i5-2500K stable?

[Tullphan]

I ran 12+ hours of OCCT followed by 100 passes of IBT using maximum stress. No errors on either.
I realize it's probably not 100% stable as nothing probably is that isn't running at manufacturers specs, but is it pretty close?

 

[Silent Knight]

100 passes of IBT on maximum level sounds good for me.Althought you need only 20-30 passes to declare stable.

 

[tangoseal]

I declare stable when it doesnt crash.

 

[Tullphan]

I declare stable when it doesnt crash.

Oh how true

 

[Silent Knight]

I declare stable when it doesnt crash.

So if it doesn't crash in these 20-30 rounds then can i declare it stable?

 

[BababooeyHTJ]

So if it doesn't crash in these 20-30 rounds then can i declare it stable?

No, prime blend seems to be the best way to stress test sandy bridge. I could pass around 50 runs of IBT but fail after a couple hours in a prime blend test. I've seen quite a few other people report the same.

It might be worth trying to do a custom prime run with prime blend settings and all of your memory enabled. I've heard that works well for sandy but I haven't tried it myself.

I declare stable when it doesnt crash.

Thats awful advice and I will never understand why so many people seem to give that advice. I've seen an unstable overclock cause artifacting in a few games, it can caues random rare crash to desktops, or worse of all data corruption because this doesn't go away when you set your rig back to stock speeds. You have got to wonder how many people complaining about "buggy games" really should be looking at their overclocks.

If a little proper stress testing can help me to avoid all of that headache then I'm all for it.

 

[saedrin]

My benchmark is usually 60 rounds of IBT, somewhere around 15-20 minutes worth of stress. If I can't crash it with that, it's very unlikely gaming is going to redline at 100% for that long anyways.

 

[BababooeyHTJ]

My benchmark is usually 60 rounds of IBT, somewhere around 15-20 minutes worth of stress. If I can't crash it with that, it's very unlikely gaming is going to redline at 100% for that long anyways.

How many threads have we seen where Linpack stable sandy bridge systems are crashing while doing simple tasks like web browsing?

 

[Silent Knight]

No, prime blend seems to be the best way to stress test sandy bridge. I could pass around 50 runs of IBT but fail after a couple hours in a prime blend test. I've seen quite a few other people report the same.

Actually installing the AVX pack for IBT/LinX will stress the SB CPUs to their limits.Older versions don't support AVX so they can't push it as hard as the new one does.In that case,Prime can't compare to IBT/LinX.
But it's just relative.Sometimes Prime crashes while IBT/LinX run flawlessly and vice versa.So it's necessary to run both of these tests.

Just my thought.

 

[tangoseal]

Thats awful advice and I will never understand why so many people seem to give that advice...

How many threads have we seen where Linpack stable sandy bridge systems are crashing while doing simple tasks like web browsing?

You just contradicted yourself mate.

So many people go overboard with stress testing etc .... just OC and use it until it crashes and then play with it some more, if it crashes change a few settings etc.... no different than stress testing for days on end.

Plus stress testing is just raw math run over and over again. Real testing is true dynamic usage. How is my advice bad? Its simple. You dont put a car on a Dyno and declare if 500 horsepower because it made those numbers 5 runs in a row. You have to get it on the street and drive the thing so you can really put some dynamic load on it and then you can see how stable that 500hp really is.

There are hills, stops, acceleration. humidity, temperature, and so many other factors that you might be surprised that car actually stalls a few times even though you "Overclocked the engine".

Same for a computer. You have to get some road testing on it so to say to find out if it is stable. I used to be all high and mighty by these so called burn in test etc... but I have seen countless times where a 100 runs of IBT or whatever was rock solid but the second I played a game BAM BSOD.

Overclocking is an Art, there is a feeling to it, there is a sense of being on the same wavelength with the PC. You dont hit the golf ball, you are the golf ball. That kind of thing.

But test away all you want just to find out that Photoshop or a game crashes your 250 run IBT tested PC with a BSOD.

 

[Haiku214]

It's stable if there will be no BSOD/crashes for 30 days straight.

 

[BababooeyHTJ]

You just contradicted yourself mate.

So many people go overboard with stress testing etc .... just OC and use it until it crashes and then play with it some more, if it crashes change a few settings etc.... no different than stress testing for days on end.

Plus stress testing is just raw math run over and over again. Real testing is true dynamic usage. How is my advice bad? Its simple. You dont put a car on a Dyno and declare if 500 horsepower because it made those numbers 5 runs in a row. You have to get it on the street and drive the thing so you can really put some dynamic load on it and then you can see how stable that 500hp really is.

There are hills, stops, acceleration. humidity, temperature, and so many other factors that you might be surprised that car actually stalls a few times even though you "Overclocked the engine".

Same for a computer. You have to get some road testing on it so to say to find out if it is stable. I used to be all high and mighty by these so called burn in test etc... but I have seen countless times where a 100 runs of IBT or whatever was rock solid but the second I played a game BAM BSOD.

Overclocking is an Art, there is a feeling to it, there is a sense of being on the same wavelength with the PC. You dont hit the golf ball, you are the golf ball. That kind of thing.

But test away all you want just to find out that Photoshop or a game crashes your 250 run IBT tested PC with a BSOD.

No, I'm saying that you can't run just and declare yourself stable. I have yet to see an unstable overclock that can't be replicated by some sort of stress test, for the most part. You have to know what to look for. I didn't start overclocking yesterday and this isn't my first machine.

Prime blend is very, very good to determine stability with Sandy. If you can pass a few hours of that odds are you are stable. There is the occasional exception due to loadline, or a little bump to vccio.

I see no good reason not to run a a couple of hours of prime blend when you first overclock you pc to help determine stability but you can go by whatever feeling that you get. just make sure to remind me to ignore you when you are calling some random game or driver buggy.

Plus stress testing is just raw math run over and over again.

What do you think that a cpu does?

 

[tangoseal]

No, I'm saying that you can't run just and declare yourself stable. I have yet to see an unstable overclock that can't be replicated by some sort of stress test, for the most part. You have to know what to look for. I didn't start overclocking yesterday and this isn't my first machine.

Prime blend is very, very good to determine stability with Sandy. If you can pass a few hours of that odds are you are stable. There is the occasional exception due to loadline, or a little bump to vccio.

Remind me to ignore you when you are calling some random game or driver buggy.

I never will call a game buggy except for Crysis 2 and Duke Nukem Forever which both are just plain suck ass horrible console ports.

 

[MaxxxRacer]

I remember a number of years ago when I was overclocking TCCD to 300Mhz on A64 X2 that it would crash in any Source game (HL2, Counter Strike, etc), but was prime and memtest stable.

That said, 24 hours of Linx with AVX and Prime 95 blend and a run of Super Pi 32M for good measure is a good enough to call it stable.

 

[BababooeyHTJ]

I remember a number of years ago when I was overclocking TCCD to 300Mhz on A64 X2 that it would crash in any Source game (HL2, Counter Strike, etc), but was prime and memtest stable.

Very true, there are always exceptions.

Actually installing the AVX pack for IBT/LinX will stress the SB CPUs to their limits.Older versions don't support AVX so they can't push it as hard as the new one does.In that case,Prime can't compare to IBT/LinX.
But it's just relative.Sometimes Prime crashes while IBT/LinX run flawlessly and vice versa.So it's necessary to run both of these tests.

Just my thought.

AVX Linpack does a great job. I use about 30 passes of Linpack with AVX support to get me into the ballpark and then go for a long Prime blend run.