SOLVED
This will sound [ Counter Intuitive ] but trust me it works.
I have an ML110 G6 with a Xeon 3440 CPU, and 16 GB of RAM.
I thought the [ Rear Fan ] was loud and went looking for a solution. I tried setting a threshold in the IPMI BMC controller using one of the network IPMI tools. But that did not work. The IPMI thresholds are only for reporting or intiaiting actions based on crossing the thresholds, not for controlling the Fan speed.
I now believe it is the BIOS and something called the [ Fan Curve ] in the BIOS version that controls how it ramps up and down and where it stablizes based on Ambient room temperature (there is a sensor at the front under the Bezel) or whether there are additional Option Cards installed.
What [ Works ] for me is to go into the BIOS at boot POST time, you must press [ F10 ] then go into the [Advanced] column menu, then go into the IPMI submenu.
[Main]->[+Advanced+]->[+IPMI+]
There is a option called the [Blow-Out feature].
If you read about it it says it will [ Run the Fans at High Speed ] presumably to "Blow-Out" the dust in the chassis and then settle down to a lower RPM if -->[+Enabled+]. I would guess if there is a Management agent installed on the Operating system it could then take up the task of monitoring sensors and ramping the Fans speed up and down appropriately.
I thought "Disable" would lead to a Quieter start up and run, It Does Not
If "Disabled" the Fans will run at "High Speed" until told otherwise by an HP Management Agent in a supported version of a supported operating system. If you run an unsupported Operating system without a Mangement Agent, or do not install an HP Management Agent installed on a supported Operating system. The Fans will start and run at "High Speed" all of the time.
I also suspect if anything like "antivirus" or a million other accidental things interferes or prevents the Management Agent from controlling the Fans, the Fans will always run at "High Speed" all of the time.
The safe option is to always leave the "Blow-out feature" [+Enabled+] its the only real control you have over the Fans without some special software agent that knows how to control them.
The Rear Fan still runs high, around 2000 RPM versus 4000 RPM, but it is audibly much more tolerable.
The Rear Fan is a standard 90 mm x 25 mm with a Tach sensor and PWM control, and a locked Rotor Sensor.
The IPMI thresholds seem to suggest anything below 900 RPM would log an event in the SEL.
And a locked Rotor signal (+V ) would keep the system from starting. This is to prevent thermal damage should the fan be unable to turn, for example if something gets jammed in it from the outside.
The pin out is also non-compliant with the Intel specs for Fan pinouts
A few people on the Internet have mapped the pin outs and they come down to approximately the same number:
Tach
PWM
Locked (Rotor) Sensor
+12V
Grd
It is the Locked Rotor Sensor that I suspect keeps people from successfully swapping out the Fans for quiter options.
A Warranty replacement Fan can be purchased, but it will conform to the same audible characteristics already observed.
A few people on the Internet have reported success at wiring a custom crossover connector between the motherboard and a new Fan with lower audible accoustic properties, similar to the Nocturna line of Fans. Which would most likely violate your warranty.
But making sure the "Blow-out feature" is enabled has made it possible for me to update the BIOS and the LO-100 firmware up to the latest and still have the fans loud, but tolerable. Its a far better option than letting them run at Full High Speed for extended periods of time.
Note: Resetting the BIOS, removing the battery, ect.. will I assume return the BIOS to its default settings, which will also reset the "Blow-out feature" to the [+Enabled+] position. After listening to the Fans whine at their top speed for a while in frustration, "any" relief can sound quite major.
After thinking about this for a while it makes a kind of good sense.
The default position leaves the system in a "relatively" quieter state if there is nothing that will be subsequently starting up to control the Fans. However the best control (and possibly the lowest RPMs) would be attained if the Management agent comes online and appropriately throttles the Fan speed back as much as possible.
The LO-100/BMC controlling the fans appears to be made by ServerEngines Pilot II by a company in San Jose, CA which was recently acquired by Emulex to combine their Converged Network adapter tech with their Fiberchannel adapter technology. So there may not be a lot of after market information available to write a custom control program.
The best examples of that would probably be the finished Management Agents provided by HP.
I'll go out on a limb and [guess] that the HP Management agents "might" expose an SNMP OID or WBEM attribute in the CIM which could be used to [set] the Fan motor PWM or equivalent RPM, to grant greater control over the Fan speed. But that also assumes you have the appropriate "agent" or "provider" and know where to look.
HP makes some MiB schemas available for their management tools, like SIM -Systems Insight Manager, there might be something that could be "found" walking the MiB with snmpwalk or browsing the CIM that would expose a controllable parameter.. but it does require you to have the HP provided software already installed.
So I chose to carefully look at my options for replacing the Rear Fan.
It turned out the Rear Fan connector had only four "connected" wires out of a six pin connector on my ML110 g6:
1. Black
2. Red
3. Yellow
4. Blue
The two outer pins were connected to nothing.
The Black wire was closest to the back of the chassis.
I purchased a [ Noctua 92 x 14 mm Low-Profile Cooling Fan with A-Series Blades (NF-A9x14) pwm]
It had four wires with the Noctuna website indicating the wires were for:
1. Black - gnd
2. Yellow - +12v
3. Green - tach
4. Blue - pwm
It made a kind of sense to me they might universally translate.
Black and Blue and their positions made sense, red is almost always +V, so 3 out of 4 guess seem to make a kind of sense.
And although the connector on the Noctuna was a four pin not a six, it could be plugged into the 6 pin header on the mother board, leaving the outer pins unconnected.
Since the original Fan was "Black wire towards back of chassis" I oriented the new Fan connector the same way, and plugged it into the motherboard. Centering the Fan connector to "skip" and not connect to the first or last outer pins on the motherboard connector.
