Mellanox SX6036 fan mod

[ilikerackmounts]

I've been searching these forums for a definitive thread on swapping these fans out but I can only find bits and pieces where people have tried things but with only varying degrees of success. The SX6036 is just a tiny bit above what I consider tolerable noise. I haven't taken the fan drawer out to have a look but, what sort of replacement fans should I buy? I found some evidence that people were able to use Sunon Maglev fans - are these good for the job? I see a guy on ebay selling a lot of 4 for a similar mellanox switch but it's not mine. What sort of pin compatibility should I be looking for? I can canabalize a fan's molex connector if need be, but it'd be nice to preserve the original fans + connectors in case something goes wrong.

From my switch's documentation it appears as though the fans are 48v, all the fans everybody are linking to are 12v fans. Clearly there's some sort of PWM or level adjustment with the fans because at first boot they are full blast and later they spin down some. Can I buy standard 4 pin PWM fans, or do I buy normal 3 pins? It looks like people were initially happy with noctuas but later found an issue where they were spinning all the way up and back down again (this is something that frequently happens with server mainboards because the noctua fans hit a lower RPM than they expect and they just assume the fans are dead).

As far as I can tell, these are molex picoblade connectors? According to this, anyway: https://forums.servethehome.com/ind...d-sx6018-switch-emc-only-300-obo.12792/page-2

If so, I can order these: https://www.molex.com/molex/products/part-detail/cable_assemblies/0151340400 and splice and solder the wires. I just need to know what will likely work. My "CPU_BOARD_MONITOR2" temp seems to be the highest with stock fans and it's at high 50Cs. Could this be what's driving the crazy high RPMs on the existing fans? What exactly inside the switch is the CPU_BOARD_MONITOR2? It kind of sounds like it's the embedded PPC CPU. Is the heatsink / thermal paste on it just kind of garbage or did they do this brand new?

Edit: ahh regarding the 48v thing, I was looking at the generic mellanox switch OS product documentation found on the web interface. I think these are in fact 12v fans.

 

[Tyrant82]

Hi,

I just have bought 6x 3pin Noctua NF-A4x20 FLX Fans that should have the same wiring layout as mentioned in this thread:
https://forums.servethehome.com/ind...018-switch-emc-only-300-obo.12792/post-238079

I also would love to switch the connectors from the Noctua fans to the ones needed by the switch, but i havent found the correct form or norm or size how those are named so i am still looking.

What would be the best option to switch the fan connectors? are those plugged into the picoblade connectors or pressed?
Should we use the ones from the original fans if they are easy to dismount?

Would be amazing if there was a 3pin to picoblade connector adapter so you just plug them in

 

[ilikerackmounts]

I mean adapting the connectors isn't really the challenge. The real issue is that those noctua fans have pretty poor airflow, if you compare the datasheets they just aren't up to snuff. What seems to be the most urgent component that needs to be cooled is, sadly, the embedded powerpc processor used for management and serving the web pages. For some reason that thing idles in the upper 50c's. Has anyone opened up one of these switches and have photos of this thing? I'm curious what the heatsink looks like on this thing and how the airflow is configured. The load average on the CPU is modest (like .33), it should not be this hot.

I'm wondering if it has a heatsink at all and if so, if it might benefit from an application of a decent thermal paste instead of maybe the thermal pad it could be using. The linked thread above shows somebody using those Noctua fans and the CPU temp rising up into the upper 70 C which for me seems a little too close for comfort. That's likely with near 0 load.

@Freebsd1976 You seemed to have the most luck in that thread, what are your thoughts? I looked at some fans with a similar profile with a few decibels lower rating and lower max RPM (which should mean they might PWM down to a lower speed on average), but I couldn't definitely say. At the very least it might be nice to know the pinout to follow for these fans in relation to a normal 4 pin PWM header.

