I seemed to have run into the PWM Fans Revving problem described by others here on the their SuperMicro X9 and X10 motherboards, with one difference being that I'm using a
SuperMicro X11SSM-F motherboard. Also, I'm using an Intel E3-1245 V5 boxed processor, with the packaged heatsink and PWM fan (can't seem to find specs on that fan). I picked up a trio of Noctua fans (
NF-F12 PWM,
NF-P12 PWM and
NF-A14 PWM), in the hopes of quieting my system a bit further than what the three
Fractal Design Silent Series R2 non-PWM fans were producing. I've tried using:
- IPMI Web/GUI interface (changing between Optimal and Standard)
- ipmitool in Part 1 of @Ericloewe 's post on how to change sensor thresholds
- ipmiutil mentioned in this thread (Part 2)
Both the ipmitool and ipmiutil utilities did change the threshold settings, according to the sensor listings, and I honestly can't tell if the IPMI Web/GUI interface made any difference (from what I read, it's just changing the default power setting between 30% and 50%). Before changing any settings, when I turned on the system, the fans each went to a high RPM, but once I power-cycled the system (unplugging the system to get IPMI to reboot/reset) it started revving instead. :(
Currently, my fans are configured thus:
Code:
ipmiutil ver 3.01
isensor: version 3.01
Connecting to node 192.168.0.211
-- BMC version 1.13, IPMI version 2.0
ID | SDRType | Type |SNum| Name |Status| Reading
025f | Full | Fan | 41 | FAN1 | OK | 1900.00 RPM
| Entity ID 29.1 (Fan), Capab: arm=auto thr=write evts=state
| Volatile | hi-unrec 25500.00 | hi-crit 25400.00 | hi-noncr 25300.00 | lo-noncr 700.00 | lo-crit 500.00 | lo-unrec 300.00
| SdrThres | hi-unrec 25500.00 | hi-crit 25400.00 | hi-noncr 25300.00 | lo-noncr 700.00 | lo-crit 500.00 | lo-unrec 300.00
| nom 12800.00 | nmax 17000.00 | nmin 2000.00 | smax 25500.00 | smin 0.00
02a2 | Full | Fan | 42 | FAN2 | OK | 1400.00 RPM
| Entity ID 29.2 (Fan), Capab: arm=auto thr=write evts=state
| Volatile | hi-unrec 1900.00 | hi-crit 1800.00 | hi-noncr 1700.00 | lo-noncr 200.00 | lo-crit 100.00 | lo-unrec 0.00
| SdrThres | hi-unrec 1900.00 | hi-crit 1800.00 | hi-noncr 1700.00 | lo-noncr 200.00 | lo-crit 100.00 | lo-unrec 0.00
| nom 12800.00 | nmax 17000.00 | nmin 2000.00 | smax 25500.00 | smin 0.00
02e5 | Full | Fan | 43 | FAN3 | Absent | 0.00 na
| Entity ID 29.3 (Fan), Capab: arm=auto thr=write evts=state
| SdrThres | hi-unrec 25500.00 | hi-crit 25400.00 | hi-noncr 25300.00 | lo-noncr 700.00 | lo-crit 500.00 | lo-unrec 300.00
| nom 12800.00 | nmax 17000.00 | nmin 2000.00 | smax 25500.00 | smin 0.00
0328 | Full | Fan | 44 | FAN4 | OK | 1200.00 RPM
| Entity ID 29.4 (Fan), Capab: arm=auto thr=write evts=state
| Volatile | hi-unrec 1600.00 | hi-crit 1500.00 | hi-noncr 1400.00 | lo-noncr 200.00 | lo-crit 100.00 | lo-unrec 0.00
| SdrThres | hi-unrec 1600.00 | hi-crit 1500.00 | hi-noncr 1400.00 | lo-noncr 200.00 | lo-crit 100.00 | lo-unrec 0.00
| nom 12800.00 | nmax 17000.00 | nmin 2000.00 | smax 25500.00 | smin 0.00
036b | Full | Fan | 45 | FANA | OK | 1400.00 RPM
| Entity ID 29.5 (Fan), Capab: arm=auto thr=write evts=state
| Volatile | hi-unrec 1900.00 | hi-crit 1800.00 | hi-noncr 1700.00 | lo-noncr 200.00 | lo-crit 100.00 | lo-unrec 0.00
| SdrThres | hi-unrec 1900.00 | hi-crit 1800.00 | hi-noncr 1700.00 | lo-noncr 200.00 | lo-crit 100.00 | lo-unrec 0.00
| nom 12800.00 | nmax 17000.00 | nmin 2000.00 | smax 25500.00 | smin 0.00
FAN1 is the Processor/Heat Sink fan, FAN2 is the NF-F12 PWM, FAN3 is empty, FAN4 is the NF-P12 PWM, and FANA is the NF-A14 PWM.
