Seven and SpeedStep

Hi. I beleive I have a problem with one of both my notebooks that doesn't idle at the lowest CPU multiplier. It's a Core i3 330M running Seven x64, Balanced power profile, Minimum Processor State of 5 percent - 0 to 5 tested, SpeedStep enabled in the mainboard BIOS. It idles at the average multiplier of 10.5 - idle indeed: it's been monitored for long periods with various applications ThrottleStop, CPU-Z, RealTemp, CoreTemp, Open Hardware Monitor etc, under maximum load of less than 1 to 3 percent.

I am worried because my other notebook idles at the lowest multiplier steadily. It's a Dothan Pentium M and runs XP.

Am I right about the Core i3 not working properly by not steadily idling at the lowest multiplier of 7?
Does anyone have any input for direct comparison with my Core i3? For instance, another Core i3 running Windows 7 x64 and steadily idling at multiplier 7.

Thanks a lot for any input!

 

Mine idle's at 7 -

I am however running 32bit but i don't think there should be much difference.

 

There might be stuff going on in the background preventing it from going full idle. I recommend looking in your task manager to see which processes are using the CPU and how many processes are running. (Well, actually I recommend ignoring it and getting on with your life, but that's not what you're looking for.)

 

Something's wrong there too! 1.2GHz is not multiplier of 7.
You may be interested in my reports here.

This is mine:

​

 

The problem seems to be with BCLK calculation. At idle it seems that BCLK is calculated at ~179MHz instead of the normal 133MHz. This might make 7x multi appear as 179/133*7= 9.4 in TS for instance.

Maybe it's a chip bug and the performance counter(s) misbehave at idle.

While idle with TS what does using the "Update BCLK" do? Do the multi's momentarily go to 7x?

 

Only change is the FSB being displayed at top in the monitor window: 133MHz.

I've made several tests already. Please have a look of this thread where I fist tried to obtain help.

 

So you hit the "BCLK" button on TS to recalculate BCLK the frequency reported was still 133 and the reported multi didn't drop for a very short time when you did that?

TMonitor is load based and reports are only accurate for multi when there is 100% load for the sampling time.

I don't think your multi is running higher than 7 when set but the problem is to try and prove this.

 

Dufus, I so wish you were right. TMonitor is broken, I've noticed that as mentioned in the quoted thread. However ThrottleStop seems very accurate and succeeds reporting the multiplier 7 in the very specific occasion it occurs.

I also wish more of you would check this out more carefully. It's odd - I think you are not quite getting it.

CPU idle state is associated with the higher multiplier. Whenever any process takes place, either intensive or not, multiplier changes up and down. It's so odd that CPU actually downclocks when the task doesn't demand as much process as the actual locked multiplier can do.

Here and in the thread I quoted, two other accounts have been confirmed.
HWiNFO32 reports correct actual frequency, compared to ThrottleStop, CoreTemp, RealTemp and Open Hardware Monitor, however it fails with the associated multiplier when idle - that just confirms the oddity.
If SHoTTa35 runs ThrottleStop, reported multiplier will be around 10.5.