The fan that looked the best I cannot buy (MOQ 1000):

DC Cooling Fan with 40mm x 40mm x 28mm Series - User is considering micro cooling fans, traditional DC cooling fans, and 5V DC cooling fans. | Made in Taiwan Custom RV Fans and PC Cooling Fans Manufacturer | TITAN Technology Limited

TITAN is a Taiwan high-performance DC Cooling Fan with 40mm x 40mm x 28mm Series - User is considering micro cooling fans, traditional DC cooling fans, and 5V DC cooling fans., computer cooler, cooling fan, mobile cooling, CPU cooler, DC fan, PC cooling kits, laptop cooling pads, power inverter...

www.titan-cd.com

TFD-4028HH12X in that table. That has a significantly lower dBA rating with not too much worse airflow and pressure.

 

[Freebsd1976]

I mean adapting the connectors isn't really the challenge. The real issue is that those noctua fans have pretty poor airflow, if you compare the datasheets they just aren't up to snuff. What seems to be the most urgent component that needs to be cooled is, sadly, the embedded powerpc processor used for management and serving the web pages. For some reason that thing idles in the upper 50c's. Has anyone opened up one of these switches and have photos of this thing? I'm curious what the heatsink looks like on this thing and how the airflow is configured. The load average on the CPU is modest (like .33), it should not be this hot.

I'm wondering if it has a heatsink at all and if so, if it might benefit from an application of a decent thermal paste instead of maybe the thermal pad it could be using. The linked thread above shows somebody using those Noctua fans and the CPU temp rising up into the upper 70 C which for me seems a little too close for comfort. That's likely with near 0 load.

@Freebsd1976 You seemed to have the most luck in that thread, what are your thoughts? I looked at some fans with a similar profile with a few decibels lower rating and lower max RPM (which should mean they might PWM down to a lower speed on average), but I couldn't definitely say. At the very least it might be nice to know the pinout to follow for these fans in relation to a normal 4 pin PWM header.

The fan that looked the best I cannot buy (MOQ 1000):

DC Cooling Fan with 40mm x 40mm x 28mm Series - User is considering micro cooling fans, traditional DC cooling fans, and 5V DC cooling fans. | Made in Taiwan Custom RV Fans and PC Cooling Fans Manufacturer | TITAN Technology Limited

TITAN is a Taiwan high-performance DC Cooling Fan with 40mm x 40mm x 28mm Series - User is considering micro cooling fans, traditional DC cooling fans, and 5V DC cooling fans., computer cooler, cooling fan, mobile cooling, CPU cooler, DC fan, PC cooling kits, laptop cooling pads, power inverter...

www.titan-cd.com

TFD-4028HH12X in that table. That has a significantly lower dBA rating with not too much worse airflow and pressure.

I use 8100 rpm (4020 delta fan) on sx6012 , and set mini fan speed 70% （the os need fan at least 4500 rpm） , no noise , and wire order need carefully changed , otherwise it will fried the fan connector！！！

 

[ilikerackmounts]

Ah cool, can you specify the pinout, maybe? I know it's a bit much to ask you to open it up if you didn't have photos. How are you controlling the speed? Via a resistor or are you interposing something else that's messing with the PWM? I'd like to maintain some PWM function but if the noise is very much more tolerable at a fixed RPM like that, I might consider it assuming that the airflow can cope with a typical load.

It doesn't seem like the switching ASIC is getting terribly hot - it just mainly seems to be the CPU.

Also, "4020 delta fan" actually just describes dimension. Can you specify which model of fan? The ones I'm looking at that fit that description at 12v have kind of poor airflow ratings and not terribly great static pressure. What is your CPU's temperature on that switch?

 

[Freebsd1976]

IMHO , you don't need replace fan, use fae to debug and set fan speed is the best way

fae mlxi2c set_fan /FAN1/FAN 1 30
fae mlxi2c set_fan /PS1/FAN 1 30
fae mlxi2c set_fan /PS2/FAN 1 30

this will set speed to 30 or what ever you want , if want a better way you can try the following way (it should work , I did not try it since i replace delta fan), please also search and read the sx6012 thread , there lots information in there!

