Prime 95

[OhHenry]

I OC'ed my computer and ran prime95 which gave me a round off error. My computer seems stable, but i'm curious to know if i can fixed that. Secondly, is there any other programs for me to find out the stablity of my system?

 

[nycxandy]

Increase the vcore.

Run Memtest86 to test memory.

 

[Shooters]

Round off error means that something isn't stable, either CPU or maybe RAM. That being said, Prime95 stresses your CPU more than any other benchmark and it's possible that it will fail yet your computer will be perfectly stable during normal use. That's why some people don't even use Prime95 to test their system because they say that it goes far beyond what most people will put their systems through in every day use. Also, I've heard of some people running Prime95 even with no overclock and getting round off errors, so go figure.

It's up to you to decide if you want to system to be able to handle Prime95 with no errors even if all other applications run stable. You can try bumping the voltage up a little or decreasing your overclock. Also, how long did did you have Prime95 running for before it gave you the error? I assume you were running the torture test? I would say that if you had it running for at least 24 hours then you should okay.

 

[VirtualLarry]

Originally posted by: Shooters
Round off error means that something isn't stable, either CPU or maybe RAM. That being said, Prime95 stresses your CPU more than any other benchmark and it's possible that it will fail yet your computer will be perfectly stable during normal use. That's why some people don't even use Prime95 to test their system because they say that it goes far beyond what most people will put their systems through in every day use. Also, I've heard of some people running Prime95 even with no overclock and getting round off errors, so go figure.

It's up to you to decide if you want to system to be able to handle Prime95 with no errors even if all other applications run stable. You can try bumping the voltage up a little or decreasing your overclock. Also, how long did did you have Prime95 running for before it gave you the error? I assume you were running the torture test? I would say that if you had it running for at least 24 hours then you should okay.

If an AMD system gives errors during the Prime95 "torture test" routines, then it's not stable, period. AMDs tend to fail in their FPUs first, and Prime95 stresses them heavily. The same cannot be said of P4's and Prime95, because P4's tend to fail in their double-clocked integer ALUs first. Therefore it is possible for a P4 to be "Prime95 stable", and actually still not be totally stable.

To be complete accurate though, it's theoretically possible for an Athlon XP's FPU execution unit to give errors, while the integer pipeline runs code seemingly unaffected, but personally I wouldn't trust that.

 

[Sonic587]

Originally posted by: nycxandy
Increase the vcore.

Run Memtest86 to test memory.

Also...specs???

 

[OhHenry]

Originally posted by: Shooters
Round off error means that something isn't stable, either CPU or maybe RAM. That being said, Prime95 stresses your CPU more than any other benchmark and it's possible that it will fail yet your computer will be perfectly stable during normal use. That's why some people don't even use Prime95 to test their system because they say that it goes far beyond what most people will put their systems through in every day use. Also, I've heard of some people running Prime95 even with no overclock and getting round off errors, so go figure.

It's up to you to decide if you want to system to be able to handle Prime95 with no errors even if all other applications run stable. You can try bumping the voltage up a little or decreasing your overclock. Also, how long did did you have Prime95 running for before it gave you the error? I assume you were running the torture test? I would say that if you had it running for at least 24 hours then you should okay.

The error popped up no less then 5 minutes or so... I oc'ed my 2.6c to 2.9ghz with 223fsb. I could increase my vcore but my cpu temp avgs around 49C with stock hsf. It seems kinda high from reading around the forums, and i'm not sure even if it is accurate. Also, i ran memtest85 and got errors in the address range test, does that mean that i have bad sectors in the ram or so?

 

[VirtualLarry]

Originally posted by: OhHenry

Originally posted by: Shooters
Round off error means that something isn't stable, either CPU or maybe RAM. That being said, Prime95 stresses your CPU more than any other benchmark and it's possible that it will fail yet your computer will be perfectly stable during normal use. That's why some people don't even use Prime95 to test their system because they say that it goes far beyond what most people will put their systems through in every day use. Also, I've heard of some people running Prime95 even with no overclock and getting round off errors, so go figure.

It's up to you to decide if you want to system to be able to handle Prime95 with no errors even if all other applications run stable. You can try bumping the voltage up a little or decreasing your overclock. Also, how long did did you have Prime95 running for before it gave you the error? I assume you were running the torture test? I would say that if you had it running for at least 24 hours then you should okay.

The error popped up no less then 5 minutes or so... I oc'ed my 2.6c to 2.9ghz with 223fsb. I could increase my vcore but my cpu temp avgs around 49C with stock hsf. It seems kinda high from reading around the forums, and i'm not sure even if it is accurate. Also, i ran memtest85 and got errors in the address range test, does that mean that i have bad sectors in the ram or so?

