Basic question regarding PWM-controlled fan headers

[tiresias]

Dear Sirs,

Some background info:

I have recently gained possession of a number of fantastic Sanyo Denki fans (by courtesy of our forum's "Galvanized"), sized 120x38mm and rated at 2600rpm at 12V. Two of these fans have now been running without incident powered by the PWM-controlled fan headers on a dual opteron Tyan K8SR serverboard, and one attached to my Tyan K8E.

Which brings me to my question:

Is it possible, by inspection of the circuitry or otherwise, to estimate the amount of power one can safely draw from such a motherboard fan header for long-term usage?

The motivation behind my question is that I was planning to replace the rather cheap and cheerful "Made in China" 120mm fan in my wife's machine with the remaining one Sanyo Denki. She is, however, using a MSI Neo2, which unlike the Tyans I mentioned, is certainly not a board designed to deal with 5 or 6 15000rpm 40mm screamers in a 1U rackmount unit.

For the record, these Sanyo Denki are a variant of this model, with identical current rating (0.52A): http://sanyodb.colle.co.jp/product_d...master_id=1140 .

[Galvanized]

Generally, desktop boards are speced at 8W per fan jack but I would not load all four ports at 8W. I had to really dig a bit in the ASUS manual to find the wattage rating for the fan jacks. Generally, server boards have about twice that rating, maybe more.

After 5 minuets, I gave up on down loading a MSI manual. Thier site is slow.

[PeteS in CA]

Be careful of DC fan current ratings. That .52A fan actually draws current in pulses that may be 1.5A-3A, but the average current is .52A.

[gonzo0815]

Originally Posted by tiresias

Is it possible, by inspection of the circuitry or otherwise, to estimate the amount of power one can safely draw from such a motherboard fan header for long-term usage?

.

To pic up the question, i think yes it is probably possible to do a educated guess on sutch a circuit. If the layout of the board is given, probably with some math this could be a very accurate guess.
I think the most interesting questiuon is the absolut power rating of the mosfets, and the cooling area of them.
The next thing is the absolut area of the coper traces and the thickness of the cooper laminate, to calculate the power.

[Rainbow]

The power to the fan connectors is often switched by tiny SMD transistors which can't withstand too much.

[gonzo0815]

Yes, but in recent (especially AMD C&Q) boards there are some not that small mosfets (those 2 to3 mm long rectangular, soldered like the bigger mosfets in vrm with the hs to the pcb) which may be can handle up to severall amp`s. The limiting factor would probably the heat transfer from thos devices. But with PWM there arent that mutch losses and therefore not that mutch heat.

[tiresias]

Originally Posted by gonzo0815

Yes, but in recent (especially AMD C&Q) boards there are some not that small mosfets (those 2 to3 mm long rectangular, soldered like the bigger mosfets in vrm with the hs to the pcb) which may be can handle up to severall amp`s. The limiting factor would probably the heat transfer from thos devices. But with PWM there arent that mutch losses and therefore not that mutch heat.

Interesting. Would you like me to take pictures of the headers and surrounding area in question on the two motherboards I was talking about?

[gonzo0815]

Well, no problem. if the device is identifiable i could easily look into the datasheet and give a roughly guess what this could be handle. For the more accurate guess i am not shure if i am educated enough to do it on a safe way.