Taking the Noctua RPM specs for each fan, and what
@Ericloewe posted, I believe my settings with ipmiutil should be as follows:
- Noctua NF-F12 PWM (FAN2)
Code:
ipmiutil sensor -N 192.168.0.211 -U user -P pass -n 42 -l 240 -h 1650
- Noctua NF-P12 PWM (FAN4)
Code:
ipmiutil sensor -N 192.168.0.211 -U user -P pass -n 44 -l 240 -h 1430
- Noctua NF-A14 PWM (FANA)
Code:
ipmiutil sensor -N 192.168.0.211 -U user -P pass -n 45 -l 240 -h 1650
IPMI seems to round to the nearest 100, which is why the lower thresholds went to 200, 100 and 0. Alternatively, I tried using ipmitool:
- Noctua NF-F12 PWM (FAN2)
Code:
ipmitool sensor thresh "FAN2" lower 200 100 0
- Noctua NF-F12 PWM (FAN2)
Code:
ipmitool sensor thresh "FAN2" upper 1700 1800 1900
- Noctua NF-P12 PWM (FAN4)
Code:
ipmitool sensor thresh "FAN4" lower 200 100 0
- Noctua NF-P12 PWM (FAN4)
Code:
ipmitool sensor thresh "FAN4" upper 1400 1500 1600
- Noctua NF-A14 PWM (FANA)
Code:
ipmitool sensor thresh "FANA" lower 200 100 0
- Noctua NF-A14 PWM (FANA)
Code:
ipmitool sensor thresh "FANA" upper 1700 1800 1900
I also tried using the -u option with ipmiutil (e.g.
Code:
ipmiutil sensor -N 192.168.0.211 -U user -P pass -n 42 -u 200:100:0:1700:1800:1900
), as well as varying the lower values to get rid of the 0 and try different values (e.g. 300:200:100, 200:100:100). After that I tried playing with the upper values as well, but with no successful results at stopping the revving. In between each attempt I:
- powered off the system
- flipped the power-switch off on the power-supply
- unplugged the power
- waited 5 or more seconds
- plugged the power back in
- flipped the power-switch on on the power-supply
- waited for IPMI to startup
- booted the system
So the system seems to be taking the threshold values, updating its internal tables, and putting those new values into service when the system is power-cycled. But, checking the Event Log, it seems to keep getting "Lower Critical - Going Low - Assertion", and "Lower Non-Recoverable - Going Low - Assertion" events, before kicking the fans up. From what I can see in the IPMI Web/GUI interface, the sensors are reporting the RPMs on FAN2 and FAN4 varying between 0, 200 and 300 on the low end, when it gets the events.
Anyone have any ideas on how to get this poor system to not rev the fans, and still utilize the PWM feature of the fans?
Also, a note to anyone with an X11SSM-F motherboard, while poking around the IPMI Web/GUI interface, I discovered under Maintenance there is an option called Unit Reset, with the description "Reboot IPMI Device"; this seems to have the same effect on the system as power-cycling, and might be worth further investigation.