It means that your system is unstable.

What you should do first, is back off on all of the overclocks, and run Memtest86+ at stock speeds. That will give you a baseline on whether your RAM is bad or not.

If your mobo allows running the RAM faster (as fast as it might while the system was otherwise overclocked), but allows running the CPU at stock speeds, then do that as well, to test if the RAM will overclock.

Likewise, then test running the CPU at a faster speed, using Prime95 or other programs as appropriate. At that point if it fails, you know that it is most-likely the CPU.

 

[TheGrandHooHa]

Originally posted by: Shooters
Round off error means that something isn't stable, either CPU or maybe RAM. That being said, Prime95 stresses your CPU more than any other benchmark and it's possible that it will fail yet your computer will be perfectly stable during normal use. That's why some people don't even use Prime95 to test their system because they say that it goes far beyond what most people will put their systems through in every day use. Also, I've heard of some people running Prime95 even with no overclock and getting round off errors, so go figure.

This is my logic. I got rounding errors after 23 hours without any overclocking at all. Meanwhile, my system has not truly "crashed" once since I built it (August 2003), so who cares if Prime95 works or not? As long as it does what I need it to do, I don't need a fancy torture test.

That being said, I would go with what other said: remove the overclock, test the RAM, then put the overclock back and test the CPU.

 

[DarkKnight]

when ur using prime95, make sure your running it on priority 10 (it's in the readme somewhere), stresses the system much more than default. On my comp, when the system failed after 2 hrs on default, it failed 2 minutes on prriority 10. After, I clocked down a little, worked fine.

 

[VirtualLarry]

Originally posted by: TheGrandHooHa

Originally posted by: Shooters
Round off error means that something isn't stable, either CPU or maybe RAM. That being said, Prime95 stresses your CPU more than any other benchmark and it's possible that it will fail yet your computer will be perfectly stable during normal use. That's why some people don't even use Prime95 to test their system because they say that it goes far beyond what most people will put their systems through in every day use. Also, I've heard of some people running Prime95 even with no overclock and getting round off errors, so go figure.

This is my logic. I got rounding errors after 23 hours without any overclocking at all. Meanwhile, my system has not truly "crashed" once since I built it (August 2003), so who cares if Prime95 works or not? As long as it does what I need it to do, I don't need a fancy torture test.

Well, I sincerely hope that you will never do any serious/important financial or spreadsheet calculation on that machine; nor gov't or engineering work; nor participate in any distributed-computing projects, that could be hampered by your machine returning subtly-incorrect results. I would hate to see someone from NASA with your attitude towards system stability, that could prove disastrous.

 

[Delorian]

I got the same thing while OCing my barton 2500 and "the problem" never seemed to bother any other program, just Prime95. Perfectly stable otherwise. I used multiple stability testers other than Prime95 and none had a problem so I just ignored the error. The others I used were SiSoft Sandra, Memtest, CPU-burn in, SETI@home and a couple others I can't remember (all were free/shareware). Just leave 'em on for a little while and see if these cause errors, that should tell you whether or not the error was serious or just a small glitch that could happen regardless of OCing.

 

[Delorian]

Originally posted by: VirtualLarry

Originally posted by: TheGrandHooHa

Originally posted by: Shooters
Round off error means that something isn't stable, either CPU or maybe RAM. That being said, Prime95 stresses your CPU more than any other benchmark and it's possible that it will fail yet your computer will be perfectly stable during normal use. That's why some people don't even use Prime95 to test their system because they say that it goes far beyond what most people will put their systems through in every day use. Also, I've heard of some people running Prime95 even with no overclock and getting round off errors, so go figure.

This is my logic. I got rounding errors after 23 hours without any overclocking at all. Meanwhile, my system has not truly "crashed" once since I built it (August 2003), so who cares if Prime95 works or not? As long as it does what I need it to do, I don't need a fancy torture test.

Well, I sincerely hope that you will never do any serious/important financial or spreadsheet calculation on that machine; nor gov't or engineering work; nor participate in any distributed-computing projects, that could be hampered by your machine returning subtly-incorrect results. I would hate to see someone from NASA with your attitude towards system stability, that could prove disastrous.

Yeah but I don't think NASA would rely on Prime95 to test thier supercomputers and why would anyone doing that kind of work worry about speed driving thier over-the top expensive PCs anyhow? I've never heard of any kind of firm or laboratory or such trying to run any equipment over-spec as far as PC equipment goes. If it's not fast enough for them, they buy something new.