You are right. Reviewing the log showed that it was just seconds too early. tc overwrote it. Here is the script that finally worked (/etc/rc.d/rc.local, or /mnt/root2/etc/rc.d/rc.local from rootdev mtdblock6

#!/bin/sh

touch /var/lock/subsys/local

echo "Waiting for clusterd"
COUNT=1
while true
do
    CMD=$(ps aux|grep "/[o]pt/tms/bin/clusterd")
    if [ $? -ne 0 ]; then
        sleep 30
        let COUNT=${COUNT}+1
        if [ $COUNT -gt 10 ]; then
            echo "time out, aborting"
            break
        fi
    else
        sleep 60
        echo "Changing fan speed"
        /opt/tms/bin/mdreq action /system/chassis/actions/set-fan-speed fan_module string "/MGMT/FAN1" fan_number int8 1 fan_speed int8 27 set_max uint8 50
        /opt/tms/bin/mdreq action /system/chassis/actions/set-fan-speed fan_module string "/MGMT/FAN2" fan_number int8 1 fan_speed int8 27 set_max uint8 50
        /opt/tms/bin/mdreq action /system/chassis/actions/set-fan-speed fan_module string "/MGMT/FAN3" fan_number int8 1 fan_speed int8 27 set_max uint8 50
        /opt/tms/bin/mdreq action /system/chassis/actions/set-fan-speed fan_module string "/MGMT/FAN4" fan_number int8 1 fan_speed int8 27 set_max uint8 50
        break
     fi
done

I am waiting for clusterd instead of hwd because that is one of the last daemons launched by pm.
Thanks for your help!

 

[ilikerackmounts]

IMHO , you don't need replace fan, use fae to debug and set fan speed is the best way

fae mlxi2c set_fan /FAN1/FAN 1 30
fae mlxi2c set_fan /PS1/FAN 1 30
fae mlxi2c set_fan /PS2/FAN 1 30

this will set speed to 30 or what ever you want , if want a better way you can try the following way (it should work , I did not try it since i replace delta fan), please also search and read the sx6012 thread , there lots information in there!

ibswitch [standalone: unknown] (config) # fae mlxi2c set_fan /FAN1/FAN 1 30
Jan 31 22:52:54 INFO LOG: Initializing SX log with STDOUT as output file.
-E- mlxi2c: Failed to locate device /FAN1/FAN : MLXI2C_NOT_EXIST_DEV (18)

Hmm that command did not work. I tried a number of other I2C paths that seems to be closer to my device but none of them worked:

