I did replace the fans on my 6450 with 120mm fans on top of the lid , now this switch is dead silent , i might get some 7250s in a few weeks and try this method on them.
Brocade ICX Series (cheap & powerful 10gbE/40gbE switching)
- Thread starter fohdeesha
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Because "business class" switches like these are intended to be installed in network rooms and server rooms where a low noise threshold is much less important than longevity through adequate cooling. They were never designed to be put in a quiet residential setting.
Actually on one of my switches, one of the fans is having reported as failed, but all 3 of the fans inside the chassis are spinning.
I have an identical setup between two 7250-24Ps, two non-original fans and one of them is wired up to a splitter for a cheapo 60mm fan. Any reason fans would still spin but one header would report a failure?
I have an identical setup between two 7250-24Ps, two non-original fans and one of them is wired up to a splitter for a cheapo 60mm fan. Any reason fans would still spin but one header would report a failure?
Yeah I ordered new fans from Digikey per previous recommendations as well as removed the heatsink off the broadcom chip which I assume is one of the main contributors to fans having to ramp up, hence why everyone is strapping fans to the heatsink.
I'm going to take some time to lap the heatsink and use some of my own thermal paste to see if that'll improve things. Replace the fans per previous mods recommendations and see what happens! Hopefully I'll get things to run cooler and the fans will ramp down on their own. If it's too loud for me even after that I'll be selling a modded switch shortly, hah.
Absolutely this. Regardless of any recommendations these things were made to be in a data center not my closet. Another thing to consider is the age of the equipment we're dealing with. Who knows how many hours and hours of operation these guys have been chugging away at. Thermal paste dries. Heatsinks fall off. Fans die. Components fail. It's all the nature of the beast when dealing with second hand and older- yet still very capable equipment for our homelab purposes.
Though I did take @fohdeesha's OP as face value. Hey you said nearly silent! haha. It's all good. I enjoy this homelab tinkering work. I don't mind tearing a switch apart just to see if I can make it run cooler and quieter. If I totally bork the thing it's all good. Was curious to see if anyone else had redone heatsink of these guys.
Just some photos of the chip and it's original thermal interface.
BTW here's an off the wall question. I see that that POE is delivered via a separate board and several headers. The headers are clearly labeled for what range of ports PoE is supplied to on each of them. Any idea what would happen if I were to say, leave some disconnect for a range of ports? What if I didn't attach it at all in a an attempt to make the switch non-PoE? Just curious if anyone ever experimented with doing something like that and what negative or possibly positive results may be?
I'm obviously just guessing here. I'm not a hardware expert.
i literally just did a repaste on one of my switches last night, i was curious if it would make a difference. i would say, not really. i used arctic mx4 2019 edition. it was kind of annoying getting that thermal pad off but didn't take too much effort. i would probably apply heat and try to pull it off in one piece, because it is kind of fabric-like? it doesn't really come off like paste at all, it tears and rips. but again i probably wouldn't bother. it didnt make a difference temperature wise.
Attachments
Yeah why I said I'd lap it off. Literally sand it down a bit and get the aluminum heatsink as smooth as possible. Too bad on not seeing a difference. Will see how I fair.
I learned last year while preparing to replace B0-stepping Atom C2000s the "paste" typically used in enterprise-class non-upgradable chips aren't thermal paste. Some are made of thermal plastic that melts and forms a bond between the chip and heatsink during first use power up. These TIMs have thermal conductivity of ~7.5 W/mK per their datasheet. https://www.mouser.com/datasheet/2/987/THR_DS_Tpcm_7000_Data_Sheet_03172023-3194660.pdf. My guess is these TIMs have more lasting properties than typical thermal paste.
I'm always concerned that improperly removing the heatsinks would delid the chip from its package. I have read some vendors use thermal glue instead of what I've linked above. Luckily by twisting the heatink and not force lifting I was able to remove the CPUs without incident. It would have been fine if the chips were damaged because they would have been tossed anyway due to the LPC bug.
I'm always concerned that improperly removing the heatsinks would delid the chip from its package. I have read some vendors use thermal glue instead of what I've linked above. Luckily by twisting the heatink and not force lifting I was able to remove the CPUs without incident. It would have been fine if the chips were damaged because they would have been tossed anyway due to the LPC bug.
I also just got a 7250-48p. For only $110 shipped! I'm pretty pumped about it. What a powerful switch for homelab. Unfortunately, it is also power hungry. Mine draws 63 watts without anything plugged into it. I would agree with others that "nearly silent" is... relative, I suppose. My homelab was nearly inaudible standing right next to it until I added this switch. Ha!
