False primes!

[MooooMoo]

During a hardware test, I was randomly testing some primes reported by PrimeGrid. Unfortunately, I found these to be composite:

1227950103*2^195000-1
1949806593*2^195000-1

This shouldn't have happened since all primes found by PrimeGrid were double-checked. Since this was a random check, I haven't checked all the primes reported by PrimeGrid. However, those results warrant a complete double-check of all PrimeGrid primes

[pacionet]

I remember that KEP had reported some false primes some time ago.

[Rytis]

These are from the ancient (hehe) times when primes weren't doublechecked. The two are from users Finbar and The Bandit. I have reduced their prime number count.

[edit] I'll just add that if anyone wants to doublecheck PG primes, I can zip up and send the logs of all LLR tests.

[KEP]

Originally Posted by pacionet

I remember that KEP had reported some false primes some time ago.

I sure did, but it all seems to be solved with the BOINC LLR, since the CPU there gets just a few seconds of 0% running, while loading the next assignment, and it is apparently enough to prevent overheating, despite the fact that my computer isn't overclocked, and barely gets over human temperature... Just wanted to let you all know that the course of my false primes apparently is overheating, and not any socalled hardware issue

So left is only to see how deep we have to go before we make it with our first prime to the top 5000 list

Regards!

[paulunderwood]

From: http://www.formfactors.org/developer...12V_PS_1_1.pdf

              Table 3. DC Output Voltage Regulation
   Output            Range            Min.          Nom.          Max.         Unit
           (1)
+12VDC                ±5%          +11.40         +12.00        +12.60         Volts
  +5VDC               ±5%            +4.75          +5.00        +5.25         Volts
+3.3VDC               ±5%            +3.14          +3.30        +3.47         Volts
   -5VDC            ±10%             -4.50          -5.00        -5.50         Volts
 -12VDC             ±10%            -10.80         -12.00       -13.20         Volts
  +5VSB               ±5%            +4.75          +5.00        +5.25         Volts
 (1)
     At +12 VDC peak loading, regulation at the +12 VDC output can go to ± 10%

You told me your PSU had these readings:

CPU Vcore 1.47 V
+3.3 V 3.15 V
+5.0 V 5.16 V
+12.0 V 11.90 V
5V(SB) 4.11 V
Voltage battery 3.15 V

Something is causing that low VSB reading -- most likely a bad Power Supply Unit (or something)

[KEP]

Originally Posted by paulunderwood

From: http://www.formfactors.org/developer...12V_PS_1_1.pdf

              Table 3. DC Output Voltage Regulation
   Output            Range            Min.          Nom.          Max.         Unit
           (1)
+12VDC                ±5%          +11.40         +12.00        +12.60         Volts
  +5VDC               ±5%            +4.75          +5.00        +5.25         Volts
+3.3VDC               ±5%            +3.14          +3.30        +3.47         Volts
   -5VDC            ±10%             -4.50          -5.00        -5.50         Volts
 -12VDC             ±10%            -10.80         -12.00       -13.20         Volts
  +5VSB               ±5%            +4.75          +5.00        +5.25         Volts
 (1)
     At +12 VDC peak loading, regulation at the +12 VDC output can go to ± 10%

You told me your PSU had these readings:



Something is causing that low VSB reading -- most likely a bad Power Supply Unit (or something)

It might be so, but it's just weird, if it is so, that there hasn't been any errors since switching to BOINC LLR, despite those, where the result is none, which yield an validation error, so it must go deeper than that, unless it has something to do with the fact, that the fan to cool the CPU doesn't has to work so hard because of the minute pauses between assignments, in which the CPU has a chance of cooling down without the need of additional power use. I don't know, it's just a theory!

Regards!

[smh]

I think it's time for you to run a prime95 torture test for a day or two

[KEP]

Originally Posted by smh

I think it's time for you to run a prime95 torture test for a day or two

It makes no sence, since it stops after the first test, and then concludes "somekind of hardware error", and that doesn't make me any more alert to what the problem is, so I'm at best left confused by using the Prime95 torture test, and to think that I should run it for 48 hours, first off it's not possible, since it stops as mentioned after the second test, and secondly, i wan't waste time on that, since there are lots of Distributed Computing projects, that my computer can still return perfectly valid results to, despite the apparent hardware error, so I think it also may have something to do with the construction of the LLR 3.7.1 program, since there hasn't been any problems, since I switched to BOINC LLR. Sorry for the long sentence...

I'm by the way glad that you decided not to leave the TPS project anyway smh, you're very welcome to support the project, and I guess that you may have considered me hostile, as well as I considered you to be hostile towards Rytis, because we have no visiuals of one an other, but since this is a new year, I'll offer you my hand as a good guesture, and hope to see ud co-exist on the message board, as well as various prime projects, as friends in stead of enemies. Life is simply to short to be enemies, and sadly not long enough, to accomplish all the prime goals that one can imagine...

Kind regards!

KEP

[Skligmund]

Sounds like you have some sort of mermory issue. How hot is your RAM getting when running the PRIME95 torture test? Can you hold it in your fingers, or is it too hot?

At the very least, I would suggest (if possible) to reduce your memory timings a little and see if that helps.

[MooooMoo]

I found another false prime by PrimeGrid:

1418514507*2^195000-1

[KEP]

Originally Posted by Skligmund

Sounds like you have some sort of mermory issue. How hot is your RAM getting when running the PRIME95 torture test? Can you hold it in your fingers, or is it too hot?

At the very least, I would suggest (if possible) to reduce your memory timings a little and see if that helps.

I've actually thought of the RAM myself, maybe I could try that, if you can tell me how to reduce the memory timings, I guess I could try to see if that helps, because if it is a faulty RAM, then it is easy to fix, but it is still not very much energy I'll put into the issue, since the computer is calculating with a very small error frequenzy, so it is not wort the effort after all. But now that I come think about it, it might actually be a RAM error, since, when I did work for BURP, sometimes the WU disappered from the RAM, and had to be forced start, before it could be completed.

Regards!