SOLVED ASRock Rack C2750D4I :: CPU temperature issue

Marcet

Contributor
Joined
May 31, 2013
Messages
193
Hi all,

I run a newly build NAS based on ASRock Rack C2750D4I motherboard (detailed configuration in my signature). This motherboard has a passive heatsink and I started to look closely at the temperatures thru the IPMI web dashboard.

The motherboard stays around 38°C with a room temperature of 22°C.

The thresholds for CPU temperature are set to those defaults :
  • Upper Non-Recoverable (UNR) : 90°C
  • Upper Critical (UC) : 85°C
  • Upper Non-Critical (UNC) : 80°C
Observing the live values on the curve shows that the CPU temperature varies from 54°C to 72°C, which I believe is fair for this kind of CPU.

But, the sensor event log shows events in the UNC, UC and UNR zones :
Code:
14    02/18/2016 22:01:35    CPU Temperature    Temperature    Upper Non-Critical - Going High - Deasserted
13    02/18/2016 22:01:26    CPU Temperature    Temperature    Upper Non-Critical - Going High - Asserted
12    02/18/2016 22:01:15    CPU Temperature    Temperature    Upper Non-Critical - Going High - Deasserted
11    02/18/2016 22:01:08    CPU Temperature    Temperature    Upper Non-Critical - Going High - Asserted
10    02/18/2016 20:02:31    CPU Temperature    Temperature    Upper Non-Critical - Going High - Deasserted
 9    02/18/2016 20:02:25    CPU Temperature    Temperature    Upper Critical - Going High - Deasserted
 8    02/18/2016 20:02:24    CPU Temperature    Temperature    Upper Critical - Going High - Asserted
 7    02/18/2016 20:02:23    CPU Temperature    Temperature    Upper Critical - Going High - Deasserted
 6    02/18/2016 20:02:21    CPU Temperature    Temperature    Upper Non-Recoverable - Going High - Deasserted
 5    02/18/2016 19:50:11    CPU Temperature    Temperature    Upper Non-Recoverable - Going High - Asserted
 4    02/18/2016 19:50:10    CPU Temperature    Temperature    Upper Non-Recoverable - Going High - Deasserted
 3    02/18/2016 19:50:08    CPU Temperature    Temperature    Upper Non-Recoverable - Going High - Asserted
 2    02/18/2016 19:50:06    CPU Temperature    Temperature    Upper Critical - Going High - Asserted
 1    02/18/2016 19:50:06    CPU Temperature    Temperature    Upper Non-Critical - Going High - Asserted


Intel specs sheet for this processor shows a max temp before shutdown to 97°C.

Beside this CPU temp concern, all the rest of the system is cool, the HDD temps for exemple are :
Code:
Current Temperature:                    32 Celsius
Power Cycle Min/Max Temperature:     27/33 Celsius
Lifetime    Min/Max Temperature:      2/35 Celsius


Trying to find a solution, I've build a bracket for two 80mm fans. They are blowing toward the passive fan.
The results shown here are after this modification (the number of UNC, UC and UNR observed was way greater before).

Any advice of what I should do ?
 

Yatti420

Wizard
Joined
Aug 12, 2012
Messages
1,437
I'd put in the fans.. See if push or pull is better.. Maybe replace heatsink compound see if that helps.. As for the IPMI thresholds I think those are set by mobo manufacturer.. I wouldn't want to run near max temp (as stated by Intel) forever.... Best you can do I think is just upgrade fans etc if dedicated to case.. Make sure cabling is clean.. No dust etc.. Maybe lower the ambient temperature and lower the rack if possible etc..
 

Marcet

Contributor
Joined
May 31, 2013
Messages
193
I'd put in the fans.. See if push or pull is better..

I've 3D printed this fan holder :

904488case.jpg


But, I believe it's not close enough to the heatsink anyway.
I'm printing a modified one which will be closer.

And as you suggested, I will try le pull configuration also.

