I might have delivered some bad results

[iconized]

Originally Posted by Uncwilly

Norway = cold + cheap hydropower = Norwegians love PC space heaters.

Here it is choosing between Russian gas or French nuclear energy, the Russian gas is cheaper, for heating that is.

Meanwhile all reruns I have done seem to be correct, just checking the results after my last factor makes more sense to me. But if people rerun my tests I have to help as well.

[Dubslow]

It seems most of us are convinced the results are good. Are there any who object to stopping this double-check project? (I only have a few hours before I'm away from my comp for 9 days, so please reply quickly)

[chalsall]

Originally Posted by Dubslow

It seems most of us are convinced the results are good. Are there any who object to stopping this double-check project? (I only have a few hours before I'm away from my comp for 9 days, so please reply quickly)

As I have found in politics, few will publicly agree with you even if they privately agree. Something about "it's better to remain silent and be thought a fool, than open your mouth and prove it".

KyleAskine, above, makes a very compelling mathematical argument as to why this double checking of TF work should not be done.

He ain't no fool....

[Dubslow]

I should hope that our discussions are not alike to politics. Everything here supports this decision, as far as I can tell; as Kyle writes, the effort isn't really worth it even if they are wrong, there were the statistically correct number of factors found, the effort isn't worth and no reruns have found factors, so it seems as if they were right. Since I'm rather on a limit, if no one objects by then, then I will terminate the project (with iconized's blessing, but I'm interpreting his last sentence to mean that).

[iconized]

You have my blessing and thanks for helping me out!

[kladner]

Originally Posted by Christenson

I'm all wet tonight...working through the thunder.... Kladner, time to find out where the heat from your PC is leaking out of the house/apartment! Also, multiply the watts by the hours per month and divide by 1,000 to get KWH/month, and figure that on your electric bill. Then figure out what the heat costs, and what part of the heat is replaced by the heat from the PC, and cost that out and subtract, and you have a good real cost of running GIMPS.

I don't have many options as renter for weatherizing. This 3-flat was built sometime in the 1920s, though it has some updates like double-pane windows and forced air gas heat instead of in-wall space heaters. It seems pretty clear that the insulation in the walls and ceiling (this is the top floor) is pretty poor. At least we get the heat coming up from the apartments below.

As to the cost per time period, this model, Kill a Watt EZ does all those tricks for you. The two computers have been running through it for a bit over 75 hours. The main computer (1090T@3.5GHz) gets shut down when I go to bed. The headless Ubuntu box (Opteron 180, dual core, 2.4GHz) runs all the time doing P-1, but by itself that's only 185 watts.

As the 1090T comes online it looks like this (including the 185w):
Idle:           300w
+monitor:    390w
+1x mfaktc: 515w
+2x mfaktc  560w
+3x mfaktc  585w
+P95 3wrkrs 645w

At $0.07733/kwh, with the above schedule, it breaks down like this:
$0.03/hr
$0.83/day
$5.82/week
$24.94/month
$303/year

My rational for shutting down the 1090T at night is that I'm not sitting here to appreciate the heat it puts out. It is also by far the highest drain since it's feeding the GPU and a much hungrier CPU. On the other hand, I'm only reducing costs by $75/year. I should try a days with everything shut down at night.

@uncwilly -Illinois has more nukes than any other state. Coal and gas plants only provide peak power supplements. In your situation I might have a more liberal attitude about running 24/7 with everything.

@lycorn -That is the rationale I've been using. These boxes are space heaters which do calculations on the side. It's just that we turn the heat way down at night and rely on blankets which have no operating costs.

EDIT: If it weren't for these painful economic realities, I'd have my partner's i7 920 doing P95, too.

[Dubslow]

The time of reckoning has come. This project is henceforth shut down. My computer will soon be disconnected from the internet for the next week (for Thanksgiving break) however fortunately I did receive permission to run it over break :). Either way, I won't be handing out any more assignments, and I will be removing all such assignments from my worktodo.txt. And now I can join GPU to 72 in earnest!

[Uncwilly]

Originally Posted by Uncwilly

Norway = cold + cheap hydropower = Norwegians love PC space heaters.

Originally Posted by lycorn

Why you mention Norway? Is it the "Somewhere you aren´t" location, by any chance?

No, I am not in Norway. I have visited Norway on a holiday of about 3 weeks. I have seen at least one forum member from N bring up this up before. Also, Norway seems to 'punch above their weight' in GIMPS. And is closer to Chi-town in weather than Lisboa.

Originally Posted by kladner

@uncwilly -Illinois has more nukes than any other state. Coal and gas plants only provide peak power supplements. In your situation I might have a more liberal attitude about running 24/7 with everything.

My personal situation is different. I have opted in for 'all green power' from my local utility, it is all wind power. Even with my laptop running Prime95 24/7 I am still in the lowest 10% of power usage for those similarly situated (at least until last period, which I went to 25% level [ran both an AC and an electric heating element at different times during the period]).

[Uncwilly]

Originally Posted by iconized

Meanwhile all reruns I have done seem to be correct, just checking the results after my last factor makes more sense to me.

Originally Posted by Dubslow

It seems most of us are convinced the results are good. Are there any who object to stopping this double-check project?

Originally Posted by chalsall

KyleAskine, above, makes a very compelling mathematical argument as to why this double checking of TF work should not be done.

Originally Posted by iconized

You have my blessing and thanks for helping me out!

Originally Posted by Dubslow

The time of reckoning has come. This project is henceforth shut down.

I concur Dr.

[Christenson]

So concur I, too....having been the first to ask about the statistics...

[LaurV]

I will finish my assignments. I got 100 from Dubslow, most of them from 70 to 71 bits, some of them from 68 or 69 to 71. First 6 I reported were only inside of the bit-ranges, and I got curses from the server (error 40 - result already known). So I decided for the rest of them to extend one bit, to go to 72, so I could also get some credit.

Up to now, over 70 exponents finished to 72 bits. No factor found neither inside of iconized's range (which was properly double checked) and neither from 71 to 72 bits. For the last I got credit, but not for all. As I did this without any reservation, about 12 times happened that the exponent was taken to 72 already by someone else (maybe GPU272, or other user) who most probably reserved the exponent properly. So it came to my mind that such extension is also not good, because there is a high probability I am doubling someone else's work, and only by luck (and very fast hardware) I succeeded to beat them to the credit.

So, in short, I will finish all that is left (about 26 expos) to 72 bits, and I am not going to request more. Switching to GPU272, where at list the expos are properly reserved and I am not duplicating anybody else's efforts.

With no factor found in 100 expos (about 5% from all) and with results from the other participants, I can assume iconized's tests were right.