4790K Overclock

[CarbonSteal]

Hello everyone,

I have a 4790K laid inside a Gigabyte Gaming 7 Z97X. The chip runs stable at 4.7 GHz and 1.275 V 24/7, although I have been running 4.5 GHz at 1.24 V with offset typically.

As previously mentioned, I have had no issues with the overclock's stability with actual tasks, only waking out of sleep. I know most overclockers don't value sleep, but I do. I have tried fixed voltages with the stock 100 MHz BClock, and offset from the stock VId. Either overclock will wake out of sleep successfully 2 or 3 out of 5 times. The others will result in a BSOD. Every single BSOD has the stop code: 0x8000400000000002, which also happens to be the stop code I get usually when my VCore is knowingly too low. If my overclocks are stable up top at a fixed voltage, how can the sleep states not be? It seems voltage related. I have not messed with Load Line Calibration or PLL Overvoltage. I have heard that you must disable PLL Overvoltage to get Sandy Bridge to sleep. Is this true? Same with Haswell?

My PSU is an EVGA G2 750 W, which EVGA swears is "Haswell-compatible" (C6/C7).

I like to know how everything works, so can someone explain to me why the PLL Overvoltage would need disabled, if relevant at all?

Thanks for any insight.

 

[Tír na nÓg]

REduce LLC and iincrease offset (- 1 notch llc=-0.005/0.010 vcore on load more or less, depending on the motherboard).

 

[CarbonSteal]

Stable Out of Sleep Now

REduce LLC and iincrease offset (- 1 notch llc=-0.005/0.010 vcore on load more or less, depending on the motherboard).

Thanks for your fast reply. I set the load line calibration to the second highest setting, and gave to +20 mV more fixed than what I had previously given it. No issues coming out of sleep.

Is there any way to increase stability out of sleep without increasing the voltage that much?

 

[Tír na nÓg]

You're welcome.

You mean +0.020, right?

CHeck the vcore in CPUz: goal is to have the same vcore on load than previously, with lower LLC. Which will give higher vcore when idling.

 

[CarbonSteal]

You're welcome.

You mean +0.020, right?

CHeck the vcore in CPUz: goal is to have the same vcore on load than previously, with lower LLC. Which will give higher vcore when idling.

Yep. +20 mV = +0.020 V.

Having the LLC higher actually seems to help sleep stability at the same VCore.

 

[EarthDog]

It raises the vcore on load...llc... you apparently have some vdroop on load.

 

[Tír na nÓg]

Yep. +20 mV = +0.020 V.

Having the LLC higher actually seems to help sleep stability at the same VCore.

Well, sorry, it starts to be late here, heavy day, lol! Missed the"m"

As E_D says : LLC compensates for VDrop (vcore reducing on load)..

 

[EarthDog]

Vdroop (2 o's) = voltage difference between windows idle and load.

Vdrop = difference between whats set in bios and idle in windows.

 

[Tír na nÓg]

Vdroop (2 o's) = voltage difference between windows idle and load.

Vdrop = difference between whats set in bios and idle in windows.

Thanks for clarification

 

[CarbonSteal]

Vdroop (2 o's) = voltage difference between windows idle and load.

Vdrop = difference between whats set in bios and idle in windows.

At 4.7 GHz offset, CPU-Z is reporting 1.279 V under load. With a 22,000 count multimeter, I measured 1.3033 V. Guessing this is due to the LLC being set to high. So the real voltage difference is like 0.02 V.

I'm pretty happy with this result now, i'm running 2133 CL9 RAM too.

Anything below 1.35 V isn't bad for Haswell daily, is it? People run Skylake at these voltages. I think people were just conservative due to 4770K cooling issues?

 

[Tír na nÓg]

No worries : all good below 1.35v!

Nice overclock you have here, and nice fast DDR3! This PC shall serve you well for quite some more time!