OC-ing a E6850 on a IN9 32X Max

I'm having a lot of trouble trying to reach 3.6 GHz with my E6850 on the IN9. I've tried both BIOS 1.2 and 1.4. My voltages seem way to high for Prime95 to be having errors at such a low OC. Would a corrupted Windows Vista installation cause Prime95 to have errors? I rarely get a BSOD or OS freeze. When I say Prime95 has errors, i mean i can't get past 5 minutes of stress testing.

Prime95 is failing trying to OC with the following settings:
My BIOS (version 1.4 or w/ 1.2) settings thus far are as follows:
OC uGuru Utility Menu:
CPU Operating Speed - Linked
FSB <QDR> - 1600
Multiplier Factor x9
FSB: Memory Ratio - Sync Mode
N/B Strap CPU As - PSB1333
NB PCIEe x16 Clock - 100MHz
SB PCIEe x16 Clock - 100MHz
SB PCIEe x8 Clock - 100MHz
SPP<-> MCP REF CLOCK - 200 MHz

Voltages Control Sub Menu:
CPU Core Voltage - 1.47V
DDR2 Voltage - 2.1V
CPU VTT Voltage 1.2V
NB Voltage - 1.5V
SB Voltage - 1.65V
HyperTransport Voltage - 1.25V

My components are as follows:
Mother Board: ABIT IN9 32X-MAX LGA 775 NVIDIA nForce 680i SLI
Case: COOLER MASTER Stacker 830 Evolution
Processor: Intel Core 2 Duo E6850 Conroe 3.0GHz 4M shared L2
Ram: OCZ 4GB(2 x 2GB) 240-Pin DDR2 SDRAM DDR2 800 (PC2 6400)
Power Supply: PC Power & Cooling Silencer 750 Quad (Black)
Graphics Card: Shader OC 1650MHz MSI NX8800GT (2 in SLI)
After Market Heat Sink Fan: Tuniq Tower 120 P4 & K8 CPU Cooler

Thanks for any help in advance

Is it failing small FFTs or bigger ones?

Moffetts said:

Is it failing small FFTs or bigger ones?

Click to expand...

When I'm running the torture test, I always chose the "Blend" option.

Which tests at 1024k first, correct? A run of any given size lasts around 15 minutes, so you are failing during a RAM critical period. Try loosening your RAM timings and giving them a little more voltage if the sticks will allow you to (don't give them more than they're rated for.) If that doesn't work try upping your NB voltage a peg.

1024k still tests your CPU, but it is more RAM-oriented. Even if it fails during this test, it could still be a CPU issue.

The sticks are rated for 5-4-4-15 @ 800MHz. I've failed that test with 6-6-6-23 as well as leaving it on Auto. Is there a different combination of timings I should try?

Wow, those are loose. Do they pass memtest?

I am memtesting now at those timings and report back

On Test #7 there were 154 errors with the 6-6-6-23 timings with the 3.6GHz OC.
What does this mean? What should I change to overcome these errors?

Would a memtest error like that relate to the NB,SB or DDR2 voltage?

Try the test at stock memory clocks and voltage settings.

Ran fine with stock memory clocks.
In the advanced settings for the ram, the timings were all left at auto.
I changed the tRFC from auto to 42 and then I got OCCT (30 minute test) to run stable. Prime95 seems to run blend tests stable now too.

With the E6850 @ 3.825 GHz, the 8800GT's in SLI, 4GB of Ram (stuck at 2.8 in Vista32), I was able to land about 16.7k in 3DMark 06.

I think i should be good on my own from here. It looks like setting that single ram setting ultimately was holding me up from keeping a stable OC.

Why is that? I saw someone else set it to that setting on another board and they said they got stability as well so I figured I would try it and low and behold it worked...