What is considered stable when OCing a cpu

[perdomot]

When I had OC'd my E4400 to 3Ghz, it was able to pass various stress tests like Orthos and Memtest for hours and yet during daily usage, I would have Explorer crash on me with the taskbar disappearing for a few seconds before returning. Other times, I would have a message about Dr. Watson popping up. Would these things constitute an unstable system or are they just things one has to put up with in order to OC? Curious on how people feel about this.

 

[nerp]

Sounds like an unstable system to me.

To me, stable means being able to run ORTHOS 24/7 while encoding XviDs in the background and being able to poke around explorer with no crashes, ever. I don't let my machines exhibit ANY symptoms of instability like you describe or consider a system with explorer crashes stable. To me, an explorer crash and Dr. Watson messages regularly tells me that something is wrong and either I need to adjust voltage or reduce my overclock somewhat. My primary concern in your shoes would be if there is the chance for data corruption. If explorer crashes, I wouldn't be surprised if I was going to be dealing with corrupted files later.

 

[Gillbot]

If there are any crashes for any reason, it's obviously not stable. IMHO, none of the various stability tests are 100% reliable either. Just because you can run prime all night doesn;t mean you'll never crash in a game or some other application.

To me, if my rig will run 24/7 and do everything I ask it to without crashing, it's stable enough for me.

 

[VirtualLarry]

You may have corrupted windows while you were overclocking. Try a reinstall.

 

[perdomot]

The odd thing is that these programs which stress the system are supposed to be more intense than most apps users run and yet a rig can complete them without a problem and still develop errors like I described above. I started to get the feeling that most OCers put up with these annoyances from time to time as the price they pay for OCing.

 

[Deadtrees]

If you run certain tasks that utilizes lots of CPU power for a long time(games, encoding...etc), the computer shouldn't have any issues during those periods.
In those cases, I'd run prime, occt, and whatever programs for a long time to ensure that the system is rock soild for a long time.

If you are like me...if you only use general programs that don't utilize much of CPU power and only need high CPU power for a short amount of time(like Photoshop and lightroom), you shouldn't worry much about stability issues.

For instance, My Q6600 @ 3.2, 1.216 Cpu-z reported Core-Voltage runs fine as long as I don't stress the CPU more than a few hours. I know I can't claim that this computer is perfectly solid as it'd crash after a few hours of heavy usage but for what I do, it's rock solid and I'm fine with it.

P.S: In your case, that's very and very unstable system you got there.

 

[stuff311]

Originally posted by: VirtualLarry
You may have corrupted windows while you were overclocking. Try a reinstall.

:thumbsup:

Doesn't matter how much a program stresses your cpu, if it is corrupt, it will crash every time.

 

[perdomot]

I installed XP under stock conditions and made an image for quick restoration as needed. I reinstalled the image several times but eventually ended up with problems. I can't figure how you can run stress tests for hours that pass and yet simple internet surfing,picture viewing and music playback cause errors.

 

[Modelworks]

Originally posted by: perdomot
I installed XP under stock conditions and made an image for quick restoration as needed. I reinstalled the image several times but eventually ended up with problems. I can't figure how you can run stress tests for hours that pass and yet simple internet surfing,picture viewing and music playback cause errors.

If the computer can run a memtest for hours and hours and then run orthos for hours you have eliminated two items . You still cannot account for the various drivers that load. The software and services running in the background. The plugins that IE or Firefox loads, etc.

Sometimes the crashes have absolutely nothing to do with overclocking.

My personal stress test is to render a 12Million polygon scene in 3dsmax @1080P.
On a 4GB machine it causes constant paging of the hard drive since 3dsmax uses about 6GB to render the scene using Gelato , so I can stress the GPU as well. So all at one time I stress the cpu, ram, gpu, hard drive. If I run that for several hours and the system is still intact, I know the system is stable. Too bad 3dsmax isn't free , maybe someone could come up with such a test using blender.

 

[VirtualLarry]

Yeah, it sounds like it could be drivers.

 

[Drsignguy]

Ok so you have installed the OS under stock levels, How about your drivers?as "Modelworks" and "VirtualLarry" have stated, I think that sounds about right.

 

[perdomot]

I had two different mobos with my E4400, an intel and a gigabyte, and both developed similar issues. The Phenom X3 I'm currently running has not had any issues similar to the other rig. Perhaps it was a combination of software and hardware. Seems like quite a few of the P35 mobos suffered issues like double booting and cold booting issues. Haven't heard any issues yet from the P45s but its early.

 

[mindless1]

Maybe I missed it, but did anyone mention the age-old test of un-overclocking the instable system, using it like that for awhile to see if you still get errors, crashes, etc? Granted, if you have file corruption you would have to restore the image again but the basic idea is use the system longer while it isn't o'c.

If it runs fine non-o'c, keep using it non-o'c. Next, adjust one thing, whether it be memory timing or FSB or whatever and use the system for awhile again. Thoroughly test each change instead of jumping from non-o'c to fully tweaked o'c that may be instable.

 

[cusideabelincoln]

I don't think you missed it, mindless. You seem to be the first person to say he should un-overclock after experiencing instability after overclocking. And I agree. Perdomot, you should have returned your E4400 to stock settings after you experienced system instability. If the system doesn't crash, then the overclock was bad. But if the system still crashed, then you have some other problem.