ibswitch [standalone: unknown] (config) # sae mlxi2c set_fan /FAN/F1 1 30
% Unrecognized command "sae".
Type "?" for help.
ibswitch [standalone: unknown] (config) # fae mlxi2c set_Fan /FAN/F1 1 30
Jan 31 22:58:45 INFO    LOG: Initializing SX log with STDOUT as output file.
-E- mlxi2c: Unknown command: set_Fan
ibswitch [standalone: unknown] (config) # fae mlxi2c set_fan /FAN/F1 1 30
Jan 31 22:58:56 INFO    LOG: Initializing SX log with STDOUT as output file.
-E- mlxi2c: Failed to locate device /FAN/F1 : MLXI2C_NOT_EXIST_DEV (18)
ibswitch [standalone: unknown] (config) # * mlxi2c set_fan /FAN/F1 1 30
% Unrecognized command "*".
Type "?" for help.
ibswitch [standalone: unknown] (config) # fae  mlxi2c set_fan /FAN/F1 1 30
Jan 31 22:59:12 INFO    LOG: Initializing SX log with STDOUT as output file.
-E- mlxi2c: Failed to locate device /FAN/F1 : MLXI2C_NOT_EXIST_DEV (18)
ibswitch [standalone: unknown] (config) # * mlxi2c set_Fan /FANF1 1 30
% Unrecognized command "*".
Type "?" for help.
ibswitch [standalone: unknown] (config) #
ibswitch [standalone: unknown] (config) # fae mlxi2c set_Fan /FAN\\F1 1 30
Jan 31 22:59:38 INFO    LOG: Initializing SX log with STDOUT as output file.
-E- mlxi2c: Unknown command: set_Fan
ibswitch [standalone: unknown] (config) # fae mlxi2c set_fan /FAN\F1 1 30
Jan 31 22:59:48 INFO    LOG: Initializing SX log with STDOUT as output file.
-E- mlxi2c: Failed to locate device /FAN\F1 : MLXI2C_NOT_EXIST_DEV (18)
ibswitch [standalone: unknown] (config) # * mlxi2c set_fan FAN\F1 1 30
% Unrecognized command "*".
Type "?" for help.
ibswitch [standalone: unknown] (config) # fae mlxi2c set_fan FAN\F1 1 30
Jan 31 22:59:58 INFO    LOG: Initializing SX log with STDOUT as output file.
-E- mlxi2c: Failed to locate device FAN\F1 : MLXI2C_NOT_EXIST_DEV (18)
ibswitch [standalone: unknown] (config) # fae mlxi2c set_fan FAN\\F1 1 30
Jan 31 23:00:06 INFO    LOG: Initializing SX log with STDOUT as output file.
-E- mlxi2c: Failed to locate device FAN\F1 : MLXI2C_NOT_EXIST_DEV (18)
ibswitch [standalone: unknown] (config) # show monitor
% Incomplete command.
Type "show monitor ?" for help.
ibswitch [standalone: unknown] (config) # show monitor ?
session                        Shows monitor session configuration and status
ibswitch [standalone: unknown] (config) # show monitor session
% Incomplete command.
Type "show monitor session ?" for help.
ibswitch [standalone: unknown] (config) # show monitor session ?
session id <1-7>               
summary                        Shows monitor session configuration and status summary
ibswitch [standalone: unknown] (config) # show monitor session summary
Flags : i ingress, e egress, b both
Session  Admin   Status  Mode       Destination  Source
ibswitch [standalone: unknown] (config) # show power
======================================
Module         Device           Status
======================================
PS1            PS               FAIL
PS2            PS               OK
ibswitch [standalone: unknown] (config) # show inventory
=================================================================================================
Module           Type             Part number      Serial Number    Asic revision    HW Revision
=================================================================================================
CHASSIS          SX6036           MSX6036F-1SFS    MT1446K00591     N/A              A7
MGMT             SX6036           MSX6036F-1SFS    MT1446K00591     2                A7
FAN              SXX0XX_FAN       MSX60-FF         MT1445K02193     N/A              A1
PS1              SXX0XX_PS        MSX60-PF         MT1445K01765     N/A              A1
PS2              SXX0XX_PS        MSX60-PF         MT1449X06206     N/A              A1
CPU              CPU              SA002203         MT1443K02365     N/A              A2
ibswitch [standalone: unknown] (config) # fae mlxi2c set_fan MGMT/FAN/FAN1 1 30

 

[Freebsd1976]

try
set_fan /FAN/FAN 1 65
set_fan /PS1/FAN 1 65

 

[ilikerackmounts]

Cool, that did work. So an important question - is it permanent? Or at least, is there a way to adjust a fan curve in case the temperature does rise it will spin up as necessary?

 

[Freebsd1976]

Cool, that did work. So an important question - is it permanent? Or at least, is there a way to adjust a fan curve in case the temperature does rise it will spin up as necessary?

no ， it not permanent ，it will automatically spin up if needed

carefully read my post ， the quoted contents is how make it permanent

 

[ilikerackmounts]

I did but I think I'm still missing some context. It looks as though the user modified an init script on the OS's firmware image itself. It's unclear how to do that exactly - is there an editor accessible from the shell? Is that what the "fae" command is actually doing (as it, invoking shell commands directly rather than using their custom CLI)? Do I have any hope of invoking nano or vi?

And how exactly does the PWM work with that modification versus the i2c command to set the speed? Since setting the speeds I've not seen the fan RPMs change if much at all. Am I modifying the average value or am I capping the maximum value? Also, given that there's no temperature monitor on the power supplies, how low can I afford to set them? Will those still spin up to higher speeds if they get too hot?

