QX9650 and FSB 450... it works but...?

[Trivia89]

Hi there!
I'm going to edit this because i'm really disappointed with my vendor:
i thought the config they sold me te pc (450 fsb) would have been right, but on ther last warranty intervention they almost blamed me for breaking the pc, the same pc i bought 7 months ago and has been only an useless piece of iron all this time.

Anyway, there was no way making it work by using the frequency declared by Corsair: the whole pc is just :banana::banana::banana::banana:, i think they just sold me broken components.

[xgman]

Try a tad more vcore?

[Trivia89]

no, it just doesn't work

[Cronos]

Increase vMCH. X38 needs higher MCH voltage to b stable at higher FSB, compared to P35.

[Trivia89]

Increase vMCH. X38 needs higher MCH voltage to b stable at higher FSB, compared to P35.

Is vMCH the Northbridge voltage?

[xgman]

Running all voltages on auto will give you problems, Try with vcore and NB voltage bumps first and maybe experiment with vdimm voltages as well. Get stable there then tweak other settings.

[Trivia89]

Running all voltages on auto will give you problems, Try with vcore and NB voltage bumps first and maybe experiment with vdimm voltages as well. Get stable there then tweak other settings.


Thank you!

[Hornet331]

it's typical for a X38.

That chipset needs a lot of voltage to get stabel at anything higher then 400... i need vMCH +0,325 (1,575V) to get 450 bootable on my DQ6.

[safan80]

what multi are you using?

[crazyea]

Thank you!
I mean, before this DDR3 set i never had the need to raise NB voltage, so i'm a little dubious if, on the same frequency (i mean, like 333x10 and 450x7,5) i needed to raise the vcore this much.

I'm also really curious to know on what vcore value i can say i'm pretty sure to not shorten my cpu life too much, are 1,34/1,35 "safe settings"? =)


If you are running the same clock, just different fsb/mutli, then it is NOT a vcore issue. This is a NB issue. May need more volts there. Also, it could be a poor divider as well. try running 1:1 or another divider. Make sure you memory is at or below spec to test the dividers.

[Trivia89]

If you are running the same clock, just different fsb/mutli, then it is NOT a vcore issue. This is a NB issue. May need more volts there. Also, it could be a poor divider as well. try running 1:1 or another divider. Make sure you memory is at or below spec to test the dividers.

Interesting...
Thank you for the infos ;)

[crazyea]

If you raise your cpu speed above stock, then you need to increase vcore/fsb vtt.
If you raise your fsb above stock, you will need to raise NB/Mem volts.
If you raise you mem above stock, you will need to raise NB/Mem volts.

You "should" be able to run at 4.0ghz with approx 1.31v. I know evry cpu is different, but at least its a starting point.

You really need to isolate them one at a time.

Test the memory subsystem (NB) first.

1. Set cpu/mb/mem to default voltages (not auto). Ex: My QX9650 is 1.24v, 1.1FSB, 1.24NB and my memory is rated for 2.1v. Also set your memory to the 1:1 divider. (450/1350)

2. Raise your fsb to the 450 you desire, and lower your multi 6.5 (closes to stock without going over)

3. Reboot, if it does not post, you need raise the NBv a little a a time until it does.

4. Once in windows at these settings, run Prime95 Blend for at least 1/2 hour. If it is stable (without errors) you can move on, if not raise the NBv a bit more.

5. Once stable here, you can move on.

Test the memory (ddr/divider)

1. Change to the desired divider getting your memory closest to spec (timings/speed)

2. Since you memory is already at its rated voltage, if you cannot boot, you will need to raise the mch again. Don't go too high, because if you cannot boot it could also be a bad divider. Some dividers are so bad that they only work with an insane amount of volts. If that is the case for you move on to the next divider after an increase of 0.1v on the NB, and then start again at your stable voltage in the prior tests.

3. If you get to windows, run prime95 blend again for a least 1/2hour. If it fails, try raising the NBv slightly again.

3. If you are without errors, you can now tighten your timings, and test (prime95 blend) again raising NBv until stable. (you can also slight increase DDR volts if you start to head way out of spec)

4. As your memory speed increase or timings tighten, you will need to slowly raise NB/DDR volts as you go.

Test your CPU

1. Assuming that you are now stable on your NB/Mem, you can now be assured that raising your multi will isolate any instabilities to your CPU.

2. Change your multi to 450x8. Reboot. If it does not boot raise your vcore until it does.

3. Once in windows, run prime95 small fft for at least 1/2 hour without errors. If it fails raise your cpuv until you can pass. (Remember to try and keep your FSB vtt or sometimes called cpu vtt to at around 88% the voltage of your vcore. this ration seems to work best)