Overclocking 2600k - Suggestions needed

[T101]

I have been using the auto tuner to get 4.3Ghz, then tweak those settings to get a stable system. However I want to reach 4.5Ghz, and I have. But I do not like the voltage spikes I get with the settings I have been forced to use to get it stable at 4.5Ghz.

CPU: Core i7 2600k
Motherboard: P8P67 Deluxe Bios 2001
CPU Cooler: Corsair H100

BCLK - 100
Multi - 45
Vcore: Manual mode, then setting the value to Auto (gives it a 1.28V to 1.32V during load).
ATV - changed from auto to 0.004 (least possible, since auto gives even bigger voltage spikes).

Using AUTO for the manual mode vcore value makes the CPU lower voltage when idling at 1600 etc. Setting the Vcore manually to 1.250V or 1.288V results in a BSOD when running Intel Burn Test.

I have tried to use the offset mode to set vcore, but I have been unable to find exactly how to get it to give me the vcore I want, since it tends to overvolt the chip quite a bit no matter what I try to do with it.

I would appreciate recommendations on what settings to use to get the cpu running at 4.5Ghz with the lowest possible Vcore spikes.

 

[Zap]

If it only takes 1.32v to hit 4.5GHz stable, I'd say you were doing fine.

 

[T101]

Great then. I was expecting to hit no more than 1.28V for 4.5Ghz.

 

[bryanW1995]

Why were you execting to get there with only 1.28v?

 

[T101]

Because I read a post on HardOCP by an asus representative that specified that as vcore för 4.4Ghz to 4.5Ghz. Thats why.

 

[mrjoltcola]

From my experience, it takes a top 10% chip to get 4.5 @ 1.28v. My testing on a handful of 2600K / 2700K showed that about max speed on average at 1.28v is 4.4Ghz. Usually 4.5 requires from 1.29-1.33v.

I usually don't test the odd multipliers, but what I have is a lot of data on 4.6, 4.8 and 5.0.

What I see on my 5 chip sample for load-tested stable data:
44x - 1.28, 1.28, 1.30 (did not test 44x on two of the chips)
46x - 1.30, 1.31, 1.32, 1.35, 1.37
48x - 1.36, 1.37, 1.39, 1.40, 1.41
50x - 1.44, 1.465, 1.49
51x - 1.465, 1.51

Only 2 of 5 chips will boot and run games stable for an hour at 52x or 53x at low voltage (near 1.5v) but Prime95 would crash.

So as you can see, for your desired 45x @ 1.28v, you'd have a better chip than the best one here.

Are you trying to stay under 1.3v? If so, you should be satisfied with 44x.

I recommend 1.35v, the sweetspot for these chips is around there, at 46x - 47x

 

[T101]

Thank you for your feedback on voltages. I am happy with the performance and thermals I am at now. But I will see how far I can take it just because I want to know the limit of this cpu.

 

[T101]

turns out that after two hours of prime95 i got a BSOD. Trying 1.35V now.

 

[mrjoltcola]

If 1.35v is stable, and I'm sure it will be, drop back and try 1.325v, and so forth. It only takes a 0.005 increment once you reach the right voltage. 2 hours of Prime95 @ 1.32v sounds like you are almost there, so you shouldn't need to jump all the way to 1.35v.

 

[T101]

1.32V manually set was stable running prime95 all night.

Now I am trying with Vcore offset of -0.005 and 50% Load Line Calibration, running at 1.30-1.31V on load. Before with Auto voltage and auto load line calibration, the Vcore would drop down to 1.28V during load, which I guess was to much of a vdroop.

 

[T101]

One hour Prime95 stable at least. Dont have time to run any longer than that now, before it could not do more than 30-40 minutes before bluescreen. Will make an overnight run tonight. At least I got the vdroop lessened and the max vcore down a little.

 

[T101]

Overnight Prime95 success.

 

[Lex Luger]

Thats good

these chips can take about 1.4 volts on good air cooling, maybe 1.425 if yo uwant to push it.

I say if you have the cooler for it, try to get your chip up to 4.7

That should be a good speed for you and your chip 24/7(if your cooler is up to the task)

 

[Sunburn74]

What do you mean by voltage spikes btw?

 

[T101]

By voltage spikes I mean that the Vcore during shifting load (from fully loaded to idle or from idle to loaded) reaches a very brief higher voltage, in this case about 1.36V. It is the opposite to vdroop. Enabling load line calibration reduce the vdroop while setting an offset -0.005 voltage instead of just selecting (Auto) or manual entering or using (auto) setting, reduce the voltage spikes when the cpu changes performance state. Thus it runs cooler and should survive longer.

 

[itstsui]

If you bring it up to 1.5V you can clock it at around 5.1Ghz

 

[T101]

Well, I dont want to push the Vcore to the max, because I dont want the ship to burn out in a few months only. I try to play the middle field of the highest performance within a reasonable vcore. Just because you CAN put 1.5V through a sandy bridge, does not mean it can take that for any amount of time. If it could, Intel would churn out 5.1GHz chips at a great pace. On stock speed, the turbo brings the CPU up to 1.288V vcore at 3800Mhz. 1.32V is a marginal increase, that is not unreasonably beyond the allowed fault tolerance for Vcore, and it brings me 4.5Ghz now. In time the chip will degrade, that is a fact. And I will either have to increase Vcore or reduce clock frequency. But by trying to balance the two somewhat, I try to push that degradation problem as far as possible into the future. As simple as that.

 

[T101]

These are the settings I used to get it running at 4.5Ghz 1.32V

AI TWEAKER
Bclk/PCIE Frequency: 100
Turbo Ratio: By all Cores, 45
Internal PLL Overvoltage: Enable

Load-Line Calibration: High
Vrm Frequency: Manual
Vrm Fixed Frequency Mode: 350

CPU Voltage: Offset Mode
CPU Mode Sign: -
CPU Offset Voltage: 0.005

CPU POWER MANAGEMENT
Additional Turbo Voltage: 0.004