It looks like in that init script it's a minimum and maximum being set. So, is the range then just adjusted proportionally based on the new min and max or does it behave more like a clipping (where, it's going to hang out closer to the max to maintain a temperature)?

 

[bryan_v]

Two quick questions:

• What's the approx CFM you're aiming for the fans?
• They need to be PWM fans right?

BTW, if anyone has any recommendations would be super appreciated.

Thinking of getting these fans, but not sure if they're over or under powered:
Delta - Fan Tubeaxial 12VDC Square - 40mm L x 40mm H Ball 13.0 CFM (0.364m³/min) 4 Wire Leads 36.5dB(A)
AFB0412VHB-TP29 Delta Electronics | Fans, Thermal Management | DigiKey

 

[unphased]

Well I should have done some more sanity checking. Here's what happened

• Got my SX6036. Great switch. Got it going in ethernet mode.
• Want to quiet it down a bit. My workstation it's next to is already kind of loud probably 42dB but this thing is probably around 50. If I can bring it closer to 43 or something I'll be satisfied.
• I can clearly see when removing the power supplies and fan module that the fans installed in the switch are all Delta FFB0412SHN's.
• According to the webui, the main fans run at 9300 RPM and the PS fans were going at 11150 RPM. They def do spin down a bit compared to during bootup. Temp monitor shows CPU at 60C which is concerning but rest of temps are perfectly good at 30C or under.
• I bought some $6 FFB0412VHN delta fans from eBay. Based on the research that I did, it *should* be half as powerful and half as loud as the FFB0412SHNs. I don't have enough experience to tell if these things are counterfeit or not. But I can say they do work pretty well and look to have 4 functioning wires because I could hook it up to my test rig and control its fan speed precisely via PWM. Interestingly even setting the 0-100 PWM setting to 0 sees it spinning still at a good clip (I'd estimate 1500rpm), but nice and quiet. I am under the impression that for the mellanox switches there exists some hardware control that prevents fans from spinning down below 4000 or so RPM. That would be perfectly acceptable as 40mm fans have acceptable acoustic properties once their RPM dips below 8~9000 RPM IMO. Now these fan impellers on my new fans have an interesting shape for sure. They stick out like scoops. All I can say is the build quality seems solid and these fans even if counterfeit somehow, seem to offer solid value.
• Since I already had my soldering gear out for a separate thing I was doing, I went and snipped the headers and did a fan replacement on one of the power supplies. Very easy to do. I figured it's safe to assume the color coding of the fan wires is consistent, as it's always red, black, yellow, blue, where yellow is presumably the RPM tach and the blue the PWM signal.
• Plug power supply back in and guess what, even more noise. Check in webui and the PS2 fan speed is 14500 RPM
• I have the removed FFB0412SHN. So I tested both with my shitty bench power supply. At 12V the old fan drew around .20 amps. It sounds more high pitched, whinier, but also less powerful than the new FFB0412VHN fans I got which at 12V draw around .26 amps.

The conclusion I can draw from independent testing of the fans is that the new fans I ordered are significantly more higher powered than the fans that came with the switch. I do not know if there is PWM active and if that somehow got disabled or changed with a fan swap but for sure the fan swap results in more noise and airflow, which is perhaps not what I was aiming for here. I do not know if these original FFB0412SHN fans are mislabeled or maybe they really did degrade to the point where they went from drawing .45 (or their .6 rating, the label is hard to interpret) down to .20. They don't sound like they're on their last legs or anything, but they also clearly do not sound full of vitality, that is for sure.

So where to go from here?

Well, I'm trying not to sink many more evenings into such a narrow endeavor as tweaking fans on this old switch. However... it's already been established as a rewarding hobby for me. so I am willing to contemplate a few strategies for moving forward with this one.

