Vcore and overclocking question

[tethlah]

Something I'm doing is building out a liquid cooled system that will include an i7-3770k. I've been doing a lot of research on what is a safe VCORE for overclocking and I've been getting nothing but numbers with no theory or explanations to validate the numbers or help me learn why what they chose. So keeping in mind that I have an unlocked CPU that I can adjust clock and voltage as I see fit, how do you figure out what i safe on the CPU? I'm looking for performance, and since I'm liquid cooling in a room that is an average of 65 degrees year round with no humidity, I should be able to regulate temperature just fine. (I'm running a dual 120 rad and a triple 120 rad in my loop)

So when it comes to overclocking and VCORE, what is safe, where do you see lifetime degradation vs performance degradation? Is there a good practice or solid numbers on this specific CPU that I'm missing? And I guess in general what, if any, increase will I see upping the VCORE when overclocking the clock? I'm thinking I should be able to go up to 4.5 on the clock. (or do you think i could go higher/lower?)

[dandelioneater]

Originally Posted by tethlah

I'm thinking I should be able to go up to 4.5 on the clock. (or do you think i could go higher/lower?)

Mine runs at 4.5GHz with 1.200 Vcore set in the bios and Load line calibration set to the second highest setting. You shouldn't need to go higher than 1.3V to achieve a few 100MHz higher. I am 100% stable on air and I think that this is a very safe overclock. Temps range from 50-60 degrees during stress testing.

Keep in mind that these chips get hot when you start going past 4.6GHz/1.3V. It is due to the tri-Gate transistors used with these Ivy Bridge chips iirc. So as long as you aren't going past 1.35-1.4V and temperatures are not much higher than 70-80 degrees during stress testing you are ok. Running the chip hot with 1.4V fully loaded 24/7 will surely lower the lifespan of the chip. (You will probably replace your setup before it 'dies' though)

[tethlah]

what i figured i would do is keep records for a couple weeks for baseline at air cooled, then see what percent drop i get out of liquid cool, then try to go to about 4.2 or 4.3 to see what the temps are like and step it up every week and see how it goes. Seem like a normal way to do it?

[Rasparthe]

The temps shouldn't change from day to day, what the temps are under load after 15 minutes will be same (or should if you have the cooling setup properly) as after an hour. You don't really need a week to see the temps. Get a program like Prime 95 or OCCT (which actually provides some handy graphs during the testing) and run it for an hour then you will see what your load temps are like. If they are getting to a point that you find uncomfortable then tone down the volts (which likely means your overclock) until the load temps sit where you like them.

As was said above, Ivy Bridge chips have a real 'cliff' when it comes to temperature. You can increase voltage up to a certain amount with only minor increases in temps and then suddenly each tiny bump in voltage will bring large jumps in temperature. This usually occurs around the 1.45v range if I remember correctly. The Ivy is pretty hardy though and I've heard (although I wouldn't if it was me) that some people run regular into the 70-80C range with their daily clocks.

Every chip is different and that is why you will see some people binning chips to find the ones that give the very best performance at the very least amount of volts and why you won't really see anyone giving numbers on voltage limits or preferences. Most people have their own personal beliefs/preferences about volts/temps that they stick with.

Good luck with your loop, post some pics!

[Generic User #2]

weren't those voltage/temps concerns only for air setups? if he's on liquid, he should be to go above 1.3v and get some pretty decent results?

[dandelioneater]

Originally Posted by Generic User #2

weren't those voltage/temps concerns only for air setups? if he's on liquid, he should be to go above 1.3v and get some pretty decent results?

Sure, water is going to keep the chip cooler than air, but using higher voltage than the factory spec will still have an effect over time. If you plan on keeping the PC for a long time then it may be wise to run a more mild overclock.

This is purely speculation, but I bet even with a custom loop running an 3770k chip at 1.4V & 4.9 or 5.0GHz it still run warmer than you would like. The tri-Gate transistors on these chips just run hot once you hit that 4.6/1.3V mark. Unless you take a chance by removing the IHS of the chip and run on the bare die it will still run hot.

To my knowledge theres really no way of knowing how much it affects chip longetevity or degradation. You will probably upgrade before it bites the dust. Is an extra 300MHz and 200mV to get there worth the risk of damaging the chip?

Thats just my opinion. I don't need to run a 5.0GHz clock 24/7. If I can get there thats great, then drop back to a more sensible speed for regular use.

[Generic User #2]

fair enough. but why are you calling them tri-state transistors? they're still only have on/off states.

[dandelioneater]

Originally Posted by Generic User #2

fair enough. but why are you calling them tri-state transistors? they're still only have on/off states.

Oops. I'm mistaken. It should be ''Tri-Gate'' transistors. will edit previous posts.

[Silvgearx]

You clearly haven't don't enough research. Instead of listening to "theories" with no explanation behind them, you should go to the intel site and read the data/spec sheets, you will find more then your looking for. There is really is no "safe" voltage.

Degradation? No body knows, if you that worry about it just keep it under what your mother board suggest. Any decent P77 will have a warning/red voltage when you go over 1.4v. Or keep it stock and run it for the next 20 years.

Voltage numbers, temperature, overclocked speeds, they should be extremely easy to find. Because we are so limited to on choosing a Intel CPU nowadays.

[Generic User #2]

How long will a motherboard actually last for? I mean, the components on it must have some sort of a lifespan - I have trouble believing they're 'immortal' even at stock settings.

I still think worrying about degradation is a moot concern. Personally, I've never heard of a modern overclocked CPU dying except during suicide runs.