Maybe replace heatsink compound see if that helps..
It will be very complicated to deal with the heatsink itself.
There is no mounting options for any other cooler.
And the heatsink in place is "glued" on the CPU.

As for the IPMI thresholds I think those are set by mobo manufacturer.. I wouldn't want to run near max temp (as stated by Intel) forever.... Best you can do I think is just upgrade fans etc if dedicated to case.. Make sure cabling is clean.. No dust etc.. Maybe lower the ambient temperature and lower the rack if possible etc..
I'm a little bit disappointed to see that this motherboard cooling system is not efficient enough.
 

Yatti420

Wizard
Joined
Aug 12, 2012
Messages
1,437
You have a printer.. Another option I see which you could do there is a vent and pull the hot air out the vent and case.. Just vent that heat out the side of the case there where the psu vents.. It'll eat up more room but use a single fan.. Just one more on the fan grill there with plastic vent guide (the spacing between mobo and vent dont want to be to far way so you can pull clean air in around the heatsink and then out the case)....

How busy is the CPU? If it's thermal adhesive can be a pain to change..
 
Last edited:

Marcet

Contributor
Joined
May 31, 2013
Messages
193
You have a printer.. Another option I see which you could do there is a vent and pull the hot air out the vent and case.. Just vent that heat out the side of the case there where the psu vents.. It'll eat up more room but use a single fan.. Just one more on the fan grill there with plastic vent guide (the spacing between mobo and vent dont want to be to far way so you can pull clean air in around the heatsink and then out the case)....
I'll do some tests tomorrow.

How busy is the CPU? If it's thermal adhesive can be a pain to change..
I'm doing some transferts from my old NAS and I run an Emby jail.
 

Fuganater

Patron
Joined
Sep 28, 2015
Messages
477
IMO, that heatsink is supposed to have cool air pushed across it (with the fins obv), not directly down on to it.
 

MrToddsFriends

Documentation Browser
Joined
Jan 12, 2015
Messages
1,338
When running without your DIY fan holder: How long does it take until the CPU temperature reaches a value of 80°C or above under minimal CPU load? If this period is at least in the range of minutes the passive heatsink should reach some sort of equilibrium state, being rather hot, too (above the heat-pain threshold). If the heatsink doesn't get hot while the CPU does, I would suspect a heat transfer problem, or, in other words a manufacturing flaw.

BTW: I mounted a 60mm FAN (Noctua NF-A6x25 FLX running at 1500rpm) onto the heatsink of my A1SAi-2750F using cable ties looped through between heatsink and board. Under low CPU load the temperature readings are slightly above room temperature (CPU) and about 10 degrees above that (System, DIMMs). In this setup the CPU temperature never got substantially above 30°C even when stress testing with prime95 on all cores (room temperature roughly at 20°C, not tested during summer).
 

Marcet

Contributor
Joined
May 31, 2013
Messages
193
IMO, that heatsink is supposed to have cool air pushed across it (with the fins obv), not directly down on to it.
Like in a rack ? So you suggest that a tower is not adapted to that kind of motherboard ?
 

Marcet

Contributor
Joined
May 31, 2013
Messages
193
When running without your DIY fan holder: How long does it take until the CPU temperature reaches a value of 80°C or above under minimal CPU load? If this period is at least in the range of minutes the passive heatsink should reach some sort of equilibrium state, being rather hot, too (above the heat-pain threshold). If the heatsink doesn't get hot while the CPU does, I would suspect a heat transfer problem, or, in other words a manufacturing flaw.
I'll check that as soon as my data copy ends.

BTW: I mounted a 60mm FAN (Noctua NF-A6x25 FLX running at 1500rpm) onto the heatsink of my A1SAi-2750F using cable ties looped through between heatsink and board. Under low CPU load the temperature readings are slightly above room temperature (CPU) and about 10 degrees above that (System, DIMMs). In this setup the CPU temperature never got substantially above 30°C even when stress testing with prime95 on all cores (room temperature roughly at 20°C, not tested during summer).
30°C wow, I'm impressed considering the values I have. I hope I can find a way to attach some fan on the heatsink.
 

