De-lidded i7 3770k Test

On my 3820 the Asus software always reports the CPU at 45C under load where as other software has the various cores at 55-57C. Even using that version of AIDA shows CPU 44C, CPU Package 55C and then the various cores from 55-57C.
 
With proper voltage tuning 4.8GHZ seems to be a sweet spot for IB. After 4.8GHZ I have to add crazy voltage to push clocks.

4.8 with low voltage and decent temps is great with IB. Great for gaming.
 
Last edited:
Asus sez: CoreTemp is in serious need of updating at this point for full compatibility with Ivy Bridge as is RealTemp (both latest releases do not include E1 stepping compatibility).
Click to expand...

That's a quote from someone that doesn't know what they are talking about. All monitoring software reads core temperature data from the same CPU register and the correct TJMax value can be read directly from any Core I CPU. The location of this data within the CPU has not changed since the original Core I was released in November 2008.

Core Temperature = TJMax - Sensor Data

The result is that all monitoring software that reports core temperatures should be reporting pretty much the same thing. Turbo-V is likely getting its data from a totally different sensor so these programs are not comparable.

When a CPU is lightly loaded, there are going to be times where different programs don't exactly agree with each other but that is usually because they are reading the sensors at slightly different times. Core temperatures can change instantaneously but you shouldn't be seeing any huge differences over time.

When fully loaded running LinX or IBT, I think RealTemp has an advantage over the competition because it runs at a higher priority level. This means it has a better chance of getting temperature data from the sensors when it requests it and can process and display that data immediately. In the example above, RealTemp 3.70 reports slightly higher peak core temperatures across the board compared to AIDA 64. Monitoring software that runs at the same priority level as LinX is going to have to get in line when trying to read the temperature sensors and might have its requests for temperature data ignored or delayed. Without timely access to the sensors, you might see significantly lower peak core temperatures reported during brief bursts of full load.

Keep in mind that Intel designs and calibrates their core temperature sensors to control thermal throttling and thermal shutdown and that's it. There has never been a need for Intel to invest extra money so enthusiasts could have 100% accurate core temperatures from idle to full load. The amount of error in the latest 22nm sensors is anyone's guess but accuracy of about +/- 5C is typical. The sensor data is good enough for general guidance when swapping parts but it becomes meaningless if you think you can use it to compare your CPU to your friend's CPU.
 
Last edited:
You must log in or register to reply here.
Back
Top