When do call your OC stable?

Based on this article, I would say months or years.

http://www.tomshardware.com/review [...] 48-17.html

Sorry, I'm saying that after reading that article it made me worried that my CPU could fry in a few months, even after withstanding a day of hard testing.

Here's a quote from the article, " The problem with blaming heat alone on a failure is that moderate increases in electromigration resistance usually require drastic temperature reductions. When it comes to protecting hundreds of dollars in equipment, we always make our recommendations to you erring on the side of caution.

We've learned through trial, error, and dead processors that voltage levels beyond 1.45 V at above-ambient temperatures can kill an Intel CPU etched at 32 nm (Sandy Bridge-based parts included) very quickly. Those same processors die a fairly slow death at voltage levels between 1.40 V and 1.45 V (somewhere between weeks and months on our test benches)."

First thing I'm doing when I get home is double checking the numbers that my "auto tune" did!

Any comments on which stability tests you run?

Nice guide Wolf!

What do you run Intel Burn Test at? 20 passes?

I used to 8 Hours Prime95 most of the time. I've never experienced any real problems.

The Sandy Bridge processors when OC'd using Offset Voltage are sometime dodgy though. I once ran 8 Hours Prime95 Blend and it was fine, then I spent some time on the interwebs and it blue-screened.

Whenever I play with OCs now I tend to run 15 Passes of IBT on Max and then 2 Hours OCCT - The projected stability testing level in comparison to Prime95 when you use the combination of those 2 is around 24 Hours.

I like anywhere between 8-12 hours of prime95, and anywhere from 1-3 hours of linpack x64 to confirm 100% stability.

I've had overclocks be fine in prime95 for 12+ hours but crash within 30 minutes of linpack...

www.google.com

i'm not the fan of stressing software..

i call my overclock stable, just running cinebench (make sure there's no lag in when your cpu rendering the box, if any lag happen, then your setting is not good enough), then half hour prime95, and ibt max 5 loops.
and after all, try to play 3D game at least 2-3 hours.

why gaming..?
because there's many case that overclocked cpu which pass ibt maximum stress, but has got bsod when do gaming or even in the idle condition..

if your setting can pass it, that's stable enough for your daily activity.

There's no point to make so much strees in your processor..
it just degrading your cpu faster..

in multicore era of cpu, each core that consist in your processor is not 100% indentical. so that's mean, each core has different capability if overclocking..
you can't reach 100% persen stable for all cores

Even in the default clock, if you stress it for the (very) long time, one of the cores will be failed too..

so that's mean, in the daily overclocking it's okay for 2 core 100% stable, 1 core 99% stable, and rest 95% stable..as long as you got no BSOD when you do your activity, it's stable..

sorry for my bad english..

lol

Edit:

TY for taking the time to actually answer the question
Now we know which sites you like.

@Wolfram..

It's okay..

everything that i wrote just my opinion..
based on my (little) experience in adding hertz by hertz to my old proceesor..

overclocking is not (really) exact knowledge, many aspect depend on "think", "feel" and "okay"..

sometimes everyone has different standart in life.


as long as you enjoy with your setting, (and got no BSOD when you doing daily task) you can call stable..

Looks like 10 hours is the mid point: almost 50% over and 50% under that value.

Edit:

As of the date of this original post that is

Alot of people seem to say that intelburntest is quicker at finding problems. But I have several times run burntest for about 8 minutes (10x 2GB test) and been fine, just to open prime95 and get a rounding error within 5 minutes.
My coolest core is also the first to fail each time, which I find a a bit peculiar.
I think about 8 hours would be enough of a test in prime95. But that is based on gut feeling more than anything else.

i use prime95, for 4 hours
and OCCT for 4 hours

i dont like it to stress it too long

if you all talking about SB then i would say none if the above
try this ;
-costume P95 (90% of your RAM)


or
-Fold (FAH)

lol..OK..

Interesting ideas and attitudes.

Stable to me is overnight Prime95 + Blend (or AIDA64), and then simultaneously running loops of 3DMark11 or similar GPU stress AND Prime95/AIDA64. Otherwise you got nothing, and CPU-z validations proves Z-E-R-O! Proves you can Post and Boot for a few seconds...i.e. zip.

My systems MUST be bulletproof; imaging me turning over a system with i.e. 4.5GHz+ OC and the second the guy runs e.g. BF3 it goes poof = worthless!

I could post a worthless 4.8GHz @ 1.36v, but to achieve above my personal rig required a 1.4v with lowest LLC.

Obviously not, but then again I like my CPU in its' 'non-degraded' and intact form. I have the older C1 SB-E, and oddly the newer one C2 degrades faster.

You have to remember, >1.45v vCore degrades your CPU and shortens its life, and so does exceeding a VTT and/or VCCSA >1.25v which does the same degrading damage. Further, don't forget about LLC which adds +0.01v~+0.05v spikes on top of vCore values.

Thomas recommends not exceeding 1.38v vCore (~4.6GHz) on the SB-E, but my assumption is he uses a much higher LLC than I do so IMO (1.38v + 0.05v = 1.43v) vs my (1.40v + 0.01~0.02v = 1.41v~1.42v); to know what I'm talking about see -> http://www.masterslair.com/vdroop- [...] eally-bad/

In other forums, there have been 1.5Xv vCore and 1.35v VTT and/or VCCSA degrade the SB-E's bin by 2~3 bin a matter of hours (-0.2GHz~-0.3GHz).

Since most of what I do is coding and testing SQL I cannot afford a 3~4 hour batch to go poof and repeat it all over again, nor risk data corruption. It's not a gaming rig as its' primary use. I know OC'ing, but not a too interested in its section - I helped @vollman1 some time ago...

Hey how's it going? I hope well! The children must be happy campers!

IMO - going higher than 4.5GHz does nearly zip and nothing for gaming. There no joy in getting into a 'fire fight or just about getting "the" guy' and lights go out.

It depends on what you're doing, if like now, I load a 'Stock' profile and 10 seconds later I'm posting to threads -- what's the point doing it @ 4.8GHz?!

7 hours of OCCT, many cinebench runs ( I use cinema 4d so this is crucial) 1 run of inteburnintest, and for some reason it crashes on p95?

^Huh? xplain it?

The SB and SB-E have different tolerances, ditto with my i7-980X. All 32nm CPUs, but that the only commonality.

No that's not right, at least not on SB/SB-E, you can degrade the SB-E on voltage alone. I've seen a 1.5Xv vCore degrade on LN2. More than likely that i7-3770K is toast, and that's assuming it's not a doctored-up Photoshoped validation. Weird to use DICE over LN2 or I would have even thought LHe as used with AMD's 8GHz.

All interesting, but as useless as two butt c____s.

The dummy forgot to erase the ID above the CPU, it looks like 2182052.

The only tests I run for that short of a time are Futuremark benches. Most folks, if they half care about stability run a series of 'burn-ins' for at least 2 days. This guy tests about as hard as I do - http://www.youtube.com/user/SingularityComputers

What other folks do and/or how they test is their business, just as how hard I test is my choice and IMO the best way. I build for others and there's nothing more that will piss them off faster is to have their rigs die a few days after they get them all set up and install a their stuff. I am even harder on Workstations and often they're run 24/7 stressed for at least 4 days. I indeed have had all sorts of components fail, and it's an easy call to let the recipients know there's going to be a '4~5 day delay over a bad XYZ component'. For the rigs that are going to be water blocked, I'd be a fool not to make sure each GPU, etc isn't up to a close tolerance in OC'ing before I deconstruct, block, water loop and build. Hours and expensive cooling wasted otherwise!!!