1. First priority would be to see about the cooling solution on the PPC CPU chip. If it is already reaching 60C I would not even want to halve the airflow for its sake. First would be to search for a bit to see if I can find any pics from others on the innards so I know what I'm getting into. I would hope that this chip is cooled by something other than a heatsink fixed to it with thermal epoxy so that I can upgrade its heatsink. I would also hope that if this were the case, that it wouldn't have degraded as far as to be 60C at idle. Redoing thermal paste in old enterprise hardware is a no brainer and I do it on all these NICs too. Even if I cannot do anything about the heatsink I am sure I can jury rig one more 40mm fan blowing directly onto it in close proximity if it came down to that. Tapping 12V from somewhere and using a noctua low noise adapter may work though I'll want to test it a bit to ensure it wont melt down.
2. If I can come up with something to lower the CPU temp a lot more then the limiting factor will probably be the transceivers. Probably the only ones that may get too hot are the LR4 transceivers for which we can only install up to 4 of them on these switches anyway. I think they will be fine. I can probably try to do some experiments to see how hot they could possibly get by figuring out how to fix the fan speed to something really slow and quiet circa 4krpm and if all temps are still ok at that point it would become feasible to do an all-Noctua fan replacement. However I'm leaning towards it not being worthwhile to spend $20*6=$120 on 40mm fans for this. So I will look into getting reasonably quiet deltas or sunons or whatever, and locking the fans at a low speed.
3. Look into even lower speed fans. These VHN fans are definitely too loud at full blast. There is one large question mark here, which is why are these 9000 RPM rated fans running at 14500 RPM. I confirmed with the power supply that it is drawing the right amount of power. There are the HN and HHN and MN models which should be progressively slower and quieter. I think there are also many Sunon options. There is also the 5-packs of the 6krpm Arctic 40x28mm fans. So, the benefit of swapping for fans like this is it will allow me to guarantee low noise rather than relying on possibly unreliable/dangerous scripting and commanding. I'm reasonably confident that performing physical fan swap will consume less time for me than to figure out the ways to command them in software to my satisfaction but I may be wrong about this.

Continued here: #49

 

[erock]

I am playing around with fan speed and temperatures (thank you for getting me here!) using the following workflow in the console interface:

1. enable
2. config terminal
3. fae mlxi2c set-fan /FAN/FAN 1 <percent>
4. fae mlxi2c set_fan /PS1/FAN 1 <percent>
5. fae mlxi2c set_fan /PS2/FAN 1 <percent>

The results below indicate that we can achieve tolerable noise levels at idle (no data transfer occurring) while maintaining CPU Board Monitor (CPU_BM) temperatures less than 64-65C. Also, PS2 FAN speed seems to have the biggest effect on CPU_BM2 and noise level. This link indicates that max operating temp of CPU is 85C. I need to check temps under data transfer load but I think this is a useful reference point.

I would like to run these fans at low speed while idle and I am in my office developing code and then manually ramp them up while running large parallel computations.

So at what temperatures should I be concerned?

Initial state with default config (noise level was probably above 50db and very distracting):
SX : 24C
QSFP_T1 : 25C
QSFP_T2 : 25C
QSFP_T3 : 25C
BM : 29C
CPU_BM : 34C
CPU_BM2 : 58C
FAN1: 9360rpm
FAN2: 9060rpm
FAN3: 9060rpm
FAN4: 9060rpm
PS1: 11,160rpm
PS2: 11,160rpm

Then I ran...
fae mlxi2c set-fan /FAN/FAN 1 30
fae mlxi2c set_fan /PS1/FAN 1 30
fae mlxi2c set_fan /PS1/FAN 1 30

Final state after 10 minutes (nice relaxing hum):
SX : 25C
QSFP_T1 : 25C
QSFP_T2 : 26C
QSFP_T3 : 26C
BM : 30C
CPU_BM : 35C
CPU_BM2 : 64C
FAN1: 5790rpm
FAN2: 5790rpm
FAN3: 5670rpm
FAN4: 5670rpm
PS1: 5790rpm
PS2: 5670rpm

Then I ran...
fae mlxi2c set_fan /PS1/FAN 1 50
fae mlxi2c set_fan /PS1/FAN 1 50

Final state after 10 minutes (louder but tolerable hum):
SX : 25C
QSFP_T1 : 25C
QSFP_T2 : 26C
QSFP_T3 : 26.5C
BM : 31C
CPU_BM : 36C
CPU_BM2 : 62C
FAN1: 5910rpm
FAN2: 5910rpm
FAN3: 5790rpm
FAN4: 5790rpm
PS1: 8280rpm
PS2: 8040rpm

 

[unphased]

nice! I opted for hardware mod first before testing any fan speed commands but it will be really easy to implement automations around fan speed since the management system inside is running linux.

