Have I submitted bad data? (Or, well, about to anyway)

Nazo · 2005-08-04, 09:28

Uhm, I have to be honest. When I got my PC-3200 memory, I though yay and such and tried putting my processor at 200x12.5, which worked fine before a long time ago when I ran the memory desynced. I just assumed it'd still be fine... That one was a definite mistake on my part because I just assumed it'd be stable where it was before. My room must be too much warmer these days than it was then I guess. Sad thing is, I even edited the INI to report my new processor speed during the times I attempted 2.5GHz, so the record will look a bit weird when they see that I tried it on a 2250MHz (180x12.5, crappy old memory held me back,) then 2500, then 2300 and probably a few jumps back and forth as I tried various configurations. In the end, the only thing truly stable is 200x11.5, which hasn't reported any errors yet in the weeks I've been running it. What's worse is that I tried various memory timings, which passed initial stress testing (I didn't give them long enough clearly) and so probably caused even more errors than the CPU problems which usually resulted in a reboot I think.

I understand the whole thing is supposed to have some double checking set up, but, do I need to try to find a way to start this whole thing over? I just don't know that I could stand it though, I'm ~74% through a very very long prime here. Took me a couple of months or so because sometimes I can't leave it running all the time (though I now am doing this more often now that I have a second system -- a dual p3-500, so, despite being dual, probably of little real use to primenet anyway, so don't get any ideas, besides, I reboot it more often than I blink in a day -- to play around with.)

Do I need to find a way to just start the whole process over? In which case, it might be better just to pass it on because I just don't know that I have the patience to start this thing over. I'm thinking future projects I select are going to all be the short ones that can finish in less than a month...

garo · 2005-08-04, 10:23

Well, you could try factoring of exponents because it takes a lot less time.
Prime95 does not report CPU speed when it reports the result but it does report any errors. There is a small - very small really - about 25% or so chance that despite the errors your test will turn out okay. So since you are at 74% I woudl recommend continuing and submitting the result.

Nazo · 2005-08-04, 21:21

Whether or not PrimeNet actually stores the information, Prime95 has it in the configuration file for whatever reason. It's in the "local.ini" file. I had to manually set my CPU type and speed for some reason, and PrimeNet does know your CPU type even if it doesn't know the speed, and that's not to say that it doesn't have it in the data, just simply doesn't show it to you since the script is optomized for setups with a bunch of computers rather than just one. I guess Prime95 had trouble detecting that my "Unknown CPU Typ" (not a typo, that's what this braindead bios calls it) is still an Athlon XP.

Anyway, considering this happened already in the iterations, roughly around 50% or so, (it's just trying to find factors now, right?) I felt like it didn't necessarily do any damage, that maybe the double checking caught it. That was the impression I had anyway.

BTW, just for reference, the possible prime I'm working on is M34547803.

lycorn · 2005-08-05, 19:01

Quote:

Originally Posted by Nazo

(it's just trying to find factors now, right?)

No. It is still running the Lucas-Lehmer test. The test has to run all the way to 100% to determine whether the number is prime. But I agree with garo, just let it finish, and somebody in the future will double check it.
And yes, if you own an Athlon XP and prefer shorter tasks, I recommend that you do Trial Factoring. Athlon XPs are much better for TF than for Lucas-Lehmer tests. For the range of numbers that is currently being handed out by the server (~30.7M), your system should take no more than 2 days per exponent. You could also do double checks; although they are Lucas-Lehmer tests, the numbers tested are much smaller so the time taken is much less. But Trial Factoring is the most suitable.
Anyway, it is your system, and any type of contribution is more than welcome. Do as you enjoy it the most.
Good luck!

Nazo · 2005-08-05, 19:53

2 days sounds good to me. I'm going with that one. The main thing is that it just bugged me seeing this thing still going months down the road. I understood it would take a while, but, I have to admit I was surprised as to how long it took. I also was uncomfortable with just how hard I was pushing my system all this time. I've decided I'd rather use something where I can run Prime95 less frequently, which will be a lot easier when I have definite results to go on.

Thanks for the info. I've set it to pick factoring in the future.

garo · 2005-08-06, 05:44

Prime95 is heavily optimized code so it pushes your CPU like almost nothing else can. Trial Factoring is also easier on your CPU - runs cooler and less sensitive to errors - so you will be happier all round.

