OC voltage question

[B!nd.]

To my understanding you want a stable system with the lowest possible core voltage correct?

I ask because I will be OCing my 2500K to around 4.5 with a starting voltage of 1.35. Should I go up from there and see how much higher I can comfortably go, or should I test stability after dropping the voltage? I'm not looking to push my cpu to it's limits. I want a basic OC in combination with longevity.

Will be using UEFI btw.

 

[pelo]

There's some degradation as the processor ages and the more heat and voltage that pass through it the faster it "ages."

What happens is that the more voltage you pass through it (and that's voltage over the recommended limits) the more voltage you'll need to add after a shorter amount of time to sustain the same clock speeds. Basically, your CPU will get more vcore thirsty the higher you go in vcore and more quickly at that. What you should be gunning for is the lowest possible voltage with the highest possible clock speeds and at 4.5ghz with only a mild bump in vcore your processor will likely outlast the moment when it asks you for more juice because for an SB system that's a very mild and reasonable overclock.

http://www.anandtech.com/show/2468/6

 

[Pitbully]

Correct.
You can probably start with a lower voltage then 1.35 for only a 4.5Ghz O/C. Most 2500k's will do 4.5 pretty easily. My 2600k only needs 1.360v for 4.8Ghz, but I'd have to say its a bit above the curve if you look at some of the spreadsheets at overclockers.com.

When I get a new chip, I usually try and set some goals for it for the overclock I want, with the voltage I want. When I got my 2600k, I wanted to get at least 4.6 out of it. When I first got it, It booted right up at 5.0Ghz, so I knew 4.6 was going to be pretty easy. At 5.0Ghz, I had to push 1.48v to get it stable and the heat became an issue as well. I eventually figured out that 4.8 would be a good stable overclock for 24/7 use and then set about lowering the voltage and stress testing. Most people overclocking their Sandys tend to try and stay at 1.4v or lower. To give you an idea of how much difference tiny voltage adjustments make, my 2600k was stable at 4.8Ghz and 1.4v, but the temp was still higher then I'd like to see, (78-80C) under full load. Going from 1.48v to 1.40v the temp went from 88C to 80C. Going from 1.40v down to my final overclock of 1.360v @4.8Ghz the temps went from 80C down to 73C, under full load. At 1.32v, I would get a BSOD about 10 or 15 mins into running Prime 95 or Intel Burn Test, but at 1.360v it was perfectly stable.

 

[B!nd.]

There's some degradation as the processor ages and the more heat and voltage that pass through it the faster it "ages."

What happens is that the more voltage you pass through it (and that's voltage over the recommended limits) the more voltage you'll need to add after a shorter amount of time to sustain the same clock speeds. Basically, your CPU will get more vcore thirsty the higher you go in vcore and more quickly at that. What you should be gunning for is the lowest possible voltage with the highest possible clock speeds and at 4.5ghz with only a mild bump in vcore your processor will likely outlast the moment when it asks you for more juice because for an SB system that's a very mild and reasonable overclock.

http://www.anandtech.com/show/2468/6

Thanks, that explains a lot. But what are you saying is that overtime the CPu will sort of build up a tolerance to the voltage and I will have to give it more? Will symptoms for that be instability and BSOD?

Correct.
You can probably start with a lower voltage then 1.35 for only a 4.5Ghz O/C. Most 2500k's will do 4.5 pretty easily. My 2600k only needs 1.360v for 4.8Ghz, but I'd have to say its a bit above the curve if you look at some of the spreadsheets at overclockers.com.

When I get a new chip, I usually try and set some goals for it for the overclock I want, with the voltage I want. When I got my 2600k, I wanted to get at least 4.6 out of it. When I first got it, It booted right up at 5.0Ghz, so I knew 4.6 was going to be pretty easy. At 5.0Ghz, I had to push 1.48v to get it stable and the heat became an issue as well. I eventually figured out that 4.8 would be a good stable overclock for 24/7 use and then set about lowering the voltage and stress testing. Most people overclocking their Sandys tend to try and stay at 1.4v or lower. To give you an idea of how much difference tiny voltage adjustments make, my 2600k was stable at 4.8Ghz and 1.4v, but the temp was still higher then I'd like to see, (78-80C) under full load. Going from 1.48v to 1.40v the temp went from 88C to 80C. Going from 1.40v down to my final overclock of 1.360v @4.8Ghz the temps went from 80C down to 73C, under full load. At 1.32v, I would get a BSOD about 10 or 15 mins into running Prime 95 or Intel Burn Test, but at 1.360v it was perfectly stable.

I don't plan on going any higher than 4.5GHz. So should I set the multiplier to 45, leave the voltage and test first? Or should I go ahead and make a small increments like .05 volts off stock setting until I find the lowest value I can have for the voltage? (obviously wont do .05 every test)

I also have yet to use Prime95. What kind of notification should I expect for instability?

