airflow and fan assumptions

[NeverDie]

1. It's gradually dawning on me that SuperMicro is designing their motherboards to go into SuperMicro cases, which have particular airflow pathways and air velocities. i.e. when their designers are picking passive heatsinks, they need some assumptions, because if they assumed less airflow, they'd need bigger and more elaborate heatsinks.

However, I'm not using SuperMicro cases.

What type of airflow must I generate, and where, to create the right environment for SuperMicro boards?

2. In looking at various 140mm Noctua PWM fans, I'm noticing that different models are essentially the same fan but with different RPM ranges. Why is that? Since the fan is controlled by PWM anyway, why not just have one fan model and have the desired speed be selected by the PWM? Instead, since I've been groping in the dark regarding question #1, it would seem I have to buy different fan models until I find ones that are "just right" to satisfy Goldilocks--er, I mean the motherboard components--while at the same time trying to keep noise as low as possible by keeping RPM's as low as possible.

 

[Deci]

I mounted a fan to the top of the heatsink in my a1sai as the airflow through the ds380 it's in wasn't keeping acceptable temps with its passive cooler, it's hard wired at 7v, the fan is barely moving anything but it's enough to keep temps a couple of degrees off ambient.

Their assumptions are that it's going in a server chassis/their chassis and noise isn't a concern/factor so as much air as the fans can pull/push through it, I have noticed that there is also limited fan speed options on the bios/ipmi, normal (ramps fans based on temp)- full speed (self explanatory) - high io (more middle ground in terms of speeds and ramps on something other than cpu temp)

I found that having the fans on normal mode with pwm capable fans (3 wire doesn't ramp speed at all only 4 wire ones do) resulted in the fans dropping below 500rpm (12cm) and the motherboard considers anything below 500rpm to be a fault and ramps to full speed, then the fault goes away and it slows down (repeating this cycle endlessly), the fans I am using are capable of 2400rpm at full speed, high io mode leaves them hovering around 1400-1500 which is quiet enough.

Without any complicated manual pwm control the point of having different max speed fans makes sense, pick a fan that at max moves all the air you need and reduce noise that way. if you can map your own speed/temp maps it makes less sense to have lots of options, at the end of the day server gear is intended to be in a chassis with huge amounts of airflow, either make some custom ducting or mount fans in such a way that they can blow directly onto any points of concern.

 

[NeverDie]

I found that having the fans on normal mode with pwm capable fans (3 wire doesn't ramp speed at all only 4 wire ones do) resulted in the fans dropping below 500rpm (12cm) and the motherboard considers anything below 500rpm to be a fault and ramps to full speed, then the fault goes away and it slows down (repeating this cycle endlessly)....

As it happens, yesterday I plugged a Noctua fan into my SuperMicro motherboard to test how loud it was, and I observed this pulsing phenomenon you just described. I thought it was strange because the fan was supposedly designed to have its speed vary only in the range of 1200 to 1500 RPM (i.e. zero pulse width should map to 1200rpm, full pulse width to 1500RPM, and so forth), and yet it did seem to drop far below 1200rpm, almost as though the power had been turned off to it. Then the IPMI sensor monitor would show it's status as red, and then the fan would go to high RPM, the status would go to green, then decrease RPM, and then the cycle would repeat, going to red again, and so forth as you described. Inside the fan kit there was a "low noise adapter," so I blithely plugged that in in-line to see what would happen (after all, how much worse could it get?), and voila, it seemed to make the problem go away. Well, at least the pulsing cycle went away. I can't be sure the PWM speed control worked as intended with the low noise adapter, because I didn't look into it further, as I decided to return the fan for independent reasons. So, FWIW, perhaps it's a clue.

 

[Deci]

I'm fairly sure the low noise adaptor is simply a half watt resistor with plugs on either end underneath the sleeving, as that's a trick that's been around for a while to drop the noise of a fan without any kind of rpm control.