Dell R730xd 100% loud fan speed even with server room at 20 C/68 F temperature

[uberguru]

Server room is 68 Fahrenheit or 20 Celcius and yet the dell R730xd server is running fan at 100% loud as hell

the 2 CPU are E5-2640 v4 90 watts CPU and have a mellanox connectx-5 2 x 100G network card and only 3 drives populated of the 12 drives slots and 2 x reade OS drives

why is the fan running at 100% and how do i fix this?



Please see screenshots below

 

[i386]

Do you have third Party/Not dell branded add in cards in the system?

 

[uberguru]

Do you have third Party/Not dell branded add in cards in the system?

i mentioned the mellanox connectx-5 network card

that is the only PCIe device i added to the server
nothing else besides the 5 drives, 2 redundant OS drives and 3 other drives
the 2 OS drives 400GB are Dell branded drives but the other 2 western digital 20TB sata hdd and 1 lenovo sas ssd 3.2TB
see screenshot below




i still dont understand why they cant provide a way to custom control the fans
where i can control the fan speed, why is this rocket science for them to implement?
i mean for me server room temperature is regulated so i dont need the fan blowing loud when simple temperature is not enough for this thing to stop blowing at max speed

 

[marcoi]

I have same server and the fans will run way less then 100%
try changing fan settings


you can also run the live manager (i forgot the exact name it an option at boot up) to run inventory so the system can confirm add-on etc.
If you cannot change the fan setting in idrac web gui, go into the bios to change the thermal profile.

if your still having issue ill start up my server and grab more screen shots.

 

[uberguru]

I have same server and the fans will run way less then 100%
try changing fan settings

what percentage are your fans running at and do you have PCIe cards also installed?

when i tried to set the fan settings similar yo your from your screenshot, i got this error




i used value 20 and rebooter and still same 100% fan speed

 

[oldfett]

SSH into your iDRAC and run the following command so its not ramping the fans for your Mellanox cards.

racadm set system.thermalsettings.ThirdPartyPCIFanResponse 0

This isn't going to suddenly drop it to 20%, but it should make a difference. After you run this you should check your card temperatures after a day or so and make sure they are in a comfortable range.

Your fans are also going to be running at a higher RPM due to the two SSD's in the back (ramps to make sure they get good airflow). So if you are never going to fill the front bays, moving the SSD's to the front (with appropriate adapters) would further lower the fan RPMs.

 

[marcoi]

i grabbed some screenshots of the bios, not sure it will help




Here are the fans running ESXI 8

 

[marcoi]

my system has a usb card and a lot of drives in the front of the unit

 

[uberguru]

am suspecting it is because of the mallanox connectx-5 card
it is dual 100G and it is possible that it is behaving like this

here is a R630 with 2 x E5-2699A v4 that are 145 watts each which is much higher than the 90 watts of the R730xd
and here is the fan of that one

so i am thinking it is the mellanox card that may be causing this

i mean what do i do here because the server room is very well cooled, i need to reduce the noise of those fans

 

[marcoi]

am suspecting it is because of the mallanox connectx-5 card
it is dual 100G and it is possible that it is behaving like this

you could take the card out, run the life manager to re inventory the system then see if the fans drop.

 

[uberguru]

you could take the card out, run the life manager to re inventory the system then see if the fans drop.

i ran the server before when i had just CPU #1 and the card was not detected and it was not running at 100%
so i am 90% sure it is the mellanox card

but i need the card and i need the fan to go down

so what do i do here?

 

[oldfett]

so what do i do here?

See my above comment on the command to disable third party PCI component fan ramp (iDRAC isn't sure what the card is and what ammount of cooling it needs, so cranks the fans). It should take effect within seconds, if you decide to revert just run the same command with a 1 rather than a 0.

 

[uberguru]

SSH into your iDRAC and run the following command so its not ramping the fans for your Mellanox cards.

racadm set system.thermalsettings.ThirdPartyPCIFanResponse 0

This isn't going to suddenly drop it to 20%, but it should make a difference. After you run this you should check your card temperatures after a day or so and make sure they are in a comfortable range.

Your fans are also going to be running at a higher RPM due to the two SSD's in the back (ramps to make sure they get good airflow). So if you are never going to fill the front bays, moving the SSD's to the front (with appropriate adapters) would further lower the fan RPMs.

i totally missed this
ok will try the racadm command

i need the rear drives, will be populating all 12 drive bays in front, the rear drives are the OS drives
and are SAS SSD do not sure howm uch heat they generate really and OS drives not doing much work

as long as fan blowing some wind, i should be good, the server room is kept cool and chilled
i dont need the fans to be doing much work, just need fan to direct air flow so dont need them running at high rpms

i can lower the server room temperature to 66 F if i want or even lower just to shut these fans up lol

 

[oldfett]

Understood, if you need the drives you need them. I just know from experience playing with an R730XD I was amazed how just removing those rear drives dropped my fan RPMs by 20% or so.

