ZeKefried

Your fan is cycling because it's lower operational speed is below Lower Critical threshold, not because the temps are low. The bmc thinks that a fan has died, ramps up power to it just in case, fan crosses the threshold, temps are low so dial down the fan and go to step one. You can confirm this by opening Event Log tab in IPMI web interface:
 

Obscuring the fan speed from your motherboard is one of the dumbest solutions. Especially on a server board that will be stuck at least as remote as under the table. You are robbing yourself of monitoring that you paid for. Second thing to consider is that you introduce another failure point with that fan hub. It is chained with the fan and probability that the chain dies is rather more that the sum of probabilities that fan and hub die.
 
Back to the original problem.
 
Solution 1: Replace the fan with a fan that doesn't go so low on rpm
 
Solution 2: Set the thresholds according to your fan. My particular version of ipmi doesn't provide a web interface to do this so:
ipmiutil sensor -N <ip_address> -U <user_name> -P <password> -n <S_num> -u <thresholds>  
<thresholds> are provided as noncrit_lo:crit_lo:nonrec_lo:noncrit_hi:crit_hi:nonrec_hi for a fan with 300-1300 rpm range i would use 250:150:50:1300:1400:1500

Your fan is cycling because it's lower operational speed is below Lower Critical threshold, not because the temps are low. The bmc thinks that a fan has died, ramps up power to it just in case, fan crosses the threshold, temps are low so dial down the fan and go to step one. You can confirm this by opening Event Log tab in IPMI web interface:
 

Obscuring the fan speed from your motherboard is one of the dumbest solutions. Especially on a server board that will be stuck at least as remote as under the table. You are robbing yourself of monitoring that you paid for. Second thing to consider is that you introduce another failure point with that fan hub. It is chained with the fan and probability that the chain dies is rather more that the sum of probabilities that fan and hub die.
 
Back to the original problem.
 
Solution 1: Replace the fan with a fan that doesn't go so low on rpm
 
Solution 2: Set the thresholds according to your fan. My particular version of ipmi doesn't provide a web interface to do this so:
ipmiutil sensor -N <ip_address> -U <user_name> -P <password> -n <S_num> -u <thresholds>  
<thresholds> are provided as noncrit_lo:crit_lo:nonrec_lo:noncrit_hi:crit_hi:nonrec_hi for a fan with 300-1300 rpm range i would use 250:150:50:1300:1400:1500

Your fan is cycling because it's lower operational speed is below Lower Critical threshold, not because the temps are low. The bmc thinks that a fan has died, ramps up power to it just in case, fan crosses the threshold, temps are low so dial down the fan and go to step one. You can confirm this by opening Event Log tab in IPMI web interface:
 

Obscuring the fan speed from your motherboard is one of the dumbest solutions. Especially on a server board that will be stuck at least as remote as under the table. You are robbing yourself of monitoring that you paid for. Second thing to consider is that you introduce another failure point with that fan hub. It is chained with the fan and probability that the chain dies is rather more that the sum of probabilities that fan and hub die.
 
Back to the original problem.
 
Solution 1: Replace the fan with a fan that doesn't go so low on rpm
 
Solution 2: Set the thresholds according to your fan. My particular version of ipmi doesn't provide a web interface to do this so:
ipmiutil sensor -N <ip_address> -U <user_name> -P <password> -n <S_num> -u <thresholds>  
<thresholds> are provided as noncrit_lo:crit_lo:nonrec_lo:noncrit_hi:crit_hi:nonrec_hi for a fan with 300-1300 rpm range i would use 250:150:50:1300:1400:1500