Well done. Don't forget to post up some screenshots too.

Ur temp is normal. Ambient & casing airflow affects the temp too.
Hmm don't think u can do 4.0Ghz with stock cooler unless in non air cond room.
U might google for "x58aud3r 4ghz bios", just make sure that HT is turned off 1st.
anyway to OC I highly recommend grabbing a good 3rd party cooler ie. Corsair H7O/H5O, Thermalright Venomous-X, etc..
with these coolers, OC is easier & more fun

Suggest to go straight to 4Ghz.
Prime at least 4 hours or LinX for 20 runs.

ur dram timing should be 9-9-9-24-88-1T
and run LinX to the max memory in normal mode or full amount in diagnostic mode
max memory settings, click "All" button or pick the last memory size option from the drop down menu, the "All" option is highly recommended
20 runs should last about 1 hour
for diagnostic mode, go to START menu > Run (or hotkey win logo + R) > type msconfig > diagnostic startup

There are no performance effects of a higher QPI speed on any single socket system I've ever heard of.

QPI has nothing to do with memory performance, CPU performance, or any thing else besides connecting the CPU to the chipset, or to another CPU (in the case of high-end Xeons).

It is possible to take advantage of DDR3-2000+, but you need an extremely high uncore clock. Again, QPI doesn't matter for this.

Slower QPI mode could help in achieving higher BCLK, easier.

Some high end mobos only. FYI ur cheapo P6T doesn't hv this option.

Ur vcore also might be insufficient.

HT advantage/disadvantage depends on games, but the difference is pretty much negligible ~1% favorably HT off.
For me, I'd turn off HT and OC core clock higher since turning off HT will reduce core temps by huge margin.
But when it comes to 8-thread intensive apps, this HT feature will show its advantage.
From ur 3rd pic, it shows that ur dram is rated at 8-8-8-24 ~1.65v
But where's the 5 hours stability screenshot?
Btw, u can combine all the CPU-Z tabs into one screen by multi launching the app a few times at the same time.

Played with this batch b4.. core 4.2Ghz HT off, uncore 4.2Ghz @ 1.275v.
For me it performs like any other 920 D0's

anyway depends on ur dram and mobo as well bro.. in my case I'm using DomGT 2kc7 and Classy
vcore @ 1.275v, vtt @ 1.425v, altho not as low as my current 920 which requires 1.25vcore and 1.4vtt only.

Let the 5-hour priming screenshot prove stability,
BSOD means ur OC is not stable yet.. try upping more voltages ie. vCore, vQPI/RAM (vtt) etc.. try and error.
Large FFT is meant to determine ur max heat, power consumption, stress some ram only. Use custom blend on 5400MBs of ram to stress core, uncore & dram entirely. U will fail in 5mins, 30mins, 1 hour, 2, 3 etc...

Good job! so how's the game still crash on u with these latest settings?
Ur temp still allow higher OC good luck

This post has been edited by owikh84: Oct 11 2010, 12:01 AM

Passed prime95 but crashed during gaming.
btw what about other games?
From what I heard SC2 will crash on some custom map
Bad GPU driver also can cause BSOD.

CPU Clock Skew.........................[Delay 300ps] <-- 100ps is enuf

Voltages do affect GFlops as well bro.
Increase vCore without HT and compare.

---

Added on October 18, 2010, 6:52 pmResume mode related in bios.
Enable wake up by usb kb/ mouse.
Other than that press pwr button on ur case.

This post has been edited by owikh84: Oct 18 2010, 06:52 PM

I wouldn't call it as an calculation error. It is about the correlation between GFlops and voltage efficiency.
The most efficient voltage will yield the most optimal performance.
LinX is used more for figuring out if u're close to optimal, efficient vCore.

No and No

It's not a bug but it's called the Intel SpeedStep Technology (EIST)
To do 200x20=4.0Ghz u hv to turn off this CPU feature in BIOS.
EIST will bump ur multi from 20x to 21x that's why u got 200x21= 4.2Ghz

Some of the reasons why u got different temps:
Diff proc diff exothermic
Environment (aircond/none)
Casing airflow
Extra active cooling (table fan)
Thermal compd (TIM)
voltages, VTT, etc...

