Why do CPUs run hotter...?

Certainly it's false. Modern CPUs and OSs (even WinBlows ;-) manage power rather effectively. Not only is the processor HALTed when nothing is happening, but the oscillator and power rails are modulated to optimize power dissipation and even system temperature. This has been going on for more than a decade, but has gotten serious over the past three or four years.

You refer to "wait states" which are entirely different from a WAIT instruction. The PDP-11 and the 6800 had a WAIT instruction ("wait for interrupt") which suspended opcode fetches and reduced power until an interrupt or other exception occurred. The 68K called it LPSTOP (low power stop) which did the same thing. I don't know how the Intel abortions implement this functionality, but they do. I think the Pentiums step down CPU clock speed when nothing's going on, as opposed to having an explicit WAIT or equivalent; Intel always prefers a kluge.

Not so. PCs use a lot less power when idling than they do when there's a lot of computing going on. My Toshba laptop only runs its CPU fan when it's working hard or when it's running a DOS app, which executes full-time. Google "pentium idle" and such for details.

John

But does the Windows idle loop (System Idle Process) actually do that? That's what I'm asking. I already know about embedded system design practices; I'm asking about PCs.

OK, that explains a great deal. Many thanks. Is it WAIT or HALT? It's been a while since I did '86 assembly language, but as I recall, WAIT waits for an i/o port condition and HALT waits for an interrupt.

Thanks again for all the information, John.

I think it's time for me to read a current Pentium architecture manual... back in 8088 days I knew exactly how a PC worked :)

Is Intel's manual the best source of information?

Although I first heard of it as an undocumented instruction on the 6805 !

Graham

Googling and wickipedia-ing a bit, it looks like the Pentium doesn't have an opcode that suspends execution, but it looks like something in the chipset or maybe the clock logic can step down the CPU clock rate. You might google "pentium idle" and "pentium speed-step" and stuff like that.

Personally, I'm not interested in the gory details of the Pentium architecture; I find it too depressing.

John

The actual electrons move faster?!?!?! Aye!

a Pentium CPU will run hotter but an amd cpu will keep the same temp.. Also it will run hotter if thears dust on and in the heat sink this can make it hotter in temp current temp + up to 40 to 50 more... That is the big deal with Pentium cpu thats why i use an amd + amds are faster you can find this out on google

Only on decent preemptive OS's, a decade ago most users were still using W95 ;-)

Note that HALT is only part of a strategy. HALT can be used in the idle task of a CPU, but there is also the possibility of reducing clocks when the load is light, which is not the same thing is it :-/

The key thing that has not been said so far in this thread, is _parallelism_. Modern chips are not internally single threaded. Even units with only a single processor core, have large sections designed to perform work in parallel. The first 'obvious' one, is the maths co-processor module, but in fact the degree of parallelism, is much larger. Most chips will be executing 'parts' of multiple instructions, ahead of the current location, all at the same time, and then deciding which one to use, or combining multiple results, based upon latter instructions. When a chip sits in a simple loop, or even better 'halts', these other parallel components, can stop running completely. Historically 'halt', was introduced in the early 1980's, but was first used in NT. Latter, Intel throttled the speed of various parts, and stopped the MCP in particular, with their 'SpeedStep' system. AMD, use a similar system with their "Cool'n'Quiet" system, which slows the clock rate, and drops the votage on parts of the chip that are not currently being used. A slightly modified version called 'PowerNow', operates on the 'mobile' processors. These throttling operations, help to reduce power when work is actually being done, as against the 'halt', which only operates during the system idle state.

Best Wishes

Yes.

How many times do I have to say it?

Not true. Win95 will HALT also. ...and it's not a decent OS.

One can also modulate the power supply, but these topics weren't at issue.

CPU

Theory does not predict that.

Sorry, I meant Win3.1, not that it made much difference because the W95 systems of a decade ago mostly ran software built to the old API, the OS could halt but did not know **when** :-;

Disagree, I am not talking about tweaking of the sort that overclockers do, I am talking about the possibilities the OS has to reconfigure on the fly to alter the performance vs consumption ratio.

BTW, another point is that the FP pipelines consume less when they are empty ;-)

Microsoft's theme song for the Pentium by the 'Stones tells it all:

"Start me up...."

"You got me runnin' hot!"

There are other hidden electronic "clues" in music, as in the Beatles':

"Help me get my feedback on the ground..."

:)

Paul

But they cut if off right before "make a grown man cry!"

It's friction. All those little rods and cogs and tiny bits rubbing against each other as the processor gets busy. When idling it runs slow and consumes little energy but once you give it something to chew on it has to engage all those bits at high speed and no lubrication in the world can keep it from getting a bit warmer. Some older processors additionally need to stoke up the boiler which makes it even hotter.

- YD.

idle.

Hey, have you ever sold comercial software? I have. My programs ran in loops waiting for some action by the user - hence, no power savings.

The power reducing tricks used today are part of the operating system which has hardware access to various power-saving functions.

Luhan