 

[pelo]

Yes and it naturally happens regardless of whether you OC or not, but happens at an exponentially accelerated rate as your OC and consequently voltage, go higher. Generally speaking, unless you OC it past what are considered safe levels by the manufacturer (Intel in this case) you won't see this become a significant issue. But for more extreme overclocks and watercooling setups this can be an issue.

IDC from Anandtech posted some great stuff that outlines power consumption clock speed and voltage increases along with heat and cooling conditions.

http://forums.anandtech.com/showthread.php?t=2200205
http://forums.anandtech.com/showthread.php?t=2195927

If it's too technical for you, and admittedly some of it is for me as well, the basic gist of it is that you shouldn't hammer down as much voltage as you can to get an epic overclock just because your heat output is low as that's only 1 determining factor in processor degradation. What you should do is find that sweet spot which your CPU has where it only takes a small vcore increase to get a significant overclock out of it. Past a certain point CPUs will require an exponentially increasing amount of voltage.

For example, my 955 Phenom II can get over 4ghz but in order to do so it requires voltages of over 1.47v and anything over 3.8ghz requires an insane amount of voltage. I can get to 3.8ghz on 1.36v but to get to 4ghz I would need to bump it up nearly .1 vcore just in order to get 200mhz! That would be a sign of a ceiling for the chip and just what "safe" is.

Hope this helps

 

[Pitbully]

Thanks, that explains a lot. But what are you saying is that overtime the CPu will sort of build up a tolerance to the voltage and I will have to give it more? Will symptoms for that be instability and BSOD?



I don't plan on going any higher than 4.5GHz. So should I set the multiplier to 45, leave the voltage and test first? Or should I go ahead and make a small increments like .05 volts off stock setting until I find the lowest value I can have for the voltage? (obviously wont do .05 every test)

I also have yet to use Prime95. What kind of notification should I expect for instability?

Normally you look for either errors in Prime 95/IBT or BSOD. This usually indicates you need more VCore for your chip. Usually you can start with stock voltage and increase it from there, but you can also start at say 1.30v and DECREASE it from there as well. If it boots fine at 1.30v, run some stress testing, if it passes, reboot and set your core to 1.25 or something and retest. To get a good stable overclock, you'll need to do this several times obviously. If you get a BSOD, check the error message, if its 0x124, 0x1A, 0x101 or 0x1E, this normally indicates you need MORE Vcore.

 

[B!nd.]

Yes and it naturally happens regardless of whether you OC or not, but happens at an exponentially accelerated rate as your OC and consequently voltage, go higher. Generally speaking, unless you OC it past what are considered safe levels by the manufacturer (Intel in this case) you won't see this become a significant issue. But for more extreme overclocks and watercooling setups this can be an issue.

IDC from Anandtech posted some great stuff that outlines power consumption clock speed and voltage increases along with heat and cooling conditions.

http://forums.anandtech.com/showthread.php?t=2200205
http://forums.anandtech.com/showthread.php?t=2195927

If it's too technical for you, and admittedly some of it is for me as well, the basic gist of it is that you shouldn't hammer down as much voltage as you can to get an epic overclock just because your heat output is low as that's only 1 determining factor in processor degradation. What you should do is find that sweet spot which your CPU has where it only takes a small vcore increase to get a significant overclock out of it. Past a certain point CPUs will require an exponentially increasing amount of voltage.

For example, my 955 Phenom II can get over 4ghz but in order to do so it requires voltages of over 1.47v and anything over 3.8ghz requires an insane amount of voltage. I can get to 3.8ghz on 1.36v but to get to 4ghz I would need to bump it up nearly .1 vcore just in order to get 200mhz! That would be a sign of a ceiling for the chip and just what "safe" is.

Hope this helps

Thanks for the links Pelo, I am currently at work so I wont be able to view the pics posted in that thread so I'll read into it more when I get home. But I understand the gist of it.

I'm scared to start out with too much voltage lol. I want stabilitly at 4.5 with the least amount of vcore as possible

Normally you look for either errors in Prime 95/IBT or BSOD. This usually indicates you need more VCore for your chip. Usually you can start with stock voltage and increase it from there, but you can also start at say 1.30v and DECREASE it from there as well. If it boots fine at 1.30v, run some stress testing, if it passes, reboot and set your core to 1.25 or something and retest. To get a good stable overclock, you'll need to do this several times obviously. If you get a BSOD, check the error message, if its 0x124, 0x1A, 0x101 or 0x1E, this normally indicates you need MORE Vcore.

I have never heard of IBT, I will have to research more on that. Yeah I like that I'll start at 4500Mhz with 1.30vcore and adjust from there.

I know a proper Prime95 test should run for ~12 hours. But this seems a little unrealistic when you're constantly changing values. How long should I keep a test running to be confident in some stability before I sit out a 12 hour run?

I am not familiar with BSOD error codes. Are there any other common errors you could possibly refer me to so I am prepared?

Thanks guys.