Going to one HD 4545

Not at all. If you blow down instead of parallel to the heat sink fins, you need a larger heat sink, and more air flow from the fan, to get the same amount of cooling.

Of course I have seen the ones you mean. For small slower processors, they provide "enough" cooling and are simpler, mechanically, to install. For something like a 166 MHz Pentium a tiny little fan blowing on a trivial heat sink will do the job.

For a 90 watt 3.06 GHz Xeon, though, nothing like that will do the job. That is why Intel supply their own heat sink (a huge copper thing with lots of closely spaced fins that weighs about one pound), a special mounting buttembly where the bolts go right through the motherboard into some metal stand-offs that bolt to the motherboard mounting plate because the weight of the heat sink would compromise the integrity of the motherboard, a "processor wind tunnel" that completely surrounds the heat sink to ensure that the air from the fans go through the heat sink and not around it, and a 60mm x 38mm ball bearing fan whose speed is temperature controlled to get enough air through it. They recommend getting the cooling air from the front of the tower and blowing towards the back. Since Intel supply this stuff (and the heat sink grease and the heat sink mounting spring clips), we must conclude that this is a stock CPU heatsink-fan for these processors where heat rejection is of major importance.

Maybe the O.P.'s processor is cool enough as is, but reversing the fan(s) in the power supply could put it over the top in heat. And if heat is not a problem, why is he changing the cooling around in his machine anyway? He must perceive it to be a problem.

Well, yes. And it need not quit completely -- that would actually be better than what seems to happen often. Instead, its stability is reduced and later cooling will not reverse that. Then people test their memories, replace their hard drives, change their cooling, etc., to little avail because the processor has been compromised. For most things electronic, adding 10C to the temperature cuts its MTTF in half, so you should never heat things more than necessary. It just seems dumb to blow hot air on something that needs to be cooled. If a processor (actually a computer system) does not need to be reliable, why cool it at all? Or why even turn it on? Well my Xeons are redlined at 70C processor temperature, which implies a lot lower ambient air temperature. When the inside of my box gets up to about 52C, the processors go up to around 56C or so, and the cpu cooling fans are screaming at about 6000 rpm. I am blinded only by Intel's recommended cooling practices for the processors I am using. Since those thing cost something like $750 each (back when I got them; they are surely cheaper now), I see no reason not to follow Intel's cooling recommendations.

-- .~. Jean-David Beyer Registered Linux User 85642. V PGP-Key: 9A2FC99A Registered Machine 241939. ^^-^^ 09:45:00 up 71 days, 3:40, 3 users, load average: 4.44, 4.01, 3.89