Marcet

Contributor
Joined
May 31, 2013
Messages
193
Some updates :
With my modified fan holder (15mm lower), the fans are a little more efficient : Gain 3°C
So the idle temperature is 50°C, which is still high, I believe.

Next step, I'll try a pull configuration and also a right push and left pull.
 

Fuganater

Patron
Joined
Sep 28, 2015
Messages
477
Granted I am using a Supermicro 4U Active heatsink and my servers are in Supermicro SC846 chassis but here are my temps. It would be nice for others using passive heatsinks to post how they have their fans setup.

Code:
dev.cpu.0.temperature: 39.0C
dev.cpu.1.temperature: 39.0C
dev.cpu.2.temperature: 36.0C
dev.cpu.3.temperature: 36.0C
dev.cpu.4.temperature: 37.0C
dev.cpu.5.temperature: 37.0C
dev.cpu.6.temperature: 35.0C
dev.cpu.7.temperature: 35.0C


After a wee bit of Googling I found an Amazon review of the board.

However, there are a few issues to keep in mind (none of which I docked a star for):
1.) This board is meant for rackmounted servers, NOT passive home cases. It states right on the product page on ASRock's site that the heatsink needs a decent amount of airflow over it. My solution for this was to buy a 50mm Gelid Silent 5 fan and attach it to the heatsink with some professional-grade double-sided tape cut to fit.
Additionally, ASRock only included a thermal pad in between the heatsink and CPU. I purchased a pack of 15mm x 15mm x 1mm copper risers, put some Arctic Silver on either side, and put it where the thermal pad used to be.
After doing the above two tweaks, my CPU temperatures have dropped from from the 50-60C idle they were originally, to a much cooler ~35 C.

You may want to look in to "better" aka more powerful fans that what you are currently using.
 

Marcet

Contributor
Joined
May 31, 2013
Messages
193
Temperature report per build would be a very good idea.
Making the hardware choice is not always an easy task :D

I wish I had your temperatures. I'm used to liquid cooling on my main desktop which is overclocked and I have 36°C under normal operation.

Interesting Amazon review. I've work to do to improve my build.

You may want to look in to "better" aka more powerful fans that what you are currently using.
With the v2 of my fan mount and having the fans in pull mode, I have an idle temp of 50°C and a much better 65°C max temp on load.
 

Dall

Dabbler
Joined
Feb 21, 2016
Messages
13
Stumbled across this thread while searching for solutions to my same problem. At the outset, I'll note that I'm not running FreeNAS. Instead, I have Ubuntu running ZFS. (I have my server run a couple python scripts nightly as well as Sickbeard, CouchPotato, Bit torrent, etc -- that's why I run full on Ubuntu versus FreeNAS, in case anyone wonders).

But I am running the same hardware; just built it this week. I have a c2750d41 in a Fractal Node 304, and I too am seeing very high CPU temps. Right now, at idle, in our 74F degree home that CPU temp is 60c (according to "sensors" in Ubuntu). Last night I was testing the capabilities of the C2750 for transcoding multiple Plex streams.* Under a pretty substantial load, the CPU cores were hanging out at 90c! At least one problem is that the fans for the Fractal Node 304 blow air across the HDD cages (which are 3-4 inches above the motherboard and heat sink) back to the exhaust fan, which is also slightly lifted. As such, very little air flow hits the heatsink. I read one review where a C2750D4I user added a CPU fan to the heatsink and brought down idle temps to approximately 35c. I'll either do that or scrap the case for something else (I'm thinking U-NAS NSC-800).

Hope this is helpful.

*As an aside, it performed admirably, playing 4 streams of 720 or 1080p video with 3 of the streams requiring video transcoding. There was some initializing buffering, but smooth play after that.
 

Marcet

Contributor
Joined
May 31, 2013
Messages
193
I agree, the board performed admirably. It's only flaw is this high temps.