I also went ahead and updated to the latest FastIron 9 firmware. I know Ruckus recommends the latest FastIron 8, but I really wanted the snazzy new web interface. So far, so good. Next up, try and quiet this sucker down. I unplugged two fans, and that helped quite a bit. The switch is running hotter, but still around 8 deg-C lower than it bumping to speed 2. Here's to hoping that running hotter doesn't significantly shorten its life span.
I'm looking into quieter fan options, now.
I also went ahead and updated to the latest FastIron 9 firmware. I know Ruckus recommends the latest FastIron 8, but I really wanted the snazzy new web interface. So far, so good. Next up, try and quiet this sucker down. I unplugged two fans, and that helped quite a bit. The switch is running hotter, but still around 8 deg-C lower than it bumping to speed 2. Here's to hoping that running hotter doesn't significantly shorten its life span.
I'm looking into quieter fan options, now.
Yeah agreed and I too have encountered the same type of TIMs that are a thermal glue but it does vary. I've mostly messed with HBAs for my NAS and some had normal dried up paste and some were definitely thermal glues. No pads though. Not too sure the best way to remove the glue stuff other than physical force. Hell I'm not even sure if it's a good idea to mess with things glued at all and if you'll find a temp decrease using a normal paste. I know for my HBAs glue or no glue using a typical paste after cleaning it up always brought the temps down.
I've found a okayish indicator is to see how the heatsink is applied on the chip. If there are visible pushpins there's a better chance at it not being a glue (from my extremely limited experience). Obviously if the heatsink is just sitting on top of a chip it's definitely a glue or adhesive pad or something meant to be a permanent fixture. There's plenty on the board of this ICX7250-48P that I'm not going to mess with since they're attached like that. But this heatsink that was in the Broadcom chip was pushpins so I figured f it let's check it out. It still has the fabric square which is bigger than the surface of the chip (a silicone gap filler I think?) but once it came off the Broadcom chip itself almost looked untouched. The question is will removing it and using a typical paste improve things. We'll see I guess!
If the switch temps are still high and the fans are too loud for me I'll make my peace and decision with it then, hah.
mine came off very easily- an extremely small twist. Why I am hopeful that I'll see improvement. Waiting on getting the new fans before I go full send.
I’ve had my 7250-24P for about half a year, replaced the fans when I first got it and that dropped about 10 Dba, then replaced them again and got another 4-5 down. I did add one on the ASIC too.
Definitely still noisy but much better.
Also worth noting that these are designed to survive beefy temps. Just look at how high they need to go before they increase fan speed! So y’all have a little leg room there. Just make sure where-ever you’re putting it has enough airflow that it won’t build up around it
Definitely still noisy but much better.
Also worth noting that these are designed to survive beefy temps. Just look at how high they need to go before they increase fan speed! So y’all have a little leg room there. Just make sure where-ever you’re putting it has enough airflow that it won’t build up around it
I tried looking through the 470 pages but couldn't find anything for this lol. Have a 7250-24P, followed the guide here (which is awesome thank you Fohdeesha). If I upgrade the firmware to a FastIron 09xxxx, does the licensing guide still apply or would i have to get a new one from Ruckus
I'm pretty sure I did just that--I definitely did not start over from scratch. The only thing that I remember being different is the version of Fastiron on the 7250 no longer uses
dual-mode--you can just set a port to have an untagged vlan and then any number of tagged vlans. But even that didn't trip up my config restore as far as I can remember.I'd be cautious running with just a single fan, especially if/when you plug in some POE devices. Trying to exhaust 80W+ of heat through a single 40MM fan could go badly. On the other hand, there's no question that the cooling in these switches is substantially overprovisioned for homelab use. I know that I at least am pretty unlikely to push anything more than a tiny fraction of what these switches are capable of, both from a POE and a traffic standpoint.I also just got a 7250-48p. For only $110 shipped! I'm pretty pumped about it. What a powerful switch for homelab. Unfortunately, it is also power hungry. Mine draws 63 watts without anything plugged into it. I would agree with others that "nearly silent" is... relative, I suppose. My homelab was nearly inaudible standing right next to it until I added this switch. Ha!
I also went ahead and updated to the latest FastIron 9 firmware. I know Ruckus recommends the latest FastIron 8, but I really wanted the snazzy new web interface. So far, so good. Next up, try and quiet this sucker down. I unplugged two fans, and that helped quite a bit. The switch is running hotter, but still around 8 deg-C lower than it bumping to speed 2. Here's to hoping that running hotter doesn't significantly shorten its life span.
Check out the fan replacement method listed above. It definitely improved my noise situation without making my temps noticeably worse.
I blew up one of those PoE chips on that daughter board that powers a bank of 8 ports when I connected my cable modem to a switch port on my 7250-48P. Switch power cycled itself and everything but to my absolute shock it only took out that PoE chip and everything else was fine. I just lost the ability to use PoE on 8 ports. