setting ipmi fan sensor threshold to 0 for fans that auto stop

[jpmomo]

I am trying to get an ek mlc water cooler to work on a supermicro h11ssl-nc mb. the ek radiator has the new vardar evo fans that completely shutoff when the temps on the cpu go down. I have been able to adjust the fan sensors via the ipmiutil for fans that spin slow but not that completely shutoff. I am appending the following cmd: -n 41 -u 0:0:0:2200:2200:2500. I am setting the fan mode to optimal settings. when I first power on the server, the fans are off and the sensor readings show N/A for the fan and Not Present! for the Reading. As soon as I toggle the fans on by either letting it get up to speed or switching to full speed mode, the fans will spin up to the expected rpm. but when i try to put it back into optimal mode, the fans begin to throttle up and down with a Lower Non-recoverable 0 R.P.M. Reading. Is there a way to disable the Lower Non-recoverable sensor or maybe put a negative number?
thanks for any help,
jp

 

[EffrafaxOfWug]

I ran into a similar problem when using SM IPMI with some Noctua fans; once they fell below their ~300rpm lower threshold, they'd report as 0rpm and the LNR assert would go off, ramping all the fans up to maximum. It seems like your fans do this by design.

Even setting the IPMI threshold to zero won't fix it and as far as I'm aware there's no way to disable the asserts. I had to find some fans that played nice with the IPMI. Otherwise I think your only options would be to either a) run the fans off a 3-pin connection or b) otherwise spoof the PWM information on pin 4; whether this is easily doable will depend on the design of your cooler I guess.

Edit: from looking at this review of an EK AIO (dunno if it's the same one you're using), the radiator itself features a PWM backplane-cum-fan-controller of some sort, and there's a single PWM connector just for the duty cycle info connecting to the motherboard. If yours is the same/similar, does the rad still function correctly if you disconnect this (as I assume it has its own internal temperature monitoring)?

EK-MLC Phoenix 360 Review

Edit 2: from reading EKWB's spec it seems like their backplane might be dependent on the PWM signal from the motherboard and removing it will just make the fans run at full pelt - worth trying I think but if true it means if you want the fans to stop when under 20% duty cycle you'll need to mount them on a non-mobo/IPMI PWM connector from a separate fan controller or suchlike. Otherwise you'd need to modify the IPMI PWM duty cycle so it never drops below 20%, and I don't think the SM IPMI implementation supports this - with raws you can apply a max duty cycle but I don't think you can specify a min.

What is the Start-Stop function and how to use it? - ekwb.com

 

[jpmomo]

I ran into a similar problem when using SM IPMI with some Noctua fans; once they fell below their ~300rpm lower threshold, they'd report as 0rpm and the LNR assert would go off, ramping all the fans up to maximum. It seems like your fans do this by design.

Even setting the IPMI threshold to zero won't fix it and as far as I'm aware there's no way to disable the asserts. I had to find some fans that played nice with the IPMI. Otherwise I think your only options would be to either a) run the fans off a 3-pin connection or b) otherwise spoof the PWM information on pin 4; whether this is easily doable will depend on the design of your cooler I guess.

Edit: from looking at this review of an EK AIO (dunno if it's the same one you're using), the radiator itself features a PWM backplane-cum-fan-controller of some sort, and there's a single PWM connector just for the duty cycle info connecting to the motherboard. If yours is the same/similar, does the rad still function correctly if you disconnect this (as I assume it has its own internal temperature monitoring)?

EK-MLC Phoenix 360 Review

Edit 2: from reading EKWB's spec it seems like their backplane might be dependent on the PWM signal from the motherboard and removing it will just make the fans run at full pelt - worth trying I think but if true it means if you want the fans to stop when under 20% duty cycle you'll need to mount them on a non-mobo/IPMI PWM connector from a separate fan controller or suchlike. Otherwise you'd need to modify the IPMI PWM duty cycle so it never drops below 20%, and I don't think the SM IPMI implementation supports this - with raws you can apply a max duty cycle but I don't think you can specify a min.

What is the Start-Stop function and how to use it? - ekwb.com

Thanks for the helpful info (and quick!) I did some more tweaking today and found that if I switched the first fan (yes, I am using the phoenix 360 and you linked to above) with a noctua a12x25 I sort of got it to do what i wanted. The noctua would always spin but the other 2 fans would stop and then start after the system got hot (running prime for a really long time.) If the system is left at idle, it only gets to around 30C. the noctua stays at 400 rpm and the other 2 vardar evo fans (ek brand that came with the rad) are stopped. the pump is also connected to this pwm controller. I am not sure what the pump is spinning at but it does sound like it is circulating. I am still not sure what exactly is going on as the noctua is supposed to have the same start/stop capability. I sent an email to the ek folks who are usually pretty helpful (although not as quick as some of the folks on this forum!)

 

[EffrafaxOfWug]

From looking at the pics the EK uses a simple shared PWM backplane so as long as one of the fans is constantly running you shouldn't ever run into the scenario where all the fans stop, so glad you found an acceptable workaround so quickly

Wasn't aware that Noctua's also do the start/stop thing - mine didn't, they'd still run at super-low rpm but report 0rpm to IPMI, hence triggering the assert. But it's possible your behave differently to mine.

I would have assumed that even if the fans weren't running, the pump always would be...? My assumption from reading about the start/stop function was that when load was low the water would still flow but rely on passive rather than active cooling through the rad - their blurb doesn't mention anything about turning off the pump.

(As an aside, personally I prefer fans spinning at their minimum rpm rather than turning on and off - generally speaking a fan going from 300 to 700rpm is only marginally noticeable, whereas going from 0 to 700rpm is usually quite pronounced)

 

[jpmomo]

From looking at the pics the EK uses a simple shared PWM backplane so as long as one of the fans is constantly running you shouldn't ever run into the scenario where all the fans stop, so glad you found an acceptable workaround so quickly

Wasn't aware that Noctua's also do the start/stop thing - mine didn't, they'd still run at super-low rpm but report 0rpm to IPMI, hence triggering the assert. But it's possible your behave differently to mine.

I would have assumed that even if the fans weren't running, the pump always would be...? My assumption from reading about the start/stop function was that when load was low the water would still flow but rely on passive rather than active cooling through the rad - their blurb doesn't mention anything about turning off the pump.

(As an aside, personally I prefer fans spinning at their minimum rpm rather than turning on and off - generally speaking a fan going from 300 to 700rpm is only marginally noticeable, whereas going from 0 to 700rpm is usually quite pronounced)

your post got me thinking about the fan controller. reading through the docs it states that you can daisy chain the fan controllers. I didn't want to have to take one of the fans off of the rad and replace it with the noctua as it is a bit of a pain and would not be ideal. I wound up using a noctua pwm splitter (y cable) and using a small noctua fan to cool the pcm (or whatever the small heatsink is on these amd epyc boards). This seems to be a better solution as I can leave the ek components as is and just add the extra small fan to the y cable. The plan is to use 2 of these units in a dual socket epyc mb. I am getting a massive case to fit everything in and am hoping that the hose length works out. At least the fan issue got sorted!
Thanks again for giving me some ideas,
jp

 

[Tooms]

Do you use the lower p option to store the IPMI value over boot ?

-n 41 -u 0:0:0:2200:2200:2500 -p

I made an video about it for my build

/Tooms