The fans would spin and the RPM cable would be plugged in, just no PWM. The FCB would get RPM stats, it just wouldn't be able to speed up/slow down the fans. Maybe I'm missing something though.
Taming the C6100
- Thread starter PigLover
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Goin to test this next week when I get another FC HBA so i can load OS into one of the nodes.
No these are not ledfans
No these are not ledfans
Tests!
81C Seems to be absolute max for cpu1, so not too bad since This will be a ESXi server and won't be seeing any high loads on all cores simultaneously.
Still bit noisy but it's all right, I'll look into using oringinal 3.3A Deltas with 556 timer based PWM controller, just to see is it viable at all. Or the 0.58A San Aces i also happen to have
81C Seems to be absolute max for cpu1, so not too bad since This will be a ESXi server and won't be seeing any high loads on all cores simultaneously.
Still bit noisy but it's all right, I'll look into using oringinal 3.3A Deltas with 556 timer based PWM controller, just to see is it viable at all. Or the 0.58A San Aces i also happen to have
Take a look at the other temps too. Log into the IPMI and get the MB#1 and MB#2 sensors. MB#2 is on the 5520 IOP chip near the back of the sled. It sits under the mezz card so its hard to cool, plus it is one of the hottest IOP chip designs Intel has ever shipped. Not at all uncommon to see it 15-20c higher than the CPUs in Dell's design because airflow is poor around it.
When my systems with slow fans see CPU temps in the 70s I can regularly see MB#2 reading in the 90c range.
Note that while that seems hot, Intel clams Tmax for the 5520 IOP is something like 105c. I'm not so comfortable with that - having things in the chassis that can boil water bother me.
When my systems with slow fans see CPU temps in the 70s I can regularly see MB#2 reading in the 90c range.
Note that while that seems hot, Intel clams Tmax for the 5520 IOP is something like 105c. I'm not so comfortable with that - having things in the chassis that can boil water bother me.
First node in use.
How is it possible that Delta had Sense and PWM wire colors swapped. Sense being blue and PWM yellow, usually it's the other way around :I
Rather queit box for a server right now, 80mm fans going at 2900rpm in front and 140mm fans at 1000rpm at top. It's like a workstation that's letting you know it exists when you're in the room with it but doesn't bother you soon as you close the door.
Still going for the waterblocks tho
How is it possible that Delta had Sense and PWM wire colors swapped. Sense being blue and PWM yellow, usually it's the other way around :I
Rather queit box for a server right now, 80mm fans going at 2900rpm in front and 140mm fans at 1000rpm at top. It's like a workstation that's letting you know it exists when you're in the room with it but doesn't bother you soon as you close the door.
Still going for the waterblocks tho
Well, this is interesting. I just got the 3 c6100s that I ordered from lextecny2012 on eBay. I've tested a few of the nodes, but I couldn't help but notice that it really wasn't as loud as I was expecting. I pulled a fan out and realized that the fans that came with my units are not the 20,000 rpm, 60W beasts that others have reported getting, but instead are Delta PFC0812DE (3.30A) @ 65-70dB. Still too loud for my tastes, so I'm am going to experiment with other fans (first, the Evercool fans that PigLover reported using). With the "stock" fans in Prime95, I am seeing 50s-60s on the L5520 CPUs.
It gets more interesting though: These fans only have a 5-pin connector on them and only 4 pins are occupied. The headers on the FCB are also only 5 pin. The servers all came configured with dual-PSU support, and each PSU board plugs into the FCB. Apart from their non-standard connectors, it seems like the FCB mixes the power prior to the headers and presents a standard PWM connection. Maybe there were multiple revisions of the c6100 (XS23-TY3), and the few I got have a newer FCB and fan design?
I will report back with I get the Evercool fans in.
It gets more interesting though: These fans only have a 5-pin connector on them and only 4 pins are occupied. The headers on the FCB are also only 5 pin. The servers all came configured with dual-PSU support, and each PSU board plugs into the FCB. Apart from their non-standard connectors, it seems like the FCB mixes the power prior to the headers and presents a standard PWM connection. Maybe there were multiple revisions of the c6100 (XS23-TY3), and the few I got have a newer FCB and fan design?
I will report back with I get the Evercool fans in.
Last edited:
For the 5 pin connector the pinout is: GND-12V-Sense-PWM-N/C
And it should be the older version as it doesn't have any fw updates for it. I also got that model and it it's ancient, the date on that system is from 2009 iirc.
And it should be the older version as it doesn't have any fw updates for it. I also got that model and it it's ancient, the date on that system is from 2009 iirc.
Last edited:
Thanks for the pinout infomation! Also, you're right. My servers were manufactured in 2009. Date code on the motherboard is: 9B3 (November 3, 2009).
This is very possible. I'm not particularly complaining though; I was able to update my nodes to the latest BIOS and BMC (see the c6100 firmware update thread). I offered $650 and he countered met me in the middle (pretty much; bit higher). I wouldn't say the servers are "like new" as described, but they work great!