I would think disabling that third party fan resonse (your Mellanox cards may or may not be Dell branded but unsure if that generation would know what it is and just treat it like a third party) should get you down into the same range as your R630, which should make the noise much much better.

I don't think lowering the room temp will help much unfortunately. After reaching a certain inlet temp the temperature doesn't seem to play a ton into the fan RPM role. I think the fan tables are more focused on meeting specific CFM rather than temperature setpoints based on the inventory in the box. So for the rear drives it knows they are present but isn't really checking their temperatures and adjusting accordingly, just blastng air to ensure they get get enough to cool them even with all the pre-heat from everything ahead of them heating things up.

There are a few threads online for manually setting the fan RPM, I'm not a huge fan of this as its locked to that RPM unless you have a very verbose script monitoring and adjusting accordingly. For safety reasons I have never even tried this.

 

[uberguru]

honestly i wouldnt mind setting the fans speed to 40% and call it a day
the room like i said has a minisplit and it is like a freezer if i drop the temperature to like 66 F

So all i need is just direction of air i dont need it to do much more than that

so if i find a way to set the fan speed manually, i will do it!!!

 

[uberguru]

There are a few threads online for manually setting the fan RPM, I'm not a huge fan of this as its locked to that RPM unless you have a very verbose script monitoring and adjusting accordingly. For safety reasons I have never even tried this.

if you find any resources about manually setting the fan speeds, please share them

thanks!

I am highly disappointed in dell for not making this fans smart enough to see what temperature of room is and adjust accordingly
the very old ibm/lenovo blade server i had does that, it is whisper quiet when the room is well cooled and increases when the temperature goes higher

why is this rocket science for dell to implement?

 

[oldfett]

if you find any resources about manually setting the fan speeds, please share them

thanks!

I am highly disappointed in dell for not making this fans smart enough to see what temperature of room is and adjust accordingly
the very old ibm/lenovo blade server i had does that, it is whisper quiet when the room is well cooled and increases when the temperature goes higher

why is this rocket science for dell to implement?

Here is a page I have seen on this, it seems detailed. Again I haven't done this and if you do follow it, would use caution:
PowerEdge G13 Fan Control Issues - A Quick Guide

I'm torn on you being dissapointed though, I understand what you are saying and as a homelabber I agree, but this is kind of a no win situation for everybody. On one hand, if the room is cool enough you can make some assumptions that all is well. Especially if you aren't pushing the hardware that hard and its mostly idle. But these enterprise cards expect a certain amount of airflow over their heatsinks (Copilot is saying 200-300 CFM for a Connectx-5, I didn't go pull as pec sheet) which often times do not have any form of active cooling, especially the GPU's. Heck in your case sure the actual chip on the card under the nice heatsink gets toasty sure, but you throw long range 100G optics in and those cages (which also have heatsinks) get dang toasty as well but are kind of obfuscated behind the other hot heatsink. Add that to a potentially already dense box full of other hardware creating more preheat and air shadows, etc. and those optics turn into hot potato's when you try to service one. So Dell to ensure they don't destroy any cards, hardware, optics, etc. because they didn't test that specific component defaults to the safe camp. Plus side they do provide the ability to turn off this safety net, so that is nice.

Just my $0.02 that this is likely a scenario where if you poll 10 people you will get 10 different ways on how to do things.

 

[uberguru]

i get what you saying BUT are you saying fans in server should care less what the room temperature is?
depending on your answer there then will better understand why you dont think fans should react on the room temperature, not just what the hell is going on inside

if i have a room setup to suck hot air and push cold air, why the server fans have to think like there is no outside world
blowing at very high rpms is nothing if the heat are not exiting the room
so what is point of making so much noise where in fact the key thing is getting the hot air out of the room, thus the room temperature

and lastly, how did ibm/lenovo blade server with several serves able to stay quiet reacting to room temperature? did they have alien intelligence or alien technology that Dell does not have access to?

 

[oldfett]

Apologies if I ruffled any feathers, I was just pointing out its both a simple and a complicated problem and multiple ways to approach it. I think inlet temperature is just as critical if not more than CFM requirements, but I do strongly feel its not the whole picture.

Anyway hope you get that system happier on the ears.

 

[uberguru]

Apologies if I ruffled any feathers, I was just pointing out its both a simple and a complicated problem and multiple ways to approach it. I think inlet temperature is just as critical if not more than CFM requirements, but I do strongly feel its not the whole picture.

Anyway hope you get that system happier on the ears.

no not at all, no feathers ruffled at all, just a fun chat to me

for example the server room is setup where the cold air if pushed in front of the servers and the the hot air is able to be pulled out from the server room with the mini split

so for me what is more important is what type of air if being pushed not how high high the fans are spinning
i know and understand that air flow inside server is important, but what i dont like is for fans to think they are the only one managing temperature for the server thus ignores what room temperature is

for me i think that is bad engineering and like i said ibm/lenovo blade servers are able to do this easily as i used to have thes ervers befor emoving to dell and now am somewhat annoyed because i never had any issue with noise now dell fans are blowing up the whole server room with noise when room is cold like a freezer