Removing old Rear Fan fom the chassis was not hard, but not easy either, silcone stabs just needed some pulling and shoving to get them to deform and remove without damage.
The Noctuna came with fresh silcone stabs, and are a different color so nothing got mixed up.
I powered on the system (it did not fault, it did not refuse to start up). Then ran an IPMI command to check the SDR report:
# ipmitool -I lanplus -H 192.168.2.156 -U admin123 -P admin123 sensor get 'REAR FAN'
Locating sensor record...
Sensor ID : REAR FAN (0x10)
Entity ID : 29.1
Sensor Type (Analog) : Fan
Sensor Reading : 1242.529 (+/- inf) RPM
Status : ok
Lower Non-Recoverable : na
Lower Critical : na
Lower Non-Critical : na
Upper Non-Critical : na
Upper Critical : 959.233
Upper Non-Recoverable : na
Assertion Events : unc- unc+
Assertions Enabled : unc- unc+
So the new Fan was recognized and the Tach signal was interpreted. The PWM signal was understood by the Fan since, the behavior during BIOS startup at POST is the same, and I do have the "Blow-out" feature enabled. So first it revs the Fan at High Speed and then reduces RPM speed.
The end result is a much quieter system.
I did not measure the sound dB level before the Fan replacement, so I cannot say objectively how much quieter, but it is noticable.
The Lights-Out 100 [Monitoring Sensors] page says "Normal Operating Range" and has a Green Check Mark.
I realized while I was testing with both the old Rear Fan and CPU Fan, the CPU Fan is a major contributor to noise.
The CPU Fan is mounted to a tall Aluminum Cooling tower of very edged fins for dissipating heat.
I'm much less likely to attempt to remove the CPU Cooler and replace it now that the Rear Fan has been replaced.
It still whines a bit, but no more than a normal desktop PC and perhaps a bit less.
Its also notable that my ML110 g6 is somehwat older now, and I noticed the old Rear Fan has a bit of a "Ka'chunk" to it as I manually turn the blades. It is a ball bearing fan, where the Noctuna is a Self Stablizing Oil pressure bearing.
Being different could account for the difference in noise, or the old Rear Fan could have simply been in need of replacement, parts do wear out.
Two more experiments; and apparently more success.
1. I purchased a Noctua 9F-B9 PWM Fan and studied the CPU Cooler that came with the ML110 G6. Its a 92 MM Fan with a cowl to direct the air flow through the fins of the Cooler. After a little disassembly I replaced the 92 mm Fan that came with the Noctua. It was a perfect fit. But when I connected the PWM connector to the CPU header on the mainboard, as I did with the Rear Fan (skipping the first and last pin, then connecting, the Black wire towards back of chassis). It kept running at such a slow speed the IPMI Threshold for Lower Uncorrectable Critical (lcr) kept crossing and eventually within about 60 seconds shut the system down. So this did not work. But after thinking.. I got an idea.
2. The Noctua advertises their PWM fans will work in a 3-Pin and 4-Pin environment. The 3-Pin being everything but a Speed Control (PWM) signal wire, that is.. they still send a Tach signal.. along the same wire, in the same position as the original 4-Pin configuration. [And the Best part] - is the 4-Pin PWM Fan running in a 3-Pin scenario runs at its Top speed, which for this fan is 1600 RPM, far below the 2200 RPM I had seen with the original Fan.. so not only should it be quieter, it should still push a lot of air through the original fins. And the special Commutator and PWM chip they advertise should contribute to make it quieter.
Note: The Fan Tach signal would still be sent to the mainboard, only it would be "fixed" since the PWM signal from the mainboard would never reach the fan, and the Tach signal would be above the 1200 RPM, below which it throws an alert and eventually shutdowns. All criteria satisfied.. it should just work.
From my experience the CPU sensor reports 30 degrees C, so I would use that to compare to see if this was at least as good, or how much worse.
To make all this happen, I simple moved the Pin for the PWM signal out of the way and reconnected.
1. Black - gnd
2. Yellow - +12v
3. Green - tach
X. Blue - pwm
To make a long story shorter, it appears to have worked.
Here are the results after running with both fans replaced and the case closed up for about 1 hour
There have been no alerts, no shutdowns - and its a lot quieter.
A few things to keep in mind are:
A. This almost certainly voids the warrantly, but my system is a personal purchase made several years ago, and I couldn't bear to run the system at home with the terrible noise it was making. It was quite literally collecting dust and devaluating anyway.
B. The Noctua fans are not the least expensive, but also not the most expensive, nor are they hard to find. And they have like a 6 year warranty.. so there is a chance they are good at what they do.
C. I learned there are a lot of variables to the problem. The BIOS [Blow-out] feature, running with and without HP Health Driver Mangement agent. The lack of control over the IPMI thresh limits for the fan speed alerts. A lot could be exposed in the BIOS, and is on other models.. but its just not there for this model.. so using after market fans is difficult. The Xeon 3440 CPU has a Thermal Design Power (TDP) of 95 watts meaning the Cooler has to be sufficient for handling that.. and its a tight fit on a microATX mainboard, therefore finding a different cooler that would fit was problematic.. so retro fitting the original just made sense.
Also note: This in "no way" equips it for Overclocking. If anything under load it "may" be more limited.
Its also sensitive to the other Temperature sensors for the mainboard, and the outside through an Ambient Temperature sensor mounted under the front fascade, with a long black and red wire that runs back into the chassis and connects to the mainboard.
I don't exactly recommend people do this.
Since so many people seemed to be frustrated, and had little success.. I just wanted to see if it could be done.
One last comment however, replacing the [Rear Fan] was [very] noticable.
Replacing the [CPU Fan] was "sweating bullets" down right "Scary" it was so silent.. I keep checking the CPU Temp and Fan speed.