Making a CPU do more work to raise tempeature

This sounds like a rather interesting method to warm a room. You could use basically a standard PID algorithm, with the output being a duty cycle of the CPU. The CPU would thus crunch x seconds per minute, each minute to warm the room up.

You would deal with the thermal lag by working out how much heat the room needs per degree under the setpoint, the heat losses of the room, and the heat generation of the computer unloaded and loaded.

There are a number of tuning algorithms available for PID controls in the textbooks. In your case, you would measure the time it takes for the room to warm up to a new temperature from cold. From this time you would obtain the time constants of the room and then get the PID constants.

Michael

Cheers. I felt it had a chance of working.

Someone on another newsgroup: "Whether this would have any effect on the room temperature remains to be seen, but it sounds like a fun way to waste some time if you've got time to waste"

I can certainly feel the extra heat generated when the CPUs are all busy doing CPU intensive tasks. However, maybe the extra heat is only

Thanks, that makes some sence.

The problem I have is that I can't do any sort of realistic controls, since the heat generated by the Sun, removed by wind etc is outside my control. If the room warms up, how do I know if its the Sun or my computer (which happens to be made by Sun Microsystems!!) Obviously if the heat generated by the computer was large compared to that provided by the Sun, it would be an easier task to estimate the effect of the computer.

In the old days you might have more of a chance here. I doubt that a CPU uses 200 watts more while busy than while idle. My guess is that it is more in the line of 20 more watts.

Given the low heat input, you could use a simple thermostat and read its output. There are some programmable thermostats which will are also autotuning. As the heating rate is going to be so slow, simple bang bang control would give you almost the same results as model based controls.

If you go a wee bit further, you could put a temperature sensor on the surrounding external walls. This would allow you to get a better handle on the heat losses. These could be used to model the room a bit better. If you had a few hundred watts, or a few thousand watts available, then it might have a reasonable chance of working.

Michael

These are not particularly new CPUs - they are only 450 MHz and have 4MB cache. They are not Intel compatible.

And I do have 4 of them in the same computer, so a multi-threaded program could make all 4 work hard. I don't know if RAM uses extra power when read/written to a lot, but there is 4GB of that.

It's not a typical PC, but a Sun Ultra 80.

I'll try to measure the input power taken by the computer. I just need to hook something up to measure the phase angle between voltage and current, although I am not sure with switch mode supplies that the current waveform would be in the least sinusoidal.

The CPU runs off DC. Measure that, it's easier.

Jerry