[robespierre]

I believe this issue was discussed at some length on Nekochan, but I can't find it in the archives. Is the fan simply powered through a NTC resistor?

My impression is that the Indy suffered from conflicting requirements. One was to have an effectively silent computer on the desktop, for videoconferencing or audio production work (the Indigo2 was proving to be annoyingly loud for these things). Another requirement was to have sufficient cooling ability for internal HDDs, powerful processors, and option cards, since a machine with an SGI price tag needs to be upgradable. A third requirement is to control costs, both design NRE and production cost. Putting a NTC in series with a regular computer fan is a compromise, because it adds minimal part cost, requires very little additional engineering, and sort-of satisfies the first two requirements. But they are not properly satisfied, because such a primitive control mechanism cannot truly set the airflow to the needed value: it's more like a gesture than a real solution.

A real solution would maintain sufficient minimum airflow to prevent hot spots from developing as the machine warms up in a cold ambient environment, and track temperature rise by increasing heat ejection to maintain a safe operating level, without overcompensating by creating annoying fan speed surges. This may require a servo programmed to respond to tachometer feedback from the fan, zoned temperature monitors, and power supply voltage and current monitoring. Pretty soon you arrive at something looking more like a Fuel than an Indy.