Solution:
Get an aftermarket cooler and clock at least 4Ghz

slightly bumping up the QPI PLL & IOH/ICH I/O volts

i7 920 D0 @ 4.5Ghz HT OFF
EVGA Classfied E760 Bios 73
Corsair DomGT 2kC7 @ 2150Mhz 8-8-8-20-1T, RTL 60/62/64, vdimm 1.61v



This post has been edited by owikh84: Oct 23 2010, 03:48 PM

Understanding EVGA X58 Classified Voltages

CPU VTT: Stock VTT voltage defaults to 1.17V or so. The scale runs in 10mv steps, allowing you to add or subtract from the base value. VTT voltage drives the integrated memory controller and the QPI link bus. This voltage will need to be increased as you increase memory frequency and increase BCLK (QPI frequency). Remember to keep this voltage within 0.5V of VDIMM to prevent damage to the integrated memory controller.

CPU PLL VCore: 1.8V base; we found no benefits to increasing this voltage for overclocking. Clean power to PLL circuits is essential for ensuring that external influences over the accuracy of clock signals is minimized. It seems EVGA have done a good job with regards to all PLL voltage rails on the Classified. In terms of overclocking headroom and PLL's, clean power always trumps more voltage, thus there's no need to tinker with these voltages to any great extent on this board, apart from perhaps QPI PLL. Even then, you'll probaby find the level of voltage required for maximum BCLK potential is far below what some other boards need to get the same clocks...

IOH PLL VCore: Input/Output Hub Phase Locked Loop voltage. The default voltage is 1.8V. For the most part this voltage can be left at stock unless chasing maximum QPI frequencies for benchmarking. For 24/7 use we recommend you stick with a maximum of 1.89V. Processors/IOHs needing more than 1.89V to remain stable are best run at a lower BCLK and higher CPU multiplier to bring them back into their comfort zone.

QPI PLL VCore: Quick Path Interconnect Phase Locked Loop voltage. For 24/7 overclocking this voltage can be left at stock in most instances. Again, if you need to use more it's probably wise to pull back a little on QPI frequency to a speed that requires no more than 1.2V. For benchmarking, we used 1.45V to reach 239 BCLK for a CPU-Z shot.

IOH VCore: Input/Output Hub voltage uses a 1.10V base. Stock values usually suffice for 24/7 overclocking even when using multiple graphics cards; we had no problems in reaching a stable 200 BCLK X20 for 4GHz CPU speed. For more extreme QPI frequencies this voltage will need to be increased. For 3D benchmarking past 220 BCLK we used 1.45V. We needed 1.50V to reach 239 BCLK. Values over 1.50V failed to POST or locked up in the OS; use only as much voltage as you need.

IOH/ICH I/O Voltage: IOH to Intel Southbridge termination voltage. We left this voltage at stock and did not see any overclocking improvements from increasing it on this board. Base value is 1.50V.

ICH VCore: 1.05V base voltage; again this voltage can be left at default.

NF200 Voltage: 1.2V stock. We left this voltage at stock for all overclocking.

VTT PWM Frequency: Changes the switching frequency of the VTT voltage line. Higher switching frequencies can help facilitate lower output ripple, and reduce overshoot while providing a higher frequency power bandwidth to the CPU. Setting a higher frequency with this BIOS option may improve overclocking headroom somewhat depending on the current drawn. For overclocking past 220 BCLK we generally used 490KHz; for speeds below this a 250KHz switching frequency worked fine for us and also reduces the heat output from the PWM FETs.

CPU PWM Frequency: CPU PWM (VCore supply) switching frequency. Stock is 800KHz, which is more than sufficient for 99% of overclocking. For subzero benchmarking you may wish to increase the switching frequency to see if it helps stability during heavy 8-thread CPU loads. The side effect in doing so is increased heat from the power FETs, although the temperature increase is quite small due to the use of a 10-phase power delivery circuit, especially as the FETs are indirectly cooled via the CPU power plane when the processor temperature is well into the negative region.

CPU Impedance: Sets the level of signal compensation for the QPI bus to the CPU. Higher QPI frequencies often demand a higher signal current, which means that compensation levels have to be changed in order to counter any signal line reflections caused by PCB traces and input impedance mismatches. The options available for this function are AUTO and Less. On the Classified, we found that QPI frequencies approaching 4GHz can benefit from a reduction in setting Less, even more so if the CPU frequency is well above 4.5GHz. Leave at AUTO for most 24/7 overclocking at BCLKs under 200MHz. For subzero cooling at high system speeds, set this value to Less to see if it helps with overclocking headroom.

QPI Compensation: Sets the level of signal compensation for the QPI bus to the IOH. There are three options for this function, those being AUTO, Less, and More. We found that IOH compensation is far more sensitive to changes in compensation values than the CPU. A setting of Less for high BCLKs and QPI frequencies almost always brings about additional system stability during benchmarking (at least in our experience). For all other purposes, this value can be left to AUTO.