How is this possible?

[tw33k]

I've set up a new test bench with an ASRock OC Formula and a 3770K. My first attempt running the chip @ 4.7GHz I had an offset of +0.010 (VID is 1.2760) LLC at Level 1 (which is the highest level on this board) The vcore jumped between 1.256v and 1.280v with an average of 1.270v as reported by Aida64. (CPU-Z reported 1.272v) I got the system stable at this voltage.

I went back and added 0.008v Turbo Voltage, changed the offset to +0.020 and set LLC to Level 3. The vcore was between 1.232v and 1.264v with an average of 1.245v (CPU-Z reported 1.256v) The system is stable at this voltage.

So basically the voltage has been decreased by 0.016v but remains stable. I'm not complaining, just don't understand why. Any thoughts? I have a DMM on the way so I can get a better idea of the actual voltage.

 

[Puppies04]

I think you might be missing some information in your post because the way I am reading it is this....

- You had more voltage than the chip required to be stable.
- You lowered the voltage and were amazed it was still stable.

 

[tw33k]

At first CPU-Z reported 1.72v. Any lower and Aida64 failed. Then I raised the offset, lowered the LLC and increased the offset. CPU-Z reported 1.256v.

I've read this board is far from accurate when reported vcore so I've ordered a DMM.

 

[Idontcare]

I've set up a new test bench with an ASRock OC Formula and a 3770K. My first attempt running the chip @ 4.7GHz I had an offset of +0.010 (VID is 1.2760) LLC at Level 1 (which is the highest level on this board) The vcore jumped between 1.256v and 1.280v with an average of 1.270v as reported by Aida64. (CPU-Z reported 1.272v) I got the system stable at this voltage.

I went back and added 0.008v Turbo Voltage, changed the offset to +0.020 and set LLC to Level 3. The vcore was between 1.232v and 1.264v with an average of 1.245v (CPU-Z reported 1.256v) The system is stable at this voltage.

So basically the voltage has been decreased by 0.016v but remains stable. I'm not complaining, just don't understand why. Any thoughts? I have a DMM on the way so I can get a better idea of the actual voltage.

If your instability is coming from voltage drops in the transients, which is what you are trying to mitigate with a combination of LLC settings and VID offsets, then it is highly likely you will never "see" the data that shows you why one setup is superior to the other.

You are getting voltage readings that are averaged out over tens of milliseconds if not longer, but your CPU is going to hickup and stop if a Vdroop transient that is too low happens to occur in the span of micro-seconds or less.

It is quite possible one of those setups you have there, the one with the seemingly lower average operating voltage, is superior because it actually keeps the absolute minimum voltage value from the transients above the instability area.

 

[tw33k]

If your instability is coming from voltage drops in the transients, which is what you are trying to mitigate with a combination of LLC settings and VID offsets, then it is highly likely you will never "see" the data that shows you why one setup is superior to the other.

You are getting voltage readings that are averaged out over tens of milliseconds if not longer, but your CPU is going to hickup and stop if a Vdroop transient that is too low happens to occur in the span of micro-seconds or less.

It is quite possible one of those setups you have there, the one with the seemingly lower average operating voltage, is superior because it actually keeps the absolute minimum voltage value from the transients above the instability area.

That makes sense. Thanks

 

[tw33k]

My new DMM arrived so I took a reading of the v-core @4.7GHz and did not see the fluctuations reported by CPU-Z, Aida64 etc. The voltage ranged from 1.236v - 1.238v. I don't feel as bad about it now.