Is thermal load the same at 100% usage as it is when unthrottled

[Ruptga]

I have a feeling my title isn't clear. What I mean is, if I have a processor running at its maximum frequency (no throttling), but it's sitting idle, and I later have the same processor running at 100% usage with F@H or something, is it pulling the same amount of juice as it was when at full speed and idle? Is there a difference between unthrottled and 100% usage in terms of heat produced?

 

[randay]

42

 

[sao123]

Originally posted by: ADDAvenger
I have a feeling my title isn't clear. What I mean is, if I have a processor running at its maximum frequency (no throttling), but it's sitting idle, and I later have the same processor running at 100% usage with F@H or something, is it pulling the same amount of juice as it was when at full speed and idle? Is there a difference between unthrottled and 100% usage in terms of heat produced?

Yes

 

[Peter]

Yes. Even without Cool&Quiet, Speedstep or somesuch technology enabled, idling processors enter a power-saving, mostly halted state - the so-called "Stop Grant" state.

They'll be stuck with their maximum operating voltage and clock though, which will make them consume more power than at minimum.
Either way, we're talking about just a few watts, far from the maximum consumption under full load.

 

[CTho9305]

Yes. Power comes from two components: "static" power and "dynamic" power. Static power is power consumed due to transistors not turning off completely, so there's always a little bit of current leaking. Static power doesn't really change unless you reduce the voltage. Dynamic power comes from two sources, both related to signals switching. The highest power consumption occurs when blocks are active.

Take a look at http://www.amd.com/us-en/assets/content_type/DigitalMedia/die_marked_E.jpg

Even at "100% usage", power consumption varies. For example, if you were running the code

loop: goto loop

..the only block with much switching would be the box labelled "fetch scan align microcode".

If your workload were more like

loop: a = b + c
d = e + f
g = h + i
goto loop

then the box labelled "execution units" would be burning quite a bit of power, but load/store and the floating point unit would have almost no switching going on. Even though the CPU is "100% busy", the power is pretty far below the maximum.

The highest power occurs when you have the most power-hungry units switching as much as possible; programs like http://pages.sbcglobal.net/redelm/ attempt to do just that. If you're interested in seeing how "100% load" can produce different temperatures, look at your CPU temperature while running burnk7 and compare them to temperatures while running something like seti@home or folding@home.

 

[Ruptga]

Originally posted by: randay
42

Yes, the answer to life, the universe, and everything, but I like CTho9305's answer better.