On the knife's edge, again...

[Loser777]

I never quite got my Q6600 400x9 100% stable even though I managed to achieve this:

That setup would never again run 400x9 past Prime95 8 Hours or Linpack 2GB 10 loops no matter what setup I tried.
Eventually, I narrowed it down to my dying power supply (which finally popped in August)

Thanks to black friday, I got a new power supply, but I'm running into trouble again. This time, I'm on Windows 7 //Linpack 64bit so stability is going to be a little harder to achieve, but I'm running into a brick wall again @ LinX 2GB.
I just can't pass 20 loops no matter that settings I try... I've gotten to about 17 and a half with no errors, but I'll get a BSOD right after that.

Settings right now:
P5Q-E
1.3875v vCore (1.400v LOAD)
1.36v VTT
1.62v PLL
1.26v NB
These are much higher than what was originally quasi-stable on my old PSU, yet I'm no closer...
With these settings, temps are usually in the low 90s during 2GB linpack.

What's keeping me from getting there?

[leo27]

Thoses temps are way too high for linpack max intel recommends is 72C for 24/7 use, i know its just a stress test but your still very high.

Also vTT is supposed to be a bit higher than vCPU (take a look in the GTL settings thread post by mikeyakame) i.e my vCPU is 1.400 in windows and vTT is 1.42. Once temps are under control you'll problably need around 1.4-1.55v to be fully stable at 3.6Ghz. If you can't get it stable with these voltages then try for 3.5Ghz or a bit lower.

I'm not sure why you were partially stable previously and not now it could be that your new power supply supplies slightly less voltage on the supply rails.

[Loser777]

If the temps are too high, why does it pass 17 iterations with no residual errors and then crash?

[Loser777]

BUMP, achieved this


Not sure if it would pass 20 iterations though- also I got LOG in Thermal- probably means that CPU would have throttled if TM was enabled.

[AH64]

Originally Posted by leo27

intel recommends is 72C for 24/7 use,

It'll go way past 90*c before it starts throttling at 99-100*c (TJMAX) But of course, the further away you are from tjmax, the better.

[Loser777]

Well, I think there are serious voltage scaling issues going on.
3.2GHz seems to be stable at stock 1.2125v, and 3.4GHz around the 1.25GHz mark.
Honestly, if I can go from 97C load to 70C load for 200MHz less, it doesn't look like much of a sacrifice.

[unclewebb]

When the LOG word shows up in RealTemp, that's a sign that at least one thermal throttling episode was logged since you pushed the power button. If that turns red and changes to HOT, that's a sign that thermal throttling is in progress. RealTemp reads the bit within the CPU so even if RealTemp wasn't running at the time, this bit would still get set and RealTemp would be able to read it after the fact. Thermal throttling can happen hundreds of times a second so when it first starts to happen it will mostly show LOG.

The 5C difference in your temperatures at full load is in my opinion a sign that TJMax is not set consistently across all 4 cores. I think Intel does this deliberately so all 4 cores don't thermal throttle at the exact same time. If you set TJMax to 100, 100, 105, 105 then you should see your 4 cores line up better and core 2 and core 3 will report more accurate temperatures.

Most CPUs lose stability when you get them this warm. Without better cooling you should probably reduce your overclock a couple of hundred MHz. If you don't run LinX 24/7 then you might be able to get away with 3600 MHz just fine for gaming and stuff like that. Real world temps are a lot less than LinX temperatures.

It's common for people to interchange Tcase temperature ratings and core temperatures. The two are not the same and shouldn't be compared. As long as your CPU is not thermal throttling and is running stable then your peak core temperature isn't that important. A little more room before the throttling point is always nice.

RealTemp 3.49
http://www.fileden.com/files/2008/3/...alTempBeta.zip

[Leeghoofd]

Your CPU get's unstable due to high temps, think we've seen and told that before in all ya previous threads... you need to upgrade or revise ya cooling. Nothing more nothing less... If you have better cooling you can nibble of maybe some on the voltages...