How Do We Define a "Stable" Overclock

[Arsynic]

I was thinking about this question while perusing another thread. If Windows boots up fine after you overclock and you can work for hours upon days doing a normal workload, is it necessarily stable? Maybe not. However, what if you can run Prime 95 for a couple of hours without errors? How about running Cinebench and PC Mark Vantage without bluescreening or rebooting? However, when you try to run Prime 95 overnight, the computer blue screens.

Is it still stable?

 

[RaistlinZ]

For me, stable is when I can game for 12 hours straight without a lockup. If I can run Prime for 12+ hours too then that's just a bonus. "Stable" should mean stable for doing your everyday workload without hiccup. You should feel confident that your system won't randomly crash doing everyday tasks.

 

[Vic Vega]

For me stable means gaming for 7-10 hours and not having any crashes, reboots, BSODs, etc. I don't find popular stress testing methods like Prime useful for my purposes - not knocking those, some people find them useful.

On a side note, my PC basically stays up 24/7 so if I game several hours a day and for a month and the system stays up throughout that time I'm satisfied with the result.

 

[Ben90]

Everyone has a different mindset on what a stable computer is. Over on XtremeSystems, a computer might be stable if it runs SuperPi and is able to save a screen shot before the system crashes. Most of these SuperPi stable settings cannot even handle the boot into windows and must be set via software.

A bank would consider our consumer processors running stock settings to be unstable. A flipped bit in non-ECC RAM could be the difference between a transaction costing the bank 000010101010 dollars or 100010101010 dollars.

Personally I do not rely on Linpac or Prime to see if my computer is stable under my definitions. It could run Linpac for years (Albeit at a throttled speed due to my insistence on using the stock cooler for a 4ghz+ OC), Yet crash sometime overnight while it is just sitting there mostly idle. A total system crash once every other month is okay with me, and an app crash every couple of weeks isn't the biggest deal in the world, but any more often and I'm going to start changing settings around.

If you really value your data, it might be best to just stay at stock settings, or invest in a way to avert silent data corruption.

 

[dguy6789]

If you can make the computer crash with any of the torture tests after any amount of time it's unstable.

 

[VirtualLarry]

To me, stable = 24/7 operation, without ANY strange behavior, data corruption, or crashing.

See my thread, 28 days and not stable enough, for what I went through with one of my rigs. It would run any stress test for several days on end, but it would spontaniously reboot after a week to a month. So therefore not stable for me.

 

[exar333]

To me, stable = 24/7 operation, without ANY strange behavior, data corruption, or crashing.

See my thread, 28 days and not stable enough, for what I went through with one of my rigs. It would run any stress test for several days on end, but it would spontaniously reboot after a week to a month. So therefore not stable for me.

+1.

 

[exar333]

I like to think that if I overclock the chip, does it do anything that would cause me to return a stock processor doing the same thing? Lock-ups, random re-boots, crashes, etc. None of these are acceptable to me for a stock or OC'd chip. It must be 100% stable, or I don't want it.

 

[Arsynic]

To me, stable = 24/7 operation, without ANY strange behavior, data corruption, or crashing.

See my thread, 28 days and not stable enough, for what I went through with one of my rigs. It would run any stress test for several days on end, but it would spontaniously reboot after a week to a month. So therefore not stable for me.

Do you consider running P95 stress test as "normal 24/7" operation? Or is it just running mostly with the CPU idle. For example, mine can run games and other high-stress apps fine but right now I can't run Prime95 overnight without it bluescreening.

 

[FishAk]

Most of you probably haven't been around stock OEM machines for so long you don't remember what they look like. I can tell you; they still have their random crashes. By some of your more strict definitions, no OEM stock machine is stable, let alone an overclock.

 

[Scotteq]

To me: 12 hours of P95 Blend, and 8 hours of OCCT with Zero errors. No hardware related crashes or lockups, period.

That is for my home PC, which I shut off when I'm not actually using it. We don't overclock at work.

 

[Scotteq]

... no OEM stock machine is stable.

Well... Yeah... Of course they're not!

Your point is?? :whiste:

 

[dbcooper1]

For me, it's having it behave the same overclocked or at stock settings regardless of what I'm doing. (aside from the speed and heat/power differences of course)

 

[Rifter]

12 hours or more prime is stable, i dont consider any lockups or crashes acceptable. I tested my current OC with 48 hours of prime and so far have hot had any lockups/BSOD. I do spend most of my time in linux which is more stable than windows though, i only game with windows.

You can say what you want about prime being overkill but encoding a video with a well threaded encoder will get your CPU to within 5-10c of what prime will, so prime is a very good toll to test stability with. Other people fold which will load your CPU to 100%, so making sure its stable under 100% loads is vital.

 

[Habeed]

This is a lot hard to measure than you might think. My machine is a 930 at ~4.2 on a Noctua air cooler. I was getting occasional crashes to where it would only just barely fail to finish an overnight 8 hour OCCT run (usually crashing only after 5+ hours). For some reason OCCT Linpack is harder than any other test I have found so far. (much harder than prime and regular Linpack)

Of course, a machine that occasionally crashes on OCCT will game for days without issues.

I finally set disabled all power management and set the voltage really high. Woke up this morning and it had survived an 8 hour run. Now I will have to do 2 more 8 hour runs before I am happy it is stable on these settings.

Even then I bet I'll get a bluescreen once in a blue moon. Nothing you can do about totally random bluescreens because they are so rare that it could be months before you get another one.

But getting it as stable on a high overclock as it is on stock is just impractical. You can get close but you can't tweak settings to stop incredibly rare glitches.

Ironically, I have a friend who runs his 920 at stock and he gets a weekly bluescreen, more often than even my overclocked 930 at unstable settings was getting...

 

[exar333]

Most of you probably haven't been around stock OEM machines for so long you don't remember what they look like. I can tell you; they still have their random crashes. By some of your more strict definitions, no OEM stock machine is stable, let alone an overclock.

Thats entirely untrue. I have used my bone-stock Dell at work for over two years, no crashes. My laptop at home (Sony) has never crashed in almost 3 years now. What crap machines do you use? I would argue they are defective, NOT "stable".

Stable machines DO NOT crash. Period.

 

[VirtualLarry]

Do you consider running P95 stress test as "normal 24/7" operation? Or is it just running mostly with the CPU idle. For example, mine can run games and other high-stress apps fine but right now I can't run Prime95 overnight without it bluescreening.

I do Distributed Computing. Which is very stressful on a CPU, much like Prime95 or OCCT. I would say if Prime95 is BSODing on you, you need to fix your overclock.

 

[zerogear]

LinX for 8 hours, Prime95 for 24 hours, Mark 3D for 5 loops.