SUMOUT error

[Lieven]

Hello,


I'm running version 23.7.1 and I'm getting the message:
SUMOUT error occured
Waiting five minutes before restarting.

Anyone an idea what's wrong?

Thanks!

[Cruncher]

<begin stock GP2 reply>

Well, sorry to be the bearer of bad news, but this indicates that there is a problem with your hardware, possibly with memory. Do you overclock your machine at all? If so, you might be overheating when Prime95 is running.

A couple of things you could try are running a motherboard monitoring program to see what temperature you're running at, and run Memtest86 overnight on your machine and see what you get.

good luck

<end stock GP2 reply>

-sr-

[Edit: since you're crediting me with the reply, I might as well edit it to fix the link to Memtest86.]

[GP2]

If I'm not mistaken, the SUMOUT errors are
the ones that produce error code 00XX0000, no?

Usually, these are relatively harmless... you should still have a 95% chance of producing a good result.

Are you also getting other types of errors?

[Lieven]

I don't get an error code at all or I don't know where to find it ;)

Another error, during the torture test was:

FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file

And a third one when LL test started:

UID: S152637/CE40C47ED, M22352293 completed P-1, B1=250000, B2=5062500, WZ1: 42316F7E
Iteration: 1030/22352293, ERROR: SUM(INPUTS) != SUM(OUTPUTS), 320871946680683.7 != 320871946677405.8
Possible hardware failure, consult the readme.txt file

Opening my case seemed to resolve the problem, since i'm running prime95 for more than 24 hours at the moment whitout any errors at all. So it should be an overheating problem.

This is very strange cause i'm not overclocking my system

My system:

ASUS A7N8X-Deluxe rev. 2.00
AMD Athlon XP2600+ 333MHz Barton @1913MHz + stock cooler
Corsair 512MB PC2700LL TWINX
Seagate Barracuda 40GB 9ms
Lite-On JLMS XJ-HD166S DVD-ROM
Hercules 3d Prophet Radeon 8500 128 MB
Iiyama HM903DT A
Windows XP SP1

I guess I need some extra cooling to run a stable system?

[nfortino]

Originally posted by GP2
If I'm not mistaken, the SUMOUT errors are
the ones that produce error code 00XX0000, no?

Usually, these are relatively harmless... you should still have a 95% chance of producing a good result.

Are you also getting other types of errors?

Yes, you are correct. I dug up George's post describing what all the error codes mean http://www.mersenneforum.org/showthr...120&#post12120

Originally posted by Lieven
don't get an error code at all or I don't know where to find it ;)

Another error, during the torture test was:

FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file

And a third one when LL test started:

UID: S152637/CE40C47ED, M22352293 completed P-1, B1=250000, B2=5062500, WZ1: 42316F7E
Iteration: 1030/22352293, ERROR: SUM(INPUTS) != SUM(OUTPUTS), 320871946680683.7 != 320871946677405.8
Possible hardware failure, consult the readme.txt file

You don't get an error code until you finish a test. Unfortunately, the rounding and the SUM(INPUTS)!=SUM(OUTPUTS) errors are considered harmful errors. You should run Memtest, and consider running double checks for now.

[Lieven]

Originally posted by nfortino

You don't get an error code until you finish a test. Unfortunately, the rounding and the SUM(INPUTS)!=SUM(OUTPUTS) errors are considered harmful errors. You should run Memtest, and consider running double checks for now. [/B]

Do I have to run double checks also if I've never had any error messages since I add some extra cooling? This means iteration is 4.10% completed and ran more than 30 hours whitout errors and started from before the first error (iteration 1030).

[nfortino]

Originally posted by Lieven
Do I have to run double checks also if I've never had any error messages since I add some extra cooling? This means iteration is 4.10% completed and ran more than 30 hours whitout errors and started from before the first error (iteration 1030).

You don't have to do anything. If you restarted from the beginning, and you have no errors, you will likely come out with a good result. However, it would still be a good idea to run 2 or 3 double checks, and confirm they are good. If they all return a matching result, you can be confident you have fixed your machine. (Note: you do not have to loose the progress you have already completed. Once Primenet gives you the double checks, close prime95, and put the double checks above the current exponent in worktodo.ini) This is a good idea for any machine, as it is essentially a torture test that helps the project, and that you get credit for.

[Lieven]

Originally posted by nfortino
(Note: you do not have to loose the progress you have already completed. Once Primenet gives you the double checks, close prime95, and put the double checks above the current exponent in worktodo.ini) This is a good idea for any machine, as it is essentially a torture test that helps the project, and that you get credit for.

How exactly do I get double checks?

[nfortino]

Originally posted by Lieven
How exactly do I get double checks?

Open the program, and under the test menu, select Primenet. In the dialogue box, uncheck take whatever work makes the most sense, and select request exponents for double check. Then, at the top of the dialogue box, change the number of days of work so that primenet gives you an exponent or two. In your case, I think that is about 50-55 days (possibly less for one, more for two). Press Ok, and see if it gives you the number of double checks you want. Once you have the exponents you want, go back and restore the old settings so you won't have to later. Then, quit the program, and follow the instructions above.

[QuintLeo]

If opening the case ends the errors, you need additional case cooling.

Most modern machines, especially "big name" stuff, tends to design around "average usage" on cooling, with a LITTLE margin. Prime pushes the CPU 100% all the time, and also pushes the memory and motherboard's North Bridge a LOT harder than normal - this is the reason it's widely used for TESTING and burn-in of computers.

There are a lot of cases out there that have really POOR thermal design.


As for me, I've VERY picky about cooling - my machines do NOT for the most part run in an air-conditioned environment (I have so many of them that I overload the A/C), so I make sure that they have GOOD case and CPU cooling.