Low Vcore

[discors56009]

I have an i7 3770k that I have overclocked to 4.0Ghz. With auto Overclock settings only changing the clock multiplier to find my VID at 4.0Ghz and running Prime95 Blend mode, my Core Temp read temperatures of 95C+ on 2 of the 4 cores within ~7min. This was at the native core voltage of 1.2510. After researching I found that the Ivy bridge processors run very hot and require quite heavy duty coolers that put my Corsair H60 Liquid Cooling system to shame. However I also read that the i7 3770k is capable of Overclocking to high potentials with fairly low core voltages.

So I began testing my i7 to see how low I could push the core voltage levels to maintain better core temperatures.

~4.2Ghz is what I will begin with. I won't mention temp reads at 4.2Ghz+. AND 4.5Ghz? . At 1.150v I read core temps of 95C+ within 10min. Failed at 1.100 which 'might' have yielded manageable temps.

~4.1Ghz @ 1.15v ran core temps of 95C+ within ~12min. *I'm thinkin.. 'PIECE OF GARBAGE H60'*. 1.10v ran core temps 95C+ after ~20min.

~Moved down to 4.0Ghz because i accepted the insanity that is Ivy Bridge. Said.. what the heck lets go 1.05v! ... Ran beautifully. Never above 77C after 30-45min. Stamped my approval.

Core Temp gave me a VID of 1.2510v at 4.0Ghz and my desired Vcore was 1.05v after fixed mode testing. I used this to set my Offset value for -.200v. However, I feared that this big of a negative offset would cause the CPU to have trouble booting because of such a low idle Vcore.

The computer booted just fine and Iv'e been on the forums for awhile barely using any voltage.

I guess my question is more of approval of what I have done for my processor given my current cooling system (which I plan on upgrading to push/pull ASAP after all this testing.).

 

[discors56009]

After a bit more tinkering I've managed to get it stable at 4.2Ghz with ~1.110 core voltage. Max temp read on 2 of 4 cores was 90C after 30 min Prime95 max load CPU stress test.

Max while 3DMark 11 Benchmark Performance testing was 79C.

This is as good as it gets for an ivy bridge with mid-level liquid cooling I suppose.

 

[Scotty99]

90c is wayyyy too hot!

I have a 2500k @ 4.2 on stock voltage (1.296) and it never goes above 60c even at max load in prime 95.

I dont think your CPU cooler is making proper contact with CPU, try and reinstall it and see if that fixes it.

 

[discors56009]

The Ivy Bridge generation of processors is actually very widely known to get sometimes 20C hotter at the same clock speeds and OC frequency as their Sandy Bridge predecessors.

The T Case (Manufacturers guideline max temp that can run consistently without shortening the life-span of the processor is 67.4* C. The Tj. Max or max temp that it cuts off at to prevent damage is 105*C and it begins throttling at 95*C+.

It has stayed stable throughout the night, never going above an idle temp of 32* C.
The max temp I read while playing The Witcher 2 (One of the most taxing PC games on the market) on Ultra settings and Ubersampling is ~69*C. Averaged 59-60 throughout the entirety of the play session.

Correct me if I'm wrong but for the i7 3770k, these temperatures are normal and OK to run at. The i7 predecessors before Ivy bridge just plain run cooler; but with the 22nm technology and 8 thread hyperthreading its a beast of a processor.

 

[steddora]

The 3770k isn't that much of a boost over a Sandy Bridge 2600k in all reality. The real performance actually lies in the other additions, especially PCI Express 3.0, Ram controller, and the IGP unit on the chip. However, the 22nm process runs at a high voltage for it's size where more heat is bound to be produced between the transistors. Remember the older 65nm process where 1.5V didn't hardly produce any heat? Well remember, the smaller the process the less voltage it's going to take to produce massive amounts of heat. This is one of the reasons the SB overclocks so much better than the IB along with other simple things such as more consistency in the heatspreader installations, less reports of good/bad bin chips, and most importantly the lower voltage for the 32nm process.

Take all this into account, yes you have a hotter running chip that performs a little better than the previous generation.

You're problems stem more from voltage and cooling. The H60 isn't a very potent cooler and really shouldn't be held accountable for high overclocks at all. Closed loop coolers (CLC's) like this aren't nearly worth the price as similar/less priced air cooling can perform the same or better.

Setting an automatic voltage on a processor that uses variable voltage from VID can actually have really bad impacts when overclocking. Remember, the chip will want more power if it's clocked to 4.0Ghz rather than it's 3.5Ghz stock speed. So if you're wanting to sprout and keep the variable voltage system (I highly recommend this if not pushing the clock limits on the chip), you'll need to switch to offset mode to help control the voltages. This can easily offset high voltages wanted through the VID and can give you a little control over your processor. My processor is a perfect example.

When I was stock with this cooler (Hyper 212+ w/ 2x FM121 fans 110CFM) at stock settings I ranged like this.

Idle-Load voltages : 0.925 to 1.375v
Idle-Load temperatures : 28C to 70C

I used IntelBurnTest as a test for heat on my processor. Warning, IntelBurnTest will produce more heat than ANYTHING you'll ever do to your chip. Example, the 70C I reached at stock with IBT... Using Prime95 Blend would only reach about 64C.

Now after I got to clocking I resulted in an actual undervolt from stock at a setting of -0.025v offset resulting in the following...

Idle-Load voltages : 0.950 to 1.325v
Idle-Load temperatures : 28C to 68C

Mind you this was resulted at a 4.4Ghz overclock which adds a full 1Ghz clock speed to the processor. Now if I were you, I'd go through the following steps.

1. Check installation of cooler, make sure thermal paste installation is perfect. Never pull the cooler off of the CPU without cleaning the CPU/cooler with isopropyl alcohol and a coffee filter. You must reapply thermal paste every time you pull the cooler off of the CPU. If you get a good install and still have heat issues, don't overclock too far or switch out for better cooling.

2. Check all voltage settings at stock and see where you can sit safely. Remember, more voltage means more heat no matter what you do. So try to keep the voltage down while overclocking as much as possible. You may be surprised and get an undervolt at a decent clock. Then again you might have to really push voltage to get very far.

3. Absolutely make sure all fans are in working order and that they and the heatsinks/components are clean and not covered in dust. Dust insulates and causes even more heat!

I hope this helps you out. Just because the IB chips run a bit hotter doesn't mean you can't keep them cool enough to overclock. Just watch the voltage and the temperatures and make sure you know exactly what you're toying with before you do.

 

[discors56009]

Best answer selected by discors56009.

 

[discors56009]

Thank you steddora, that was everything I was asking for and then some.

I have tinkered with my chip quite a bit and have settled down on 4.2Ghz with an offset of -.110v and 75% Load-Line Calibration (probably overkill for 4.2Ghz). My chip runs a VID of 1.256 at 4.2Ghz under full load. This offset and LLC combinations gives the chip a max Vcore of 1.112 under full load. Prime95 Blend Mode got the processor to reach 90C max on 2 of the 4 cores after 30-45min. Never above.

I was worried with that this much of an offset the chip would not boot because the Idle Vcore would be too low. This is not the case however and have kept the computer on/sleeping for 48 hours now.

In conclusion, this is the max I will push it until I get a better cooling system. Aiming for push/pull system to allow me to push to 5.0Ghz sometime in the near future. Will post results.