2005-08-04, 09:28	#1
Nazo Aug 2005 Alabama, USA 25₈ Posts	Have I submitted bad data? (Or, well, about to anyway) Uhm, I have to be honest. When I got my PC-3200 memory, I though yay and such and tried putting my processor at 200x12.5, which worked fine before a long time ago when I ran the memory desynced. I just assumed it'd still be fine... That one was a definite mistake on my part because I just assumed it'd be stable where it was before. My room must be too much warmer these days than it was then I guess. Sad thing is, I even edited the INI to report my new processor speed during the times I attempted 2.5GHz, so the record will look a bit weird when they see that I tried it on a 2250MHz (180x12.5, crappy old memory held me back,) then 2500, then 2300 and probably a few jumps back and forth as I tried various configurations. In the end, the only thing truly stable is 200x11.5, which hasn't reported any errors yet in the weeks I've been running it. What's worse is that I tried various memory timings, which passed initial stress testing (I didn't give them long enough clearly) and so probably caused even more errors than the CPU problems which usually resulted in a reboot I think. I understand the whole thing is supposed to have some double checking set up, but, do I need to try to find a way to start this whole thing over? I just don't know that I could stand it though, I'm ~74% through a very very long prime here. Took me a couple of months or so because sometimes I can't leave it running all the time (though I now am doing this more often now that I have a second system -- a dual p3-500, so, despite being dual, probably of little real use to primenet anyway, so don't get any ideas, besides, I reboot it more often than I blink in a day -- to play around with.) Do I need to find a way to just start the whole process over? In which case, it might be better just to pass it on because I just don't know that I have the patience to start this thing over. I'm thinking future projects I select are going to all be the short ones that can finish in less than a month... Last fiddled with by Nazo on 2005-08-04 at 09:28

2005-08-04, 21:21	#3
Nazo Aug 2005 Alabama, USA 3×7 Posts	Whether or not PrimeNet actually stores the information, Prime95 has it in the configuration file for whatever reason. It's in the "local.ini" file. I had to manually set my CPU type and speed for some reason, and PrimeNet does know your CPU type even if it doesn't know the speed, and that's not to say that it doesn't have it in the data, just simply doesn't show it to you since the script is optomized for setups with a bunch of computers rather than just one. I guess Prime95 had trouble detecting that my "Unknown CPU Typ" (not a typo, that's what this braindead bios calls it) is still an Athlon XP. Anyway, considering this happened already in the iterations, roughly around 50% or so, (it's just trying to find factors now, right?) I felt like it didn't necessarily do any damage, that maybe the double checking caught it. That was the impression I had anyway. BTW, just for reference, the possible prime I'm working on is M34547803. Last fiddled with by Nazo on 2005-08-04 at 21:25

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Corrupted data?	Oops!	Information & Answers	2	2013-10-22 03:48
GPU TF vs DC/LL data	bcp19	GPU to 72	0	2011-12-02 16:41
Manually submitted P-1 results give more credit.	Mr. P-1	PrimeNet	1	2011-10-24 14:17
New paper from Wieb Bosma submitted to arXiv.org	schickel	Aliquot Sequences	4	2010-01-05 18:37
Data available?	Prime95	LMH > 100M	10	2007-06-22 23:55

2005-08-04, 10:23	#2
garo Aug 2002 Termonfeckin, IE 2⁴×173 Posts	Well, you could try factoring of exponents because it takes a lot less time. Prime95 does not report CPU speed when it reports the result but it does report any errors. There is a small - very small really - about 25% or so chance that despite the errors your test will turn out okay. So since you are at 74% I woudl recommend continuing and submitting the result.

2005-08-05, 19:53	#5
Nazo Aug 2005 Alabama, USA 3×7 Posts	2 days sounds good to me. I'm going with that one. The main thing is that it just bugged me seeing this thing still going months down the road. I understood it would take a while, but, I have to admit I was surprised as to how long it took. I also was uncomfortable with just how hard I was pushing my system all this time. I've decided I'd rather use something where I can run Prime95 less frequently, which will be a lot easier when I have definite results to go on. Thanks for the info. I've set it to pick factoring in the future.

2005-08-06, 05:44	#6
garo Aug 2002 Termonfeckin, IE 2768₁₀ Posts	Prime95 is heavily optimized code so it pushes your CPU like almost nothing else can. Trial Factoring is also easier on your CPU - runs cooler and less sensitive to errors - so you will be happier all round.