Weird Game and Prime 95 problems, may it be Hardware?
 

Yo guys, hope we can share knowledge about hardware stuff here ins this nice forum. I play Unreal Tournament 2004 a lot and the game gives me this error after around 3 minutes of play:

UT2004 Build UT2004_Build_[2004-11-11_10.48]

OS: Windows XP 5.1 (Build: 2600)
CPU: AuthenticAMD PentiumPro-class processor @ 2107 MHz with 510MB RAM
Video: RADEON 9600 Series (Omega 2.5.97) (6497)

General protection fault!

History: ParticleIterator::Update <- AxEmitter::Tick <- TickAllActors <- ULevel::Tick <- (NetMode=0) <- TickLevel <- UGameEngine::Tick <- Level Rrajigar <- UpdateWorld <- MainLoop <- FMallocWindows::Free <- FMallocWindows::Realloc <- 676F4C57 0 FArray <- FArray::Realloc <- 0*2 <- FMallocWindows::Free

I know that Unreal 2k4 issues are not the main concern here, but I think that more people may have problems like mine, so I hope that solving this can help a lot of people (like myself =P). Well, lets cut off the obvious things, I have already have the last game patch,direct x, drivers and windows updates (except SP2, which sucks hard), I´m not overclocking any processor nor memory settings. I have already checked my voltage and temperature settings and they are stable even under heavy loads. I must say that I have already ran many stress tests and benchmarks (like Sandra and 3DMark) and my system never failed, not even a glitch... Except fot UT 2k4 and this Prime95 program. I heard of it here and downloaded it to figure out that this program is not a stress tester, but a program to find some kind of particular prime number and send to a research center... But I used it to test and it gave me 2 messages: "Error: Round is 0.5, expected less than 0.4" for the lighter test and "Illegal Sumeout error: Possible Hardware Failure" for the heavier one; both less than 2 minutes after the test began. I have already underclocked memory to see if it helps, but it was useless. I heard that AGP aperture size bigger than 128 MB and AGP fastwrite might cause system to be unstable so I turned both off; but it didn´t work for me. I must repeat, UT 2k4 and Prime 95 are the only programs that present problems, I have already ran many other games as well as "CPU Stability Test" to test the CPU and "Memoy Gold" to test memory, and they were both aproved. I wonder if it has something to do with my BIOS settings and hope that someone can help me, because I really like UT 2k4.

These are my modest system specs:

Ahtlon XP Barton 2800+ FSB 333 (166x2) running at ~2100 mhz
One 512 MB Spektek Memory Stick DDR 333 mhz
80 GB Western Digital IDE Har Drive
Abit KV7 (KT 600 chipset) Mobo
ATI 9600 PRO 256 DDR VRAM (with latest Omega Drivers)
Windows XP PRO SP1

Thanks in Advance, guys!!!

you are correct prime 95 is not really a stress tester but it does load your system so heavly that it sometimes finds problems others cant. the errors when doing a test mean your number didnt match the correct number check for dust in the system clean it out with compressed air reseat ram and cables... get the idea sometimes overheating so check your system temp.

I will try cleaning stuf... But the I have already checked system temerature and voltage, they are fine....

Sorry, but I don´t get it, why dows prime95 get me the error "Rounding was 0.5, expectted less than 0.4" ? It seems a lot of people are getting this damn error, isnnt it maybe a program bug?

No, it is not a program bug. It is usually a result of:

1) Overclocking causing system unstability.
2) bad audio/video drivers
3) Bad cooling in the system causing overheating.

[QUOTE=Arthanis]Sorry, but I don´t get it, why dows prime95 get me the error "Rounding was 0.5, expectted less than 0.4" ? It seems a lot of people are getting this damn error, isnnt it maybe a program bug?[/QUOTE]While Prime95 is under development, the rounding test is used to fine-tune internal program constants (such as what FFT length to use for exponents in a certain range). But once the development version has run successfully on properly-functioning hardware without any of those rounding errors, the internal program constants are "locked-in" and never changed between any two runs of the released production version.

[i]Then as long as an uncorrupted copy of the Prime95 software is run with the same parameters (same type of test, same exponent values, same hardware, ...), it always executes [b]exactly the same sequence of instructions[/b] on each run.[/i] So if the final arithmetic result differs (as indicated, for instance. by the rrounding error message) from the stored expected values, the reason must be a hardware problem.

Q: Why do lots of people (who are having problems) get the rounding error message?

A: Because that's the most common type of error that will be detected for a wide range of hardware failures. The rounding test is very sensitive to errors that change even one bit during the long sequence of calculations. Also, that test happens to be the first test executed by Prime 95 in many situations, so that's the error reported if something's wrong.

2) bad audio/video drivers

This could be a bug in the software.
Not nessesarily that of the audio/video.



Respectfully,
TTn

[QUOTE]Windows XP PRO SP1[/QUOTE]

This is also suspect. (SP1)

[QUOTE=TTn]2) bad audio/video drivers

This could be a bug in the software.
Not nessesarily that of the audio/video.



Respectfully,
TTn[/QUOTE]Yes, it could be. Except that a significant number of those who have seen this error have fixed it by updating or disabling said drivers.

Re: Driver/OS influences
 

As [b]Xyzzy[/b] said in the software forum:

[QUOTE]Use the mprime floppy to minimize OS/driver variables:
[url="http://www.mersenneforum.org/zip/"]http://www.mersenneforum.org/zip/[/url][/QUOTE]

Oh, I i didn´t think that bad drivers could cause errors on prime95 stress tests, since its low-level test, very close to hardware layer. But I discovered something, my PC was crashing in other aplications too, or they reboot my machine or they just close in my face... Then I did the following: reset my BIOS and didn´t make any change on it, so now my 2800+ Barton is running as a 2500+ ... Then now it passes on every test... But processor is really the Barton 2800, so it´s not overclocked, I checked it´s ID and it is AXDA2800DKV4D , wich means its really a 2800 ... What could that be? How do I solve my problem, since I bought a 2800+, and I don´t want to underclock it? Could thiis be memory? Thanks for the help, guys, you are the best!!!!