Overclock Stability Question

[Kent10]

Hi:

I have achieved stability in overclocking my Q6600 processor to 3.33Ghz with 1.375v. I ran Prime95 for 24 hours with no errors. This was tested with my memory at its lowest setting.



Now I am trying to find out how fast my memory can go. I have my memory at 1184MHz with 5-5-5-15 timings and 2.2v. It passed 17 passes of memtest86 (10 hours) but then when I ran Prime95 again with the 1184 setting it failed after 4.5 hours.


I am going to run memtest86 on the memory for 24 hours now to be sure the memory is OK at the 1184 setting. If it passes 24 hours, do you know what could be causing Prime95 to fail. What should I try next. Should I lower the memory to the next lowest setting which is 1110. Is there any way to get anything between 1184 and 1110 on my motherboard. I can’t figure out anything.



I have read only a little about changing voltages on the north bridge and elsewhere. I don’t know if that would be a possibility too.



Thanks, Kent

[_WBaS_]

It's hard to say why it would pass memtest and not prime, but if the only thing you changed was the ram settings, then it must be a problem with the ram. When you OC your CPU, you are actually OC'ing your ram as well. So if you are setting your ram to 1184Mhz, then you are actually running higher frequencies. Try running cpu-z to see what your ram is actually running at. Just remember to double the value shown because ram is Double Data Rate (DDR). If you look in the SPD tab, you will see what your ram is rated for.

Since this seems to be an issue with the ram, I don't think increasing your NB voltage would help much. That tends to help with increasing FSB. But I'm not an expert at Intel OC'ing since I've only done it on this AMD rig. A mild increase might help this issue, but could also help you raise your FSB more :)

[Kent10]

Thanks WBaS for the info. 1184 is the OC'd frequency of the RAM. CPU-Z is 592 or half of that. I have a divider set but I can't remember what it is. We'll see, maybe the memory will still fail the Memtest86 24 hour test and then I can just lower it. I wish I didn't have to lower it so much to 1110 but I might fool around with the timings a little too. I guess I could also try Prime95 again and see what it does. Perhaps the first failure was just a fluke. Or maybe the 3.33 CPU setting is borderline. It did fail at 3.34 so maybe lowering this to 3.32 will help. Thanks again, Kent

[_WBaS_]

Oh alright, I guess that frequency makes sense. I think cpu-z tells you what divider you're using? Anyways, good luck with your OC!

[Kent10]

Here is someone else in a similar situation--a quote from another web site. "Prime95 torture test adds a lot of heat to the box. When my RAM timings were too aggressive, my box would pass Memtest but fail in Prime95 because the RAM couldn't keep up with the aggressive timing at the warmer temp!"

While Prime95 was running my tests ranged from the upper 50's to the upper 60's and very briefly hit 70. They were generally in the low to mid 60's though. My CPU and memory is being cooled with 2 1500RPM fans on the GeminII so I hope that is good but perhaps I have too much heat for the memory to pass in Windows.

[polobunny]

In my opinion, if normally it didn't fail Prime95 and now it does, your system is considered unstable.
What I personally like to do is find the maximum stable values for the RAM and CPU and just knock it back down one bit. This normally assures a reliable system overall, as computer components like everything else can deteriorate over time.

Edit: That's very hot too... I've had many computer pass memtest86+ but not even boot into Windows by the way. ;)

[Kent10]

Thanks Polobunny. I think that may be what I will do. I will try Prime95 one more time and if it fails I will use the lower memory setting and possibly a lower CPU setting and then run Prime95. I am a little concerned about the temps too. I have a G0 processor rated for 71C so I thought if it only gets to 70 rarely it should be OK but please correct me if I am wrong. I would never tax my machine as in Prime95 so I think it may never get close to those temps. The ambient temp will be a little higher in the summer too so I have to consider that. Thanks, Kent

[polobunny]

Originally Posted by Kent10

Thanks Polobunny. I think that may be what I will do. I will try Prime95 one more time and if it fails I will use the lower memory setting and possibly a lower CPU setting and then run Prime95. I am a little concerned about the temps too. I have a G0 processor rated for 71C so I thought if it only gets to 70 rarely it should be OK but please correct me if I am wrong. I would never tax my machine as in Prime95 so I think it may never get close to those temps. The ambient temp will be a little higher in the summer too so I have to consider that. Thanks, Kent

The thermal specification is the CPU maximum temp. You want to keep your processor lower than this at ALL times.
Also, as heat rises, your processor lifetime falls. Even then 3.2Ghz would be mighty fine on a Q6600. It's when you get greedy that you push your system too high, for just 100mhz.
Overall, i'd search for a nice setting over the 3Ghz bar, where you can lower your voltage and thus reduce the heat.

[Kent10]

You are right about being greedy. I had found a stable setting for 3.2GHz and the voltage was quite a bit lower than for 3.33. Just for the little extra MHz I had to raise the voltage and heat a lot. I will follow your advice and lower it back. Thanks. I don't want to lose my brand new processor. I will test it later. Perhaps Prime95 will pass with everything a step lower. Maybe I can keep a higher memory setting this way too since the heat will be less. Thanks so much for your help.

[polobunny]

Originally Posted by Kent10

You are right about being greedy. I had found a stable setting for 3.2GHz and the voltage was quite a bit lower than for 3.33. Just for the little extra MHz I had to raise the voltage and heat a lot. I will follow your advice and lower it back. Thanks. I don't want to lose my brand new processor. I will test it later. Perhaps Prime95 will pass with everything a step lower. Maybe I can keep a higher memory setting this way too since the heat will be less. Thanks so much for your help.

Exactly what I meant. To go from 3.2 to 3.33 you had to raise considerably the voltage and therefore create more heat, which could be what is stopping you from having a higher memory overclock. 3.2Ghz is nowhere near a slug anyway.