As you can see in my other thread with pictures, the stock heatsink on that CPU is very tiny and even after replacing it with an absolute chonker of a full copper heatsink, I only got a 10C improvement... I think there is a huge amount of thermal headroom on everything in these and even though I am very new to this fancy switches stuff, my understanding is that if you put a large network load on the switch it is not likely to proportionately increase CPU load on the management computer. It may increase heat output from the switch ASIC. But we can see the switch asic is cooled extremely well even under greatly reduced airflow. Definitely continue to monitor temperatures, but I'd worry more about the transceivers, in particular any high power ones, before the CPU or the ASIC. The main barrier is clearly the 4krpm fan RPM limit in some firmware. The most practical approach to go below 4krpm may be to spoof these fan rpm signals, but it seems unnecessary as the noise at 6krpm is already nearly acceptable to me so 4krpm should be very acceptable.

 

[Tim111]

Does anyone know the default number for the fan1? I was trying different percentage and at 50 my fan is surging and says rpm is lower then threshold of like 5700 rpm. At 65 it is quieter and temps seem ok. At 80 it is close to original. Between 9400 and 9800 rpm anything 90 or above is like a jet.

 

[erock]

Does anyone know the default number for the fan1? I was trying different percentage and at 50 my fan is surging and says rpm is lower then threshold of like 5700 rpm. At 65 it is quieter and temps seem ok. At 80 it is close to original. Between 9400 and 9800 rpm anything 90 or above is like a jet.

As described above you can use the following to set fan speeds to a desired percent:

1. enable
2. config terminal
3. fae mlxi2c set-fan /FAN/FAN 1 <percent>
4. fae mlxi2c set_fan /PS1/FAN 1 <percent>
5. fae mlxi2c set_fan /PS2/FAN 1 <percent>

 

[jmsq]

Does anyone know the default number for the fan1? I was trying different percentage and at 50 my fan is surging and says rpm is lower then threshold of like 5700 rpm. At 65 it is quieter and temps seem ok. At 80 it is close to original. Between 9400 and 9800 rpm anything 90 or above is like a jet.

So, this may require a bit of experimenting depending on the particular switch and fan variants that are installed in it. I have one SX6012 and one SX6036 and these are the lowest values that were stable for me at idle:

enable
conf t
show inventory
sh fan
fae mlxi2c set_fan /MGMT/FAN1 1 28 # for sx6012 only
fae mlxi2c set_fan /FAN/FAN 1 38 # for sx6036 only
fae mlxi2c set_fan /PS1/FAN 1 38 # for sx6036 only
fae mlxi2c set_fan /PS2/FAN 1 38 # for sx6036 only
sh fan

 

[Tim111]

The one I tried it on is an HP sx6036 flashed to latest firmware fully licensed. But can't get 50 or lower without fans going up and down and rpms are in the 3500 range. I will try on one of my other sx6036 and see.

 

[erock]

So, this may require a bit of experimenting depending on the particular switch and fan variants that are installed in it. I have one SX6012 and one SX6036 and these are the lowest values that were stable for me at idle:

enable
conf t
show inventory
sh fan
fae mlxi2c set_fan /MGMT/FAN1 1 28 # for sx6012 only
fae mlxi2c set_fan /FAN/FAN 1 38 # for sx6036 only
fae mlxi2c set_fan /PS1/FAN 1 38 # for sx6036 only
fae mlxi2c set_fan /PS2/FAN 1 38 # for sx6036 only
sh fan

I can get down to 20% for all fans on my SX6036 while avoiding instability. RPM's are around 4890 and temperatures are about 5 degrees C higher than at 40%. If I go below 20% then the fans oscillate between high and low speeds.