 

[perdomot]

Actually, I did return it to stock speeds figuring I would make do with the default speeds but eventually errors creeped up again. I thought it might be some hardware issue so I changed the mobo but the new one had issues to. Seems like there was always something going on that was a problem. Ram was tested and passed so I was at a loss as to the cause. I decided to try an AMD rig after my vacation and so far so good although it is slower than an intel rig.

 

[cusideabelincoln]

Well it seems like you most likely have software conflicts.

 

[Drsignguy]

Originally posted by: cusideabelincoln
I don't think you missed it, mindless. You seem to be the first person to say he should un-overclock after experiencing instability after overclocking. And I agree. Perdomot, you should have returned your E4400 to stock settings after you experienced system instability. If the system doesn't crash, then the overclock was bad. But if the system still crashed, then you have some other problem.

This is something I had to do. I was stable using the older F@h clients, but as soon as I ran the Smp clients, my main system crashed. The new clients are much more memory intensive and of course cpu intensive. So, I have been doing some more tweaking to get my stability back by bumping up the voltage and making sure my timings are in order.

By thinking your "orthos", occt,or prime stable, doesn't guarantee anything.

 

[racolvin]

I gave up caring My system runs just fine while I'm playing the games I ask it to but when I run Prime95 I get a rounding error on Core2 or Core3 after about 10 minutes on that test that produces the most heat. It drove me bananas for a while and I played with voltages, etc. When no matter what I did in terms of core voltage solved the problem yet my games still ran fine, I quit caring and just played

 

[perdomot]

Originally posted by: cusideabelincoln
Well it seems like you most likely have software conflicts.

Wouldn't a software conflict on an intel system cause the same conflict on an AMD rig?

 

[Quiksilver]

Perhaps your hard drive has gone bad? Especially if it still unstable after returning to stocks clocks and re-formatting.

 

[taltamir]

I have a new benchmark for stability. Get firefox 3, open many tabs (10 windows with 10 tabs each). Set it to SAVE open tabs when exiting and reload them on next lunch...

Run OCCTv2 one mixed CPU + RAM (tests CPU and RAM at the same time! it will not work with just CPU test). When OCCT is running open FF and watch it try to load all those pages at once.

"perfectly stable" systems will blue screen every time i did this, until the voltage was raised enough.

PS. Stable means 0 problems, none, whatsoever.

 

[Drsignguy]

Originally posted by: taltamir
I have a new benchmark for stability. Get firefox 3, open many tabs (10 windows with 10 tabs each). Set it to SAVE open tabs when exiting and reload them on next lunch...

Run OCCTv2 one mixed CPU + RAM (tests CPU and RAM at the same time! it will not work with just CPU test). When OCCT is running open FF and watch it try to load all those pages at once.

"perfectly stable" systems will blue screen every time i did this, until the voltage was raised enough.

PS. Stable means 0 problems, none, whatsoever.

Absolutly correct!

I had to do some tweaking today and I am pretty G'damn sure I got it right this time. On my main rig, 8Gb of ram puts a strain on the system and I had to bump up the memory voltage and the NB voltage a bit more to get it stable. How do I know its stable now? 2 smp clients, and 1 gpu client and its crunching on. No glithces, drips or errors. Temps are really good too.



So, what taltamir is basically suggesting, load the hell out of it and if she holds, all is good!

 

[Insomniator]

I don't know, I bet many stock systems including prebuilts would not be considered stable going by the 0 problems whatsoever criteria. Systems crash for various reasons stock or not and personally I do not feel like reinstalling
windows or playing with my bios every time I have a problem that may or may not be caused by anything I can control.

I will say however if that normal surfing and movie watching crashes the system with any bit of regularity then there probably is a problem that can be fixed. I got a BSOD while playing source (stock or not) once in a while and whenever I close a source
game windows gives me an error saying it has stopped working. Reinstalling (vista and steam) didn't help so I just deal with it. Still would consider it stable by more relaxed standards.

 

[quadomatic]

I would guess that if Orthos runs fine but you're having issues, it's an issue with your windows installation.

 

[Markfw]

Originally posted by: Drsignguy

Originally posted by: taltamir
I have a new benchmark for stability. Get firefox 3, open many tabs (10 windows with 10 tabs each). Set it to SAVE open tabs when exiting and reload them on next lunch...

Run OCCTv2 one mixed CPU + RAM (tests CPU and RAM at the same time! it will not work with just CPU test). When OCCT is running open FF and watch it try to load all those pages at once.

"perfectly stable" systems will blue screen every time i did this, until the voltage was raised enough.

PS. Stable means 0 problems, none, whatsoever.

Absolutly correct!

I had to do some tweaking today and I am pretty G'damn sure I got it right this time. On my main rig, 8Gb of ram puts a strain on the system and I had to bump up the memory voltage and the NB voltage a bit more to get it stable. How do I know its stable now? 2 smp clients, and 1 gpu client and its crunching on. No glithces, drips or errors. Temps are really good too.



So, what taltamir is basically suggesting, load the hell out of it and if she holds, all is good!

Yes, but I have one system that runs 2 F@H smp clients 24/7 for months. Then it will lock up. After reboot, and changing nothing, it goes for months again. Not even sure I call that stable.