Can you explain in a bit more detail how exactly you did this? I've played with IPMIutils and all I've gotten it to display is the current RPM and thresholds. I currently have a fan completely disconnected and `ipmiutil health` and `ipmiutil alerts` don't report any problems but report 0 rpm for that fan.
Edit: For those interested, I figured it out! ipmitool (mentioned by mulder on Page 4) is a lot better than IPMIutils IMO. To (remotely) set the thresholds, it is quite simple:
Code:
$ ipmitool -H [BMC IP ADDRESS] -U root -P [password] sensor thresh "FCB FAN4" lcr 0For those who wonder: It is possible to set the lower-critical RPM threshold to 0, although if a fan is unplugged it still complains because it is "unable to read sensor: Device Not Present".
Last edited:
I live in an appartment where one of my three bedrooms are used as an office, and where I house my computers.
From the very start I knew that I had to silence the C6100 when I got it, and I'm very gratefull for PigLover and others that have taken the small steps first, and documentet
their attemts, towards the goal of making it possible to sit next to a C6100 without going crazy from the noice
Due to lack of skill and equipment, I wanted to go the route where you swap out the fans without cutting/modifying wires. Soldering is not an option unless I pay someone to do it
Earlyer in this thread the Supermicro FAN-0074L4 came up as a good option for a more silent fan than the original Delta-fans, so I bought 4 of these from eBay.
From others in this thread I also knew that the cable on these Supermicro-fans were quite short (only 2"/5cm, see picture), so I also bought some Dell adapter-cable from 4-to-5 pins.
The adapter-cables I bought was not the ones from FrozenCPU.com (FrozenCPU.com 4-Pin PWM to 5-Pin Dell Adapter Cable - 8" Length (CB-PWM-D5) - FrozenCPU.com),
which were only 8"/20 cm long, but I found some ~14"/35 cm long cables on eBay (X2 Motherboard 4pin to Dell 5pin 4 Wires Adapter Cable Wire Converter 383 002 | eBay)
and I was very glad I did. From fan number 4 there is just enough cable to connect to the fan controller (see picture below).
The Supermicro FAN-0074L4 had the San Ace 80 model 9G0812P1F03 fan inside, as confirmed by RimBlock. However the fan had 2 closed and 2 "open" ribs going from corner-hole to corner-hole, so I guess I need to remove those to be able to reuse the rubber-plugs holding the original fans to the chassis.
overview:
fan:
fan right:
fan left:
fan cable:
Now I just need to put it all back togheter and hope that I remember where all the screws went
Edit
When test-mounting the fans now, I just realized that by putting the cables ontop of the fans, instead of below and through the fan-guard like the original cables are, the cablerun are abit shorter, and much easier to handle.
/Edit
Edit2
Dont buy the cable-extender on the eBay-link posted above, the cables do not have the same order on both connectors, and will not work as is.
/Edit2
From the very start I knew that I had to silence the C6100 when I got it, and I'm very gratefull for PigLover and others that have taken the small steps first, and documentet
their attemts, towards the goal of making it possible to sit next to a C6100 without going crazy from the noice
Due to lack of skill and equipment, I wanted to go the route where you swap out the fans without cutting/modifying wires. Soldering is not an option unless I pay someone to do it
Earlyer in this thread the Supermicro FAN-0074L4 came up as a good option for a more silent fan than the original Delta-fans, so I bought 4 of these from eBay.
From others in this thread I also knew that the cable on these Supermicro-fans were quite short (only 2"/5cm, see picture), so I also bought some Dell adapter-cable from 4-to-5 pins.
The adapter-cables I bought was not the ones from FrozenCPU.com (FrozenCPU.com 4-Pin PWM to 5-Pin Dell Adapter Cable - 8" Length (CB-PWM-D5) - FrozenCPU.com),
which were only 8"/20 cm long, but I found some ~14"/35 cm long cables on eBay (X2 Motherboard 4pin to Dell 5pin 4 Wires Adapter Cable Wire Converter 383 002 | eBay)
and I was very glad I did. From fan number 4 there is just enough cable to connect to the fan controller (see picture below).
The Supermicro FAN-0074L4 had the San Ace 80 model 9G0812P1F03 fan inside, as confirmed by RimBlock. However the fan had 2 closed and 2 "open" ribs going from corner-hole to corner-hole, so I guess I need to remove those to be able to reuse the rubber-plugs holding the original fans to the chassis.
overview:
fan:
fan right:
fan left:
fan cable:
Now I just need to put it all back togheter and hope that I remember where all the screws went
Edit
When test-mounting the fans now, I just realized that by putting the cables ontop of the fans, instead of below and through the fan-guard like the original cables are, the cablerun are abit shorter, and much easier to handle.
/Edit
Edit2
Dont buy the cable-extender on the eBay-link posted above, the cables do not have the same order on both connectors, and will not work as is.
