then it disregards the FactorOverride settings.

[ThomRuley]

Thanks, I'll be sure to check that. While we're at it, is there anything else that may be interfering with status reports?

[lycorn]

I used a previous trial-factoring to estimate that it would take about 1h10m to do each exponent to 60 bits (assuming no factor is found). However, even after setting the FactorOverride to 60 bits, the computer still tells me that each factoring assignment will take about 2 weeks.

There is something unclear in the above statement: if your computer takes 1h10 m to factor an exponent to 60 bits, it should take way longer to factor to the default limit (70 bits) for the 56.5 - 57M range: actually around 2000 times more. So, if it is not taking into account the 60 bit override, it should be giving you an estimated time of ~3 months and not 2 weeks... :?
God, I´m picky tonight ;)

[cheesehead]

Y'know, there's a reason they're called "estimates" instead of being called "gold-plated, iron-clad guarantees of exact elapsed time". :)

[lycorn]

Sure. But the estimates done by the program are usually pretty accurate. My point is that ThomRuley´s estimate must be somehow wrong. Note that 1h10m to factor to 60 bits is *very* slow. On a P90, it takes roughly 1 h to factor to 61 bits. So, unless he is using a K5 or something similar, or running some other heavy stuff in parallel with P95, I can´t figure out how he got those timings.

[ThomRuley]

Actually, I think my estimate may have been off by a bit. When I finally set all the switches correctly, the computer was giving me an estimate of about 5min per exponent. This will knock them out pretty quickly if the estimate is correct. 8)

[smh]

Originally Posted by lycorn

Sure. But the estimates done by the program are usually pretty accurate. My point is that ThomRuley´s estimate must be somehow wrong. Note that 1h10m to factor to 60 bits is *very* slow. On a P90, it takes roughly 1 h to factor to 61 bits. So, unless he is using a K5 or something similar, or running some other heavy stuff in parallel with P95, I can´t figure out how he got those timings.

The time it takes to factor an exponent to N bits depends on the exponent you try to factor.