QX9xxx overclocking results

[3Tone]

Originally Posted by karlcloudy

Hmm, this thread is dead silent

Gee I wonder why????

[karlcloudy]

Originally Posted by 3Tone

Gee I wonder why????

Now comeon, isn't this pushing it a bit too far?

[karlcloudy]

Originally Posted by cokeeffekt

bump the MCH to 1.3 just to test

OK, I just tried 1.26V. ASUS marks the voltage in Yellow when I use 1.3V, so I thought that I'd try 1.26V first and it's going great so far. Sounds like I might finally have worked out what was stopping my 4GHz OC and making my 3.8Ghz OC unstable hopefully

[toXin]

Originally Posted by cokeeffekt

bump the MCH to 1.3 just to test

This should work.

Originally Posted by karlcloudy

I'm also debating about whether to "not bother" worrying about it as I've had it running for quite a while without any detrimental effect on actual system stability. It hasn't actually crashed or done anything funny.

There's two things to worry about there.

1.
Visually things may seem like they are ok, But over time things could get bad.
In a nutshell, your CPU goes to process something like ABC12345. An unstable CPU could process ABB11345 instead which is bad data, Overtime the more bad data that is processed and written the worse the OS/functionality of your PC gets.

But things take a while to get bad or even unusable. If your like me and you re-install a fresh OS once or twice a year then should be fine.

2.
Clocking stability relies on temps, if it starts hitting the 70s then it's most probably going to fail. Running tests on full load for too long and not having a cooler without quick enough heat transfer will most likely give you tempratures like this. This would explain why your small FFTs are passing and why the large ones aren't.

Now that you have read that imagine playing performance hungry games or decoding/encoding large files. Doing either one will put your CPU under full load, and depending on how long they are under load will eventually start hitting higher temps. Higher temps = bad data.

[3Tone]

2.
Clocking stability relies on temps, if it starts hitting the 70s then it's most probably going to fail. Running tests on full load for too long and not having a cooler without quick enough heat transfer will most likely give you tempratures like this. This would explain why your small FFTs are passing and why the large ones aren't.

Not quite, I've had a prime95 session going for about 24hrs at over 80deg.

[cokeeffekt]

yeah an MCH of 1.3 is about the max safe limit for standard cooling ive read, although ive had mine as high as 1.42 to test

[karlcloudy]

Originally Posted by toXin

This should work.



There's two things to worry about there.

1.
Visually things may seem like they are ok, But over time things could get bad.
In a nutshell, your CPU goes to process something like ABC12345. An unstable CPU could process ABB11345 instead which is bad data, Overtime the more bad data that is processed and written the worse the OS/functionality of your PC gets.

But things take a while to get bad or even unusable. If your like me and you re-install a fresh OS once or twice a year then should be fine.

2.
Clocking stability relies on temps, if it starts hitting the 70s then it's most probably going to fail. Running tests on full load for too long and not having a cooler without quick enough heat transfer will most likely give you tempratures like this. This would explain why your small FFTs are passing and why the large ones aren't.

Now that you have read that imagine playing performance hungry games or decoding/encoding large files. Doing either one will put your CPU under full load, and depending on how long they are under load will eventually start hitting higher temps. Higher temps = bad data.

The temps I'm getting aren't really that high in my opinion. In reality, I've never seen the temps anywhere near the rated maximum Tcase temperature. Now Tcase's limit is obviously lower than the cores themselves, because of it's location. I haven't done any proper stability testing yet (only about 10 minutes this morning before I left), but already, setting the northbridge to 1.26V improved it a LOT. I changed to 1.3V as the previous poster said and I'm going to test it soon. LinX has never failed and I actually get much higher temps with this (70ish, although I've been monitoring while I've been playing games and the like and get no where near this normally). prime95 small FFTs heat up my processor more than large FFTs iirc. I think this problem is squarely down to the northbridge not having enough power. Does anyone know how safe it is to leave the NB on 1.3V? This is what my motherboard cooling is like (I assume this has a huge part in whether or not it's safe or not). Interestingly, if you set the NB over 1.42V (don't worry, I didn't actually save or use this configuration at any point), it changes to pink and says "BETTER COOLING SYSTEM REQUIRED" or something to that effect.

[snappy1]

I ran an X48 rampage extreme with a QX9650, when benching at 4.4-4.5Ghz I was getting crashes, it was actually my northbridge overheating. I fitted a 80mm high speed fan over my NB, problem solved.
I am still running the same fan in my i7 system, sitting on my top graphics card at the northbridge, monitor your temps and all should be ok.
I have seen temps over 80 deg c on my NB in my REX, scary stuff

[karlcloudy]

Originally Posted by snappy1

I ran an X48 rampage extreme with a QX9650, when benching at 4.4-4.5Ghz I was getting crashes, it was actually my northbridge overheating. I fitted a 80mm high speed fan over my NB, problem solved.
I am still running the same fan in my i7 system, sitting on my top graphics card at the northbridge, monitor your temps and all should be ok.
I have seen temps over 80 deg c on my NB in my REX, scary stuff

Where is NB temperature monitored from? Is it the reported motherboard temp or do you have to manually pop a probe near it?

[snappy1]

I use Everest Ultimate to monitor NB temps, but not all motherboards support this feature.

[karlcloudy]

Originally Posted by snappy1

I use Everest Ultimate to monitor NB temps, but not all motherboards support this feature.

Does it specifically say northbridge temperature?


Also, I have decided to set VTT manually as I've heard that it can be set to dangerous values automatically if left on Auto, although I don't know what setting I need for this type of overclock, so I have set it to stock... No doubt it needs more though? Could someone fill me in on how to adjust VTT properly?

[cokeeffekt]

i set mine to "normal"

[karlcloudy]

Originally Posted by cokeeffekt

i set mine to "normal"

I don't know whether mine has that (my gigabyte board had that), but "Normal" sets it to default anyway, which is effectively what I did by popping in what the BIOS said was the standard setting. I'm still having no joy with 1.3V on the NB with prime95 large FFT and I did get a BSOD when I tried to hibernate once. The test does last longer though, as on lower settings it pretty much errors out as soon as it can (probably about when the first test completes normally)

[karlcloudy]

Originally Posted by snappy1

I use Everest Ultimate to monitor NB temps, but not all motherboards support this feature.

Hmm, what are these temperature sensors?


Click to view full size!

[cokeeffekt]

do you have any fans with sensors on them ??