How do you know when to raise one of the other core voltages?

Emotep

Distinguished
Jun 21, 2006
20
0
18,510
Right now, I am trying to overclock my Q9450 from 2.66GHZ to 3.2GHZ

My Northbridge, Southbridge, and VTT voltages are pretty modest...

NB = 1.33V
SB = 1.05V
VTT = 1.20V

I've slowly worked my CPU Voltage up from 1.12, which crashed instantly, to 1.2006, which currently is running Prime95's Large FFThttp://en.wikipedia.org/wiki/Fast_Fourier_transform stress testhttp://en.wikipedia.org/wiki/Stress_Test 9 hours and counting stable.

But how do you know when your CPU voltage is enough and it is time to raise one of the other voltages? And how do you know which one to raise?
 

Lupiron

Distinguished
Feb 9, 2008
1,711
0
19,780
If it always seems to fail at the same point, no matter the VCore!

When looking for your VCore that is needed. Sometimes it is best to start a bit high, and just work your way down. That way you avoide going up in single steps, which can take forever! Once it is stable at a higher voltage, it's simple to get it lowered.

AND if it isn't stable still, though you are pretty sure it should be, because the voltage is a bit high, you can add voltage to something else. Like NB for instance.

1.40 -1.45 seems great for 400 FSB.

By the way, if you managed to become LARGE fft stable, you are wasting your time. Large ffts is like half the two tests combined.

IE, if you have a RAM problem, you may see it with large ffts, and if you have a processor problem, you may see it with large ffts.

Which doesn't help at all, does it? Since you wanna be sure neither is having problems, You have to use JUST the small ffts for the processor stability. Its much more powerful than Large FFTs.

And you wanna use the Blend test, to test your system memory! That way if either fails, you know the root cause and can mess with one or the other! Since you know what one caused the failure.

Good luck with small FFTs if you havent used the before alone!

--Lupi
 

Emotep

Distinguished
Jun 21, 2006
20
0
18,510
By failing at the same point, what do you mean by that? Do you mean the same thread fails, or do you mean the stress test fails at around the same time period every time? (EG. Always fails at 40 minutes...)

Also, if my RAM is underclocked to get 1:1 ram ratio, do I still have to test the RAM?

My ram is 1066, but I am currently underclocking it to 800Mhz to match my 400 FSB
 

Lupiron

Distinguished
Feb 9, 2008
1,711
0
19,780
It is wise to test the ram, if just past the first series of tests. 17 minutes, because if your strap to NB is off, it will cause weird failures.

But it usually fails the full threaded blend test fast if that is the case. So just a small amount of time will work.

Usually a small adjustment to QDR or FSB will cure it. IE from 1600 QDR linked to 1603 QDR unlinked with a manual ram speed of over 802Mhz entered. It will apply something like 801.1 Mhz to the ram speed. And it will hopefully try a new divider!

Failing at the same point. yes. near the same time, or with the same trigger.

As in, it always fails at 15 mins, or fails regularly when entering a new test of Prime 95 small ffts. (IE, just after passing the first series of tests, it says 8K Test Passed! And then bombs all the time.)

--Lupi
 

Emotep

Distinguished
Jun 21, 2006
20
0
18,510
What is QDR?

Also, I have my NB Strap set to auto. If I am going to set my FSB to 400, what should my NB Strap be set to?

How exactly do I test ram again?
 

Lupiron

Distinguished
Feb 9, 2008
1,711
0
19,780
Oh, sorry. Most boards are using QDR so you can fine tune the FSB. Its the FSB x 4.

But you can change the FSB in .25 increments with it, as opposed to from 400, to 401.

400 is fine, unless you notice errors that are un accountable!

Then you can try 333, and it will change the way your ram speed is calculated. And hopefully work right as a result.

--Lupi
 

Emotep

Distinguished
Jun 21, 2006
20
0
18,510
Thanks for the help... so how long would you stress test in small FFT for to ensure stability?

Also, if my ram is underclocked and is at the max recommended voltage by the manufacturer (OCZ in this case), should I bother running the blend test after running a small FFT?
 

Lupiron

Distinguished
Feb 9, 2008
1,711
0
19,780
It is always safe to at least run the blend test through the first series of tests. But 8 Hours on the CPU will small ffts on all the cores should suffice!

Then it's time for the Gaming tests!

You can lower your voltage to the RAM, or tighten the timings, which may need a lil more on top of stock.

--Lupi