CPU clock speed varies

[RuyLopez]

Running a rather old Dell Latitude D800 with a 1.80 gigahertz Intel Pentium M processor.  I just noticed today that the clock speed is varying by as much as a factor of 3, from 1.8GHz to 600MHz.

The system does not show obvious indications of overheating.

OS:  Windows XP Pro, SP3.

I ran PC Wizard 2010 v193 and that certainly shows the variations in the clock speed and variations in the voltage.  Image of those measurements is attached.  It looks as if the clock speed is falling off dramatically when the voltage drops below approximately 1V.  If I recall correctly, the nominal voltage is supposed to be 1.3V. 

Any suggestions?

[Hoov]

What does your house voltage look like? If it is noisy, it will affect the output of the power supply.

[RuyLopez]

Hi Hoov,

I do not think it is associated with the mains.  The computer is a laptop, so after reading your post I took it off the mains and reran the PC Wizard test running on a fully charged battery.

The clock was still oscillating between 1.8GHz and 600MHz, although the voltage did appear more stable than before, running quite steadily at 0.988V.

Best regards,
RL

[Hoov]

Running off a fully charged battery definitely narrows down the possibilities. Looking at the graphs again, there is definitely a correlation between the three graphs. Try turning off everything except the minimums needed to run the system and then check it again. Then check the voltage graph. If it flattens out, you may need to restrict the maximum processor loads. I don't know if PC Wizard will run in safe mode, but it would be interesting to see the graph from that.

[RuyLopez]

Just found the voltage spec's as provided by Intel.  The CPU is a Pentium M 745 and the voltages are as follows:

Maximum Performance Mode:
Frequency=1.8GHz
Voltage=1.276 -1.340V

Battery Optimized Mode:
Frequency=600MHz
Voltage=0.988V

Given that, the battery test I ran seems to be exactly what should be expected.  Also appears that on AC power, the CPU voltage is pegged at the low battery optimized value for whatever reason.

I was running PC Wizard off the internal CD drive.  I will try running it in Safe Mode and report back.

[RuyLopez]

Given that, the battery test I ran seems to be exactly what should be expected.

That statement makes no sense!    Yes the voltage is what is expected, but the clock frequency is still oscillating between 1.8GHz and 600MHz, which is certainly not expected.

[Hoov]

 

[RuyLopez]

Try turning off everything except the minimums needed to run the system and then check it again. Then check the voltage graph. If it flattens out, you may need to restrict the maximum processor loads.

I just tried the exact opposite experiment, based on a preliminary, subjective observation.  It looks as if the clock speed attains the 1.8GHz rate at high CPU loads.  An image is attached that compares the CPU load, clock rate, and voltage.  That graph is consistent with the preliminary observation that the clock rate is low at low CPU loads. 

I am confused now.  I do not recall ever noticing a clock rate of less than 1.8GHz before, so I do feel confident that something has changed, just far from certain what, let alone why.

[Hoov]

Well I just checked mine, and it looks much like yours did. I started out with a full load, and then backed it off, and its doing the same thing. So I would say its normal.

Seems strange that I never noticed it.

[PCBruiser]

Hi, guys,  That's called SpeedStep.  It is controlled by your bios settings and built into all Intel CPUs since the ill fated Prescott P4, although I don't think Dell gives you the option to turn it off.  All of the mobile, and now all the multicore CPUs, have that capability.  It reduces the CPU speed to meet your current computing requirements.  That saves wear and tear on the CPU.  There are actually a few settings between the min and max although I rarely actually see them used.  It also reduces the max CPU speed for mobile CPUs when you are running on battery and have your power settings set to conserve battery power rather than maximize performance.

[RuyLopez]

Greetings Gentlemen,

Alright, it seems the norm and by design at that.  What confused me most was the apparent oscillation--I never cared for oscillations in the electronics that I have worked with.

I will assume that all is normal.

Thanks for the assistance--much appreciated as always.

Best regards,
RL

[PCBruiser]

I agree with you on variations with electronics, but in this case it is designed right into the CPU logic.  What happened was that Intel gave up on their evolution of their then current desktop CPU designs with the end of the line Prescott, affectionately known as the Preshott, and started to redesign their mobile line of CPUs to be their followup desktop ones.  They became the very popular and fast Core series of CPUs, plus the Atom for netbooks.  A very wise choice.  Given that SpeedStep was already built into the circuit logic, it seemed unnecessary to design out a valuable feature and left it in place.  It is one of the reasons their current Core generations of CPUs run efficiently at lower overall temperatures.  To the extent that variability can be an issue, it is most likely entirely offset by the lower overall operating temps of their CPUs.

[Hoov]

Thanks for the education.   

To be honest with you, I never paid much attention to it until RL started asking about it. It do make sense. Why waste power when you don't need it and are not using it.