Unable to reach long-term stability

[tikithorsen]

Since I first started to overclock my brand new rig there was a symptom:
Any type of overclock seems stable for days some even a week or more and then fail.
My first OC was 3.8GHz at 1.375v, it worked for like 2 weeks, also passing at least 4 hours of Prime95 and then it started to fail. I told myself that it was never stabled in the first place and I proceeded to my second OC which was @ 4.0GHz 1.450V and the NB @ 2.8GHz 1.275v (which was my most successful OC) passed 8+ hours of Prime95 and around 1 month of heavy gaming).
Last one to fail was 4.0GHZ @ 1.475v after a week and 2 hours of Prime95.

They have a fancy of failing when I play BF3. The error is always the same; BF3 crashes at the same time as Skype (since I'm on a call with a friend), but strangely enough all the other open programs don't crash, though some stop responding. Other times it simply freezes. In either scenario I have to do a hard reboot to solve it.

What I have checked off the list are only temps, since I constantly keep an eye of them and they always stay within parameters.
Now, my suspicious is on the CPU IMC and/or my RAM sticks that can't really work properly at 1600MHz. It passed 2 runs of Memtest86+ back when I first did it, but I will run it again just to make sure.

I would really appreciate any type of assistance pin-pointing the culprit.
Thanks in advance for any help.

 

[RGone]

When I have any sort of failure and I am using the computer in an Overclocked condition, I remove the overclock and return to stock settings and see if the applications that are failing continue to fail. No sense in chasing all around the computer room if the failures occur even at stock speeds. That is the first, first thing I do to logically check out the cause of errors when I was overclocked. And greatly more so since you say each overclock has passed hours and hours of P95. No reason memory or otherwise for hardware to be changing in a week or two time. Not unless you have a failing piece of hardware and then you likely would know it because returning to stock speeds likely will not cure the issue.
RGone...

 

[tikithorsen]

I will set all the settings manually to the stock values (don't like the auto values) and run some tests and use it for a week or so.

 

[Johan45]

Do you have your GFX OC"ed as well if so maybe start there. It could be an issue there since it fancies crashing while you're playing BF3.

 

[trents]

What frequency is your RAM rated for and what frequency are you running it at? If you are trying to run it at 1600 mhz or higher on that CPU then this could be the problem.

 

[tikithorsen]

Do you have your GFX OC"ed as well if so maybe start there. It could be an issue there since it fancies crashing while you're playing BF3.

Already checked that, sadly is not it.

What frequency is your RAM rated for and what frequency are you running it at? If you are trying to run it at 1600 mhz or higher on that CPU then this could be the problem.

Is rated at 1600MHz 9-9-9-24 2T 1.5v.
I was running it exactly at that.
Now is at 1333MHz 7-7-7-19 1T 1.5V.

If this time it holds up, I'll try just OC'ing the CPU and leaving the rest unchanged.

Thanks for the replies.

 

[storm-chaser]

Is that the ripjaws 4gb kit # F3-12800CL9D-4GBRL?

Its also quite possible you could have one bad/failing memory module and maybe not running into the Denab memory controller issue. Due to the fact you are seeing lots of uptime between failures. You can test each memory module individually to determine whether the error may be coming from one single stick.

Should be a lifetime warranty on that memory.

 

[tikithorsen]

Is that the ripjaws 4gb kit # F3-12800CL9D-4GBRL?

Its also quite possible you could have one bad/failing memory module and maybe not running into the Denab memory controller issue. Due to the fact you are seeing lots of uptime between failures. You can test each memory module individually to determine whether the error may be coming from one single stick.

Should be a lifetime warranty on that memory.

Is the F3-12800CL9D-8GBXL model.
¿How many passes of Memtest86+ should I run?
Problem is that I'm pretty sure that G.Skill doesn't exist as such on my country, so the warranty only the year given by the retailer, which already expired.

 

[storm-chaser]

Is the F3-12800CL9D-8GBXL model.
¿How many passes of Memtest86+ should I run?
Problem is that I'm pretty sure that G.Skill doesn't exist as such on my country, so the warranty only the year given by the retailer, which already expired.

You have a lot of time between mean failures so it might be good to leave on all night to see if you get any errors. Watch your temps as the test goes on.
Might want to take the side off the case to reduce heat and monitor your temps for a while before letting it go.

I'm just brainstorming - here are some other possible causes if you still find a dead end:
CPU-NB voltage is critical: Are you at 2.8Ghz right now? If so, you might bump your CPU-NB voltage to 1.3, thats safe for air. Or try lowering your NB to 2.6ghz and test for reliability again.

If you can confirm that the ram is in fact good, I'd be willing to try to help you get it stable at 1600. The Denab IMC has its issues but nothing I havent been able to overcome before on other P2 9XX CPUs paired with XMP rated 1600 ram.

 

[tikithorsen]

You have a lot of time between mean failures so it might be good to leave on all night to see if you get any errors. Watch your temps as the test goes on.
Might want to take the side off the case to reduce heat and monitor your temps for a while before letting it go.

I'm just brainstorming - here are some other possible causes if you still find a dead end:
CPU-NB voltage is critical: Are you at 2.8Ghz right now? If so, you might bump your CPU-NB voltage to 1.3, thats safe for air. Or try lowering your NB to 2.6ghz and test for reliability again.

If you can confirm that the ram is in fact good, I'd be willing to try to help you get it stable at 1600. The Denab IMC has its issues but nothing I havent been able to overcome before on other P2 9XX CPUs paired with XMP rated 1600 ram.

When it was at 2.8GHZ the voltage was 1.3v, yes. I'll leave the the RAM sticks testing tonight.
Thanks for the ideas.

 

[tikithorsen]

Looks like this is finally stable (with the settings displayed on the sig).
Thanks for the advises and ideas.

 

[tikithorsen]

Update:
I tried upping the frequency to 4.0GHz and the vcore to 1.450v and passed an hour of Prime95 Blend Test but failed after 1 hour of WoW.

Maybe it can be stable with 1.475V, but that is just way too much increase of voltage (heat) for a mere extra 200MHz, IMHO.

 

[Johan45]

That's up to the individual. You're the one paying in the end, if something goes wrong. I have mine at 4.64 with 1.476v under load. but if not loaded it runs at 1.45. I've used the LLC to give it a bump when it needs it.

 

[storm-chaser]

Looks like this is finally stable (with the settings displayed on the sig).
Thanks for the advises and ideas.

AMD Phenom II 965 @ 3.8GHz 1.400v / NB @ 2400MHz 1.225v
Cooler Master Hyper 212 Plus
Gigabyte GA-990FXA-UD3 rev 1.1 (F7)
G.Skill RipjawsX 2x4GB @ 9-9-9-24-33 2T 1600MHz 1.500v

Great, I'm happy to hear you're running the ram at its full potential of 1600 MHz with the Denab cpu. Very nice overclock