Taming the C6100

PigLover

Moderator
Jan 26, 2011
3,071
1,375
113
Pot = potentiometer = variable resister = what that fan controller you showed in your picture actually is...
 

Dragon

Banned
Feb 12, 2013
77
0
0
Pot = potentiometer = variable resister = what that fan controller you showed in your picture actually is...
LOL I see, I was beginning to wonder why a certain vegetation could be involved in this...

Is it possible to control 4 fans from 1 pot? It is messy to install 4 controls in such tight spaces.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,353
5,488
113
When you guys figure this out... I'm going to hire you to make me one.
 

TechIsCool

Active Member
Feb 8, 2012
263
117
43
Clinton, WA
techiscool.com
Hey everyone this is something I have knowledge on since I have been working with Intel 775 heat sinks and the fans with my micro controller project. The PWM line is driven low at 0 - 3.5vdc with a maximum of 16vdc. The frequency is acceptable between 50Hz - 100Khz which is a very wide range. You can't just use a potentiometer to do this since PWM is not a resistor. Its how fast you turn it on and off. Without changing much you could just cut the pwm line on all the fans and then run them all to a single micro controller. Not the best since one host could be using 100% CPU usage it will run faster then the others. I would run them in four different zones and have some type of temperature monitoring software on the host to spin them up to 100% when the load requires it.

Patrick or PigLover have you tested Speedfan on the host? If it functions none of this talk actually needs to be done since it can be all handled via software. I know for a fact they work on a bunch of my Dell servers that I own and my P8P67 Pro
 

PigLover

Moderator
Jan 26, 2011
3,071
1,375
113
I know you can't just resister-mod a pwm fan. Well, you can to a point because lowering the operating voltage will lower the revs regardless of what the pwm control lead tries to do...

Speedfan is on my list of things to try before anything gets modified.
 

TechIsCool

Active Member
Feb 8, 2012
263
117
43
Clinton, WA
techiscool.com
I know you can't just resister-mod a pwm fan. Well, you can to a point because lowering the operating voltage will lower the revs regardless of what the pwm control lead tries to do...

Speedfan is on my list of things to try before anything gets modified.

Alright I dug into this some more it seems that the BMC interfaces with a FCB that does the controlling/sensing that chip is using a

Firmware Information:
---------------------
Chip : PIC18F45J10
Release Date : 08-26-2011
Target Product : Puma X3.5/Piggy/Loki/Sif/Thor
Major Version : 01
Minor Version : 18
Checksum : 515e

I have yet to find reference to anything like controlling it except this vauge message
Set Sensor Reading and Event Status | S/E 30H M | Only for fan devices | Supported Yes
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,353
5,488
113
Very interesting. The only issue here is that there are four systems with four fans so there are some dependencies since it is not even a 1:1 ratio.
 

TechIsCool

Active Member
Feb 8, 2012
263
117
43
Clinton, WA
techiscool.com
Very interesting. The only issue here is that there are four systems with four fans so there are some dependencies since it is not even a 1:1 ratio.
The way I understand it is that it is interfaced via IPMI. The BMC does the interface with the FCB so if one is hotter then the others I would assume all the fans would come up to speed. If in fact the FCB does actually control PWM and not just monitor the RPM sensors.
 

PigLover

Moderator
Jan 26, 2011
3,071
1,375
113
The FCB has a connection to the BMC on each MB. It also has a sensor connection to the drive bay too, independent of any MB. The only reason to get sensor data from the drive bay is so that the FCB could ramp the fans if the drives get too warm...which implies it is in control.

I'm traveling this week and should have everything needed to fire it up. We'll see what we can find out this weekend.
 

cactus

Moderator
Jan 25, 2011
830
75
28
CA
How about doing a uC reading the PWM coming from the PIC uC on the FCB and lowering the duty cycle that is seen by the fans?
 

TechIsCool

Active Member
Feb 8, 2012
263
117
43
Clinton, WA
techiscool.com
How about doing a uC reading the PWM coming from the PIC uC on the FCB and lowering the duty cycle that is seen by the fans?
I think that would work as well then you would get the control from the FCB but also have it slower then what would normally be speced. We will have to wait and see if anything works in software first before hardware is needed.
 

