Overclocking and Prime95 - my first major overclock - need help please

[Zithras]

Computer I'm making:

CPU Q6600 G0 SLACR (lapped)
MB: Gigabyte GA-X48-DQ6 (unfortunately, it's so new that its hard to find info on)
cooler: TRUE (also lapped) with scythe 120mm 100+ cfm fan, on tight with 'penny' (well...dollar coin) mod
PSU: PC Power and Cooling S75Qb 750W
video: Zotac GeForce 8800GT 512 MB
ram: G.Skill 4 GB (2x2gb) DDR2 1066 (mfr. specified voltage is 2.1, with 5-5-5-15 timings, when the motherboard is set to autodetect timings, it gets 5-7-7-20, right now I have it dialed in to the recommended 5-5-5-15 timings, but running at 800 instead of 1066 for a 1:1 ratio)
fans: 5x Thermaltake A2016 80mm variable case fan

I posted earlier about a heating problem, and later figured out that the Xigmatek heat sink was not clamping hard enough onto the processor, and two pushpins refused to hold. So, I got a TRUE, and that problem's solved - no more overheating 😀. On to the next problem!

Since the BIOS only lets me set ram timings corresponding to 2.4, 3.0 and 3.6 GHz CPU FSB speeds, I'm trying for a 3.6 overclock, which should be more than doable with this board and cooling setup - but it's not working (I'd be happy with 3.4, but the board doesn't support it, and I don't want to go as low as 3.0).

I set the stock voltage rather high, at about 1.481, (stock for the 6600 is 1.3, with Intel stated max of 1.55) and started up the computer, with the intention of slowly working my way down voltage-wise until the computer became unstable. The board posted with no problem, but I started getting errors when running Prime95 (large FFTs, 4 threads). I assumed the voltage was still too low and adjusted higher. Same story when I upped the voltage to the low 1.49s. Once I hit 1.50, however, instead of giving me an eventual rounding error, the computer just started locking up after about 2 minutes! No error message, no warnings, just a lockup. I raised the voltage again to 1.51 and the same thing happened.

At no time did the CPU core temperature go above 68C, so I doubt its overheating.

Any ideas as to what I'm doing wrong? Are my voltages too high? Too low? (Is it even safe to put a Q6600 at/above 1.5V even if it's not overheating?)

How can I tell if my voltages are too high or too low?

Anyone with a similar setup have an idea on what voltages I should try?
Any other information I should post?

Thanks for your help!
Zithras

 

[Lupiron]

First off, you dont wanna use Large FFTs for anything. Small for CPU testing. Blend for RAM testing.

Your board will let you select other memory speeds, somewhere in there.

You also may need to establish what your vdrop and droop are, if there are no corrective measures in your Bios, like Load Line Calibration. Go into the Bios and select 1.3000 at the Bios VCore setting, boot at stock with speed step and EIST disabled and when you get into windows, run CPUz and note the idle value in windows of the Core Voltage.

Then run Prime 95 on all 4 cores, Small FFTs torture test! (You'll note that it usus more power and gets higher temps than large ffts.) And after a minute, note the now diminished VCore vlaue. You want to see the lowest it flickers to. List those here. A 1.3000 VID processor will need roughly 1.44 volts while fully loaded on small ffts.

Your temps are fine.

Anyways, you have to disable all the processor options, c1e, vanderpool, disable bit, etc. Then you have to disable all spread spectrum controls, lock the PCIe bus to 100 Mhz, disable any static read control, or transaction booster.

You may also wanna let the ram boot at spd for a while, until you get the CPU voltages and how your board applies vdrop and droop worked out, and once the processor is stable, the ram is usually easy, because you'll know where the problem lies.

(Large ffts uses both ram and some cpu power, so using it as a test doesnt tell you what one is causing the problem.)

The end result should be FSB 400 with a x 9 cpu multi, NB to 1.40 or so, Ram linked and synced with the FSB. (That'll give you your 800 Mhz.) You should have a strap setting, or unlink option to select other speeds besides for sync mode, but Sync is usually the best way to set it.

--Lupi

 

[Zithras]

Memtest86+ ran 8 clean passes, so that's not the problem. I've messed with settings for awhile, with no luck. I guess this particular CPU just won't hit 3.6. (I tried everything suggested, disabled all processor options, etc.)

Thanks for the information about Prime - I'll use small from now on.

I'm thinking I'll set it at 3.4, and unlock the memory ratio/set the memory to 1066 dual channel.

Does anyone know how to enable asynchronous memory multipliers (i.e. get them to show up) in the AWARD BIOS used in the X48-DQ6? I.e. I want the RAM linked but unsynched to the FSB. (The default synch settings only apply to 266, 333, and 400 FSB)

 

[Mondoman]

...
I set the stock voltage rather high, at about 1.481, (stock for the 6600 is 1.3, with Intel stated max of 1.55)...

Actually, the Intel stated functional max is 1.45V, with 1.55V being the voltage at which permanent damage to the CPU is likely.
I'm not sure why you started with a voltage higher than your VID -- why not start there and work your way up (voltages that are too high can cause problems, too).

 

[Shadow703793]

^+1.

45nm: ABSOLUTE MAX is 1.45
65nm: ABSOLUTE MAX is 1.5v

 

[Mondoman]

shad - read the Intel manual. 45nm max voltage is 1.35V IIRC, with 1.45V being the "absolute maximum voltage", above which permanent damage to the CPU is likely. Similarly for the 65nm CPUs, but 0.1V higher: 1.45V max voltage, 1.55V absolute maximum voltage.

 

[Shadow703793]

^I should have been more clear. By MAX I was referring to ABSOLUTE MAX. sorry for the confusion 😉.

 

[Evilonigiri]

I'm not sure if too high of a vcore will cause issues with the cpu. I know it does for ram though...

 

[Zithras]

The main problem is the high heat high vcore generates. Assuming you don't put it insanely high and fry the CPU, high vcore will cause more processor heat, which will cause unstability and increased degredation. This is why you can overclock more on water than on air (better heat dissipation).

At least, this is how I understand it.

Edit: As a quick update, it seems as if 3.5 is the best this chip can do 🙁 This is sad. Even then, it will barely hit this, with full fans enabled (noisy!)