Sign in with
Sign up | Sign in
Your question

EIST & C1E - Idle multiplier

Tags:
Last response: in CPUs
Share
January 13, 2013 2:03:16 PM

Hi,

I've searched high and low on the internet to find an answer to my question, however I've had no succes. Here's a short discription of my issue:

I'm not really sure if it can be considered an issue, but I'd like to be certain.

I've been running a core i7 3820 for quite a few months, installed a Corsair H80i and overclocked it to 4.5GHz (1.3v Vcore, LLC=Auto). I've enabled EIST and C1E to save a few watts when my PC is idle. Whenever my PC is inactive however, Realtemp v3.70 always reports a changing multiplier. 125mhz*12 is the bare minimum, but it never settles down (jumps up to 125mhz*13-16 when idle), despite the CPU being between 0% and 0.5% load.
I used to own a core i7 870 @ 3.2Ghz and when my system was idle, the multiplier settled at the minimum value until I put load on the CPU. Everything other than that works fine, but I can't help but wonder if this could damage my CPU in the long run (the constant changing multiplier when idle).

Any input would be much appreciated.
a b à CPUs
January 13, 2013 4:01:34 PM

Have a look in the Control Panel - Power Options at the Minimum processor state. Set this to a low number like 5% if you want your CPU to use a low multi when idle.

If you want your CPU to be very efficient at idle then leave this setting at 100% and enable the C3 and C6 sleep states. If you do this, your CPU will spend 99% of its idle time asleep in the C6 state where power consumption is minimal. Many people still think you need a low multiplier when idle but when using C6, this is not true. It makes virtually no difference what CPU multiplier your CPU uses because each individual core should be spending over 99% of its time asleep.

The new CStates button in RealTemp will show you exactly what your CPU is doing at idle. There are a lot of old myths about CPU power consumption that are simply not true on modern Intel Core i processors.

RealTemp T|I Edition
http://www.overclock.net/t/1330144/realtemp-t-i-edition

Here is what some smart guys at Berkeley discovered.

Power Optimization – a Reality Check
http://www.cs.berkeley.edu/~krioukov/realityCheck.pdf

Lockiing your CPU to the minimum multiplier when idle will make it feel sluggish but does nothing to save power. You are much better off letting your CPU get up to full speed as quickly as possible so it can process what it needs to do and then get back into the C6 power saving sleep state as soon as possible for as long as possible.

Intel CPUs are designed so the multiplier is constantly and rapidly changing, hundreds of times a second. Your CPU will live a long life if it is allowed to run as it was designed.
January 14, 2013 10:05:32 AM

Thanks a lot for the info! I'll check the C3 & C6 states tonight. Now here's hoping this doesn't interfere with my overclock.
a b à CPUs
January 15, 2013 4:20:25 AM

You can overclock and use C3 & C6. Most people don't do this because as soon as they run into problems, they read the forums, these sleep states get blamed for their instability and they immediately get turned off. Then they waste their time trying to come up with some magic offset voltage settings so they can reduce power and overclock at the same time which is usually the root cause of their instability problems.

Once you start to understand how a Core i CPU works, you will discover that there is no need to do this. I disabled C1E on my Asus board because that was preventing the CPU from resuming from sleep correctly when overclocking. With C3/C6 enabled, a CPU core will spend virtually zero percent of its idle time in C1E so turning it on or off does not change power consumption. On or off makes no difference because at idle, the individual cores are going to be in C6.

Same goes with using a fixed voltage. Users see a high fixed voltage or a high multiplier and they assume that the CPU is using lots of power. The C6 sleep state automatically drops the core voltage close to zero. For reducing power consumption, you can't beat that.

I am hoping more people will do some proper testing with a Kill-a-Watt meter so they can learn how to maximize performance and minimize power consumption, both at the same time.

Maybe Tom's needs to do some myth busting with a modern Sandy or Ivy Bridge CPU.
!