Dragon

Banned
Feb 12, 2013
77
0
0
Looking for simpler solutions.
Me too, one of my chassis has 5 hotswap fans and each fan sits on top of a circuit board that does automatic fan speed control. the board also allows manual override of fan speed by using a jumper.

There are different versions of fan control circuits, but they are all very simple, if that board can fit in the c6100 (placed vertically behind the disks) then all we need to do is give power to it, set jumper to low-speed and connect the fans to the board. I know the regional distributor, if they work I can order a bunch of them and send it to Patrick.

The fans mechanism(Hot swap)


The fans mechanism (non hot-swap)


The fans control board (From a non-hotswap fan chassis)


The fan speed control jumpers (From a non-hotswap fan chassis)
 
Last edited:

cactus

Moderator
Jan 25, 2011
830
75
28
CA
Dragon are you in the US? XCase sells those orange fan shroud chassis, but I haven't seen them in the US.
 

Dragon

Banned
Feb 12, 2013
77
0
0
Cactus, no I am not, I am in Hong Kong at the moment, and I was informed by the regional distributor that pretty much all those chassis with the orange ducts are made by the same company headquartered in Beijing and Shenzhen, for years big name/small name/listed companies have been buying OEM from them and slap their own labels on. It's only until recent years that they sell some of their own designs under their own brand name.

After getting access to the actual photos of the Dell fans, it looks like the simplest way is just replace the fans with slower ones. (replace the fans by cutting the 4 wires and reconnect them to the new fans, and let the 2 extra wires and socket from Dell's original setup remains unaffected and continue to do whatever it does)

 
Last edited:

Patrick

Administrator
Staff member
Dec 21, 2010
12,353
5,488
113
After getting access to the actual photos of the Dell fans, it looks like the simplest way is just replace the fans with slower ones. (replace the fans by cutting the 4 wires and reconnect them to the new fans, and let the 2 extra wires and socket from Dell's original setup remains unaffected and continue to do whatever it does)
My guess has been that this is what we will come down to but I don't want to get in the way of anyone's creative process.
 

PigLover

Moderator
Jan 26, 2011
3,071
1,375
113
I agree, which is why I already ordered the Evercool fans. But I still plan to explore other options along the way. A software control method would be best, but I don't actually expect to find one.
 

PigLover

Moderator
Jan 26, 2011
3,071
1,375
113
OK - back from Texas and all the parts arrived while I was gone. Populated one node with L5638s and 48GB RDIMM and lit it up.

The PSU fans are surprisingly quiet for older 1U redundant 1100watt PSUs. Not near silent, but not obnoxious at all. I don't think I'll need to do anything with those.

When the unit starts up it ramps all the fans to full speed for a few seconds. WOW - its pretty loud! But then they settle down to a much slower speed which is much quieter - according to the sensors in the BMC they are sitting at about 4,000RPM. In fact, compared to the Juniper 10Gbe switch its not too bad at all. Louder than the Juniper, but the tonality of the 80mm fans is much less obnoxious than the 40mm fans in the Juniper EX. If they never ramped up I might just be willing to leave them alone for my garage-based lab - but I know under load they will come up.

After thinking about it while I was traveling I think I'm just going to go straight to the fan replacement. I bought the Evercool's and connectors to do a dell 6-pin PCM to standard 4-pin PCM. Need put together the other three nodes and load an OS so that I can get temp readings to know that the Evercool's move enough air. Should have results sometime late this weekend.
 
Last edited:

PigLover

Moderator
Jan 26, 2011
3,071
1,375
113
Also - noticed that the C6100 with 4 nodes installed draws 28-30 watts when all four nodes are turned off. The MBs run the BMC on standby power, but 28watts is a lot of draw just for the BMC. Runs up to 300 watts when one node is staring up - while its running POST and the fans are full speed. Guessing the fans themselves are a bit part of this as these Delta's can draw 4.9A @ 12V (almost 60watts each) when running full speed. These things are certainly not power efficient...hopefully the Evercool replacements will drop the idle power draw a bit too.