/Edit2
Last edited:
From the picture on eBay, the wiring on the 5-pin end of that cable doesn't look quite right. I'd have to pull mine back out to be sure, but as I recall it was a six-pin connector and all six wires were in use. Two power, two return/ground, one sense, one PWM control. To make it into the 4-pin fan connector you tie the two power together and two return/ground together in-line. This allows the fan to be powered from either PSU (there are diodes on the FCB to permit the simple line-tie between the PSUs to operate safely).
Have you actually put this together and powered on? Perhaps you have the old-style FCB, but for the majority of C6100 users I'm not sure this is the right adapter cable.
Have you actually put this together and powered on? Perhaps you have the old-style FCB, but for the majority of C6100 users I'm not sure this is the right adapter cable.
Just put it together now, and when I pressed the powerbutton the leds started blinking green and amber, for which the manual says on page 20:
Since the servicetag doesnt show up on the website, can the fact that I probably have a DCS-server be the answer? I need to silence this server if I'm gonna keep it operating in my appartment.
The original fans have a 5-pin plug, but with only 4 wires. The FCB also only had 5-pin plugs for the fans.
Since the servicetag doesnt show up on the website, can the fact that I probably have a DCS-server be the answer? I need to silence this server if I'm gonna keep it operating in my appartment.
Do the new fans spin? Did it power on? Is the only problem the green/amber flashing of the node power lights?
If that is the case then it is likely that the new fans are just spinning too slowly to make the BMC alarm rules happy. If its running, log into the IPMI and look at the "sensor readings" page. You should be able to see which sensor raised an alarm - and its probably the fans. You can use "show thresholds" on that page to see what the alarm thresholds are set to. And you can use ipmitools to change the thresholds to values that work better for the new fans.
Of course, if the fans didn't spin or the node refused to power up then you've got a different problem to deal with. Either you knocked a connector loose when working on the fans or something about the new fans/cables is making the FCB/BMC unhappy.
The fans dont spin up, and when I view the "sensor readings" for one of the nodes it says "Reading/State Unavailable" on FCB FAN1-4.
Got the same status on all 6 PSU-items on that list, dont know if that means anything. I can always mount the old fans back and see if that fixes anything.
I did change the left front board ear, while changing out the fans, but I doubt I knocked something loose. I dont see anything obvious when I look at the interior now, and I was extra carefull when I disasembled the drivechassis, and made sure I didnt pull on any of the wires.
Got the same status on all 6 PSU-items on that list, dont know if that means anything. I can always mount the old fans back and see if that fixes anything.
I did change the left front board ear, while changing out the fans, but I doubt I knocked something loose. I dont see anything obvious when I look at the interior now, and I was extra carefull when I disasembled the drivechassis, and made sure I didnt pull on any of the wires.
Sensor readings are only available if the node itself is running. You can log into the BMC with the node off, but it depends on the system management bus provided by the chipset to read the sensors and the node has to be up to read them.
If the fans did not spin then odds-on the pinout for the adapter cable you used is wrong. By default, until the BMC gets sensor data from at least one node, the FCB it supposed to spin the fans at full speed.
I'll switch back to the original fans tomorrow (its past midnight here now), and see if I can get this thing back to life.
Think the node was running, because the cpu-temp was reading 92 degrees C, before I noticed and pulled the powercord. Tried to touch the cpu-cooler, but it was too hot to touch.
Think the node was running, because the cpu-temp was reading 92 degrees C, before I noticed and pulled the powercord. Tried to touch the cpu-cooler, but it was too hot to touch.
I swapped the Supermicro-fans with the original Delta-fans, and the chassis powered up like before and the FCB FAN1-4 gave the rpm of the fans.
Atleast I know that I didnt mess things up when I changed the front board ear.
The LOW CT threshold for the fans is set at 1500 rpm.
Can I have a custom FCB that dont recognize the new fans, since I likely got a Dell DCS-server instead of a "normal" Dell PEC-server? (My servicetag dont show up on the dell webpage)
Atleast I know that I didnt mess things up when I changed the front board ear.
The LOW CT threshold for the fans is set at 1500 rpm.
Can I have a custom FCB that dont recognize the new fans, since I likely got a Dell DCS-server instead of a "normal" Dell PEC-server? (My servicetag dont show up on the dell webpage)
No. There is no way for the FCB to recognize what fans are connected. If the fans you installed are not spinning up, it is more likely what PigLover said:
___
I haven't noticed this functionality on the few c6100s I have. They are the older 5-pin models from late 2009 though. The newer 6-pin revision might be a bit different. (I haven't updated the FCB firmware because I'm worried that the newer firmware will not be compatible).
Last edited:
It didn't do it on my older PIC16 (5pin) enclosure but it does spin up on the new PIC18 (7pin) enclosure. The FCB firmware are not compatible with the PIC16 (5pin) boards.