Sandy Bridge Overclocking Speed Q?

[Jman13]

Stupid question I'm sure, as I've been out of the loop on new hardware for quite some time. Just poked around a bit before my new build to see what the best bang for buck was and then went to my local MicroCenter to pick up the goods.

Anyway, I'm not planning on overclocking my 2500K...yet...though I likely will eventually (I'll save it for when I feel like I need a speed boost). Anyway, I'm curious about what speed people are actually referring to when they say "I run my i5-2500K at 4.5 GHz". Is this the nominal speed (such as 3.3 GHz on the 2500), the single core top turbo speed or the turbo speed for all 4 cores?

Also, when overclocking one of these chips, does the idle clock speed (which is ~1.6GHz according to CPU-Z) also go up accordingly, or does it still stay at that lower idle speed?

Thanks.

 

[Zap]

Usually people are referring to the Turbo speed for all 4 cores. When overclocking using Turbo and an offset voltage (as opposed to fixed voltage) the CPU can still idle at 1.6GHz at whatever miniscule voltage needed. Thus, best of both worlds. Super low power draw when idling. Super high performance when loaded up.

 

[Jman13]

Thanks!

 

[gramboh]

Usually people are referring to the Turbo speed for all 4 cores. When overclocking using Turbo and an offset voltage (as opposed to fixed voltage) the CPU can still idle at 1.6GHz at whatever miniscule voltage needed. Thus, best of both worlds. Super low power draw when idling. Super high performance when loaded up.

I also have no experience with SB, but will be building an IB system when it comes out.

So with SB, you can bump the max turbo speed on all 4 cores (say to 4.5GHz), then when any load hits the CPU, all 4 cores will always go to 4.5GHz (assuming the load is significant) and then scale back seamlessly to 1.6GHz?

Is there any reason for one to disable power gating/saving features and just run at 4.5GHz 24/7? This is how I have been operating my Q6700 since I bought it, because the power saving features make overclocking unstable on this platform, which is now about 5 years old.

 

[Puppies04]

I also have no experience with SB, but will be building an IB system when it comes out.

So with SB, you can bump the max turbo speed on all 4 cores (say to 4.5GHz), then when any load hits the CPU, all 4 cores will always go to 4.5GHz (assuming the load is significant) and then scale back seamlessly to 1.6GHz?

Is there any reason for one to disable power gating/saving features and just run at 4.5GHz 24/7? This is how I have been operating my Q6700 since I bought it, because the power saving features make overclocking unstable on this platform, which is now about 5 years old.

Nope, no reason whatsoever (unless you are going for crazy OC). Some people disable the power saving features and turbo to check their voltages are stable when first setting up their overclock but I don't bother.

 

[gramboh]

Nope, no reason whatsoever (unless you are going for crazy OC). Some people disable the power saving features and turbo to check their voltages are stable when first setting up their overclock but I don't bother.

Glad to hear this, as it will be nice to have power saving features and the OC power when you need it.

So voltage is handled with an offset then? So when the turbo kicks in, voltage is offset by +x volts?

 

[kalrith]

Glad to hear this, as it will be nice to have power saving features and the OC power when you need it.

So voltage is handled with an offset then? So when the turbo kicks in, voltage is offset by +x volts?

It is handled by offset, but it's also impacted by LLC (line load calibration). Basically LLC of 1 will keep voltage high when at load, and LLC of 5 will keep it lower when at load. I don't think LLC affects idle voltage. Usually 2 is sufficient for most overclocking (4.3 - 4.4 GHz) and 1 is used for higher overclocking (4.5 GHz).

I read a recommendation that was really helpful. I ran the voltage in fixed mode and LLC 1 while determining my overclock. I watched the voltage in CPUz and wrote down the lowest voltage I saw at load (which was 1.24v). Once it was stable with Prime95 for 24 hours, then it was time to play around with the offset.

I wound up with a -.40v offset and LLC 2 with a 43x multiplier. That puts the load voltage at 1.24v and the idle voltage quite low. I had to up the voltage too much to be stable at 44x, so I just stuck with 43x.

 

[gramboh]

It is handled by offset, but it's also impacted by LLC (line load calibration). Basically LLC of 1 will keep voltage high when at load, and LLC of 5 will keep it lower when at load. I don't think LLC affects idle voltage. Usually 2 is sufficient for most overclocking (4.3 - 4.4 GHz) and 1 is used for higher overclocking (4.5 GHz).

I read a recommendation that was really helpful. I ran the voltage in fixed mode and LLC 1 while determining my overclock. I watched the voltage in CPUz and wrote down the lowest voltage I saw at load (which was 1.24v). Once it was stable with Prime95 for 24 hours, then it was time to play around with the offset.

I wound up with a -.40v offset and LLC 2 with a 43x multiplier. That puts the load voltage at 1.24v and the idle voltage quite low. I had to up the voltage too much to be stable at 44x, so I just stuck with 43x.

Wow, I didn't realize you had that much control over voltage at idle/normal/turboboost states with SB, that is a huge leap forward from my Q6700/P965 setup. Thanks for the info, good to know this stuff for when IB comes out, since I believe IB/Z77 will be identical in terms of overclocking setup.