Seeking a second opinion overclocking some dated hardware

[BonzaiDuck]

+1 for Error8's remark about VTT or "CPU_FSB" voltage. It should depend on how high you're clocking the host-frequency. On some boards, you need to push this voltage to 1.4 (or higher) as host-frequency reaches 400 Mhz. On other boards, maybe 1.30V. 1.4V is pushing toward a dangerous limit for the Penryn (Wolfie or Yorkie) cores; on the 65nm Kenstfield, it's acceptable. And you should be able to reduce VCORE slightly after increasing VTT.

 

[California Roll]

This thread piqued my interest in oc'ing again. I settled on 3.0 back in Jan and haven't touched my rig since.

Thanks for the linpack link. I've been using Prime the whole time and this saves a lot more time.

I've been able to undervolt my late model, 1.325vid, Q6600 and keep it rock stable at 334x9. Here are the relevant voltages only:

Ram: 1.8v, 5-5-5-12 (stock)
MCH: 1.1v (stock)
VTT: 1.2v (stock)
PLL: 1.5v (stock)

Vcore: 1.2625v (anything lower = linpack fail or bsod)
(CPU-Z: 1.232v)

I entered in all the above manually. Everything else is set it AUTO.

I have LLC enabled (to reduce vdroop) but have both powersaving features on in bios (C1E and EIST).

Idle temps between 30-35C depending on core, Core Temp, Tj Max 90C.

Using linpack (IntelBurnTest 2.1), 10 runs, to test stability.

Now looking at the above you'd think I'd have a sweet chip here. I've been trying to keep it as cool and quiet as possible at 3.0, which is why I'm trying to find the lowest vcore voltage I can get.

3.2 however, requires some serious voltage increases. Is this normal, considering how low it is for 3.0?

Core Temp shows vid at 1.1625v at idle and 1.325v at load. This is normal right?

I haven't tested since I was ok with vcore at 1.26, but if I raise VTT to say 1.3v, I should be able to lower vcore even more? Is this worth testing if I'm going for low voltages?

Blitz KriegeR: If you don't mind, can you try a 3.0ghz overclock and see what your minimum settings are?

 

[Blitz KriegeR]

Sorry. kind of a bad time for me. Moving this weekend so things are a mess; but I can give you the data I've already recorded.

Of course I started with locking my RAM 1:1, loosened all timings and defaulted all voltages to begin the testing. I'll assume you're already aware of all that.

I was liberal with voltages within "safe" levels for basic stability testing, to help eliminate all failure points other than the CPU. The voltages I changed throughout the duration of the test was MCH from 1.25 (default) to 1.35 when FSB was raised to 400 (turned out I didn't need it in the end), VTT from 1.20 to 1.30 (again, turned out I didn't need it in the end), and of course Vcore as needed.
*Note that I favor the point of view that vdroop exists for a reason, and would rather not subject my chip to over-voltage spikes of up to 1.6v; regards of how many milliseconds they last. Therefore LLC is disabled in all my testing. C1E is also always disabled for testing.

The first thing I did was bump FSB up to 300x9, stable at 1.25v (never tested any lower because that's what my chip runs at on default settings.) {Stable}
Bumped up to 333x9 @ 1.3625v, still BSOD on boot. {Unstable}
Dropped to 400x6 @ 1.25 to test FSB stability. {Stable}
Increased to 400x7 @ 1.30v {Stable @ min voltage}
Increased to 400x8. 1.425v required to POST, 1.5v for stability - 1.456v idle / 1.40v load {Stable @ min voltage}

Reading your results it looks like you chip is already a good deal better than mine, since I need way more than 1.25v for anything over 2.8Ghz.

So anyway, my advice is just be patient, use trial and error, and don't be afraid to fool around (within limits of course!). I will get back to testing my CPU at 333x9 at some point, but probably won't be able to publish the results until I get internet at my new place (about two weeks *sigh*).

Best of luck, and happy OC'ing!

Edit: Yes from what I've seen/heard most Q6600 chips need huge voltage boosts in the 3.2Ghz+ range. 2.8Ghz - 3.2Ghz seems to be the "easy peasy" low voltage OC range for this chip. Again, mine needed a fair bit more voltage that most.