Geek idea

[ciphergoth]

may be unoriginal or wrong.

Processors can be damaged by getting too hot, so they have processor fans. Modern fan assemblies include a thermometer so they can adjust the fan speed to keep the processor at a constant temperature.

However, another thing that can damage processors is sudden changes in temperature. Suppose a processor is working flat out and the fan is going full whack. If the processor suddenly stops doing work and so generates much less heat, the fan won't slow down until after the sudden drop in temperature.

So really, motherboards should monitor the power consumption of processors as well as their temperature, so that sudden changes in workload can be responded to quickly to prevent sudden changes in temperature and so prolong the life of the processor.

Comments

[lovelybug.livejournal.com]

Sounds sensible enough to me :) Didn't hear you go this morning, so didn't get to say again - last night was hot ;) See you later beautiful xx

ok, you can go back to your geeking now...

[aegidian]

Umm.. isn't heat output pretty much in direct proportion to power consumption (processors viewed as heating devices), so a temperature sensor is in effect a power consumption sensor?

[wechsler.livejournal.com]

Since both CPU and motherboard have a non-zero thermal capacity, there will be a delay between a change in power input and the point at which the device reaches a new thermal equilibrium; as such a power monitor is useful in predicting the change in temperature that *will* occur, while a thermometer can only measure current state.

[conflux.livejournal.com]

The thermal capacity of the combined heat sink and chip means that cooling due to reduced workload is not a major problem if the heat sink is big enough to absorb rapid heat fluctuations. So basically all you need is the relatively low tech and cheap solution of making the heat sink out of a large enough chunk of metal. I use Zalman heat sinks, which are made from a very large and heavy piece of copper arranged in flower leaves, for this reason. In my opinion smaller cheaper heat sinks are a false economy if you want your computer to last for a few years.

[aidan-skinner.livejournal.com]

You should be able to hook cpufreqd or equivalent to do this under Linux, assuming you cared enough. ;)

- Aidan (oh, and ACPI worked for you, still doesn't on my laptop even under 2.6)

[vampwillow]

if there was *only* a fan then I'd agree with you, but the heatsink evens out / slows down the rate of change, so 'sudden' changes will never happen.

[martling.livejournal.com]

Agreed - I don't think you'd end up varying dT/dt by more than a factor of 2-3 with typical setups, and I think there's already a pretty large margin from rates that would be a problem, even with regard to long term chip lifetime.

I think it's still worth looking at the idea though, but with a view to optimising power usage over both the fan and CPU, for laptops.

[princealbert.livejournal.com]

on AMD AthlonXP compatible boards theres a thermometer directly under the CPU if it reaches a preset temp (set in BIOS) the machine shutsdown.

Some fan/cooler solutions come with their own thermometer which is placed in the heat conductive gel between the cpu and the cooler base to achieve the fan control you mention.

The three wires going from the motherboard to the fan supply on a fixed voltage and return a strobe signal to obtain the rpm.

A CPU going full whack, then returning to idle does not suffer thermal shock. Due to the conductive gel going to the base of the cooler, (under the fan), the CPU gradually loses heat as the fan does its job, it takes minutes compared to the speed a CPU can obtain idle.

So as long as you have a big bit of copper, (other materials are used on cheaper solutions), between your CPU and your fan your CPU wont suffer thermal shock.

Alternative cooling systems like water still have the copper base.

-Roy

[envoy.livejournal.com]

You could still do it with just a thermometer. Just use a more advanced logic. Make the speed of the fan be in some part or whole based on the delta of the temperature over the last 30 seconds-minute. That way if the temperature starts to rise OR fall very quickly the fan can change to compensate, tapering off as the delta of the temperature does.

I *think* delta is the right word there, but I might be wrong.