I would not go for U-NAS NSC-800 as it seems very compact and has not a standard PSU.
I'm sure I will find a durable solution for this heat problem.
Lowering my DIY fan holder has solve a part of the problem, now I just need to improve it a bit.

I'm not sure I can achieve 35°C. But I'll try to be below 50°C at Idle.
 

brando56894

Wizard
Joined
Feb 15, 2014
Messages
1,537
Like in a rack ? So you suggest that a tower is not adapted to that kind of motherboard ?
Correct, this is a motherboard meant to go in a rack. I have the same motherboard and bought an aftermarket fan for it (60mm IIRC) and superglued it to the heatsink and it works well. I'm currently around 116-120F.

Sent from my Nexus 6 using Tapatalk
 
Last edited:

Dall

Dabbler
Joined
Feb 21, 2016
Messages
13
I agree, the board performed admirably. It's only flaw is this high temps.

I would not go for U-NAS NSC-800 as it seems very compact and has not a standard PSU.
I'm sure I will find a durable solution for this heat problem.
Lowering my DIY fan holder has solve a part of the problem, now I just need to improve it a bit.

Appreciate the heads up on NSC-800. Looking at the layout for the NSC-800, I'm going to have the same issues anyways with regards to CPU temp. As it is, I'm only using 5 disks, and one is an SSD. I can migrate the SSD off to the side somewhere and have the ability to expand my Zpool with another mirrored VDev. Point being, I'm not really limited by a case with capacity for 6 disks versus 8.

I'm thinking I'll put a fan on top of the heat sink and be done with it. I also need to do cable management now that the build is completed, and that too may help with speed.
 

Marcet

Contributor
Joined
May 31, 2013
Messages
193
Correct, this is a motherboard meant to go in a rack. I have the same motherboard and bought an aftermarket fan for it (60mm IIRC) and superglued it to the heatsink and it works well. I'm currently around 116-120F.
Very interresting. Which kind of fan are you using (speed) ?
The temps you gave, is it idle or working on normal load ?
 

Marcet

Contributor
Joined
May 31, 2013
Messages
193
I'm thinking I'll put a fan on top of the heat sink and be done with it. I also need to do cable management now that the build is completed, and that too may help with speed.
Cable management is mandatory as it can lower air flow and it's getting worst if it catches dust.
 

MrToddsFriends

Documentation Browser
Joined
Jan 12, 2015
Messages
1,338
Meh, take a Noctua NF-A6x25 FLX and attach it to your heat sink however you like. Together with the provided Ultra-Low-Noise Adapter it runs at 1500rpm it is very silent (every single WD Red produces substantially more noise) and cools an Avoton very effectively. Should work in most cases (there is not much airflow over the CPU heat sink in my Node 304).
http://noctua.at/en/products/fan/nf-a6x25-flx.html

Or take the PWM version of that product if that fits your needs better (didn't exist as I bought the FLX).
http://noctua.at/en/products/fan/nf-a6x25-pwm.html

The cable tie fixing method I mentioned earlier in this thread might not be for everyone as this might induce dis- and reassembly of the server in case of a CPU fan failure.
 

Marcet

Contributor
Joined
May 31, 2013
Messages
193
Meh, take a Noctua NF-A6x25 FLX and attach it to your heat sink however you like. Together with the provided Ultra-Low-Noise Adapter it runs at 1500rpm it is very silent (every single WD Red produces substantially more noise) and cools an Avoton very effectively. Should work in most cases (there is not much airflow over the CPU heat sink in my Node 304).
http://noctua.at/en/products/fan/nf-a6x25-flx.html
Interesting. Can you tell me more about it ?
Which way (push or pull) ?
What cpu temps are you getting ?

I'm still wondering which temp range I must try to achieve.
The cable tie fixing method I mentioned earlier in this thread might not be for everyone as this might induce dis- and reassembly of the server in case of a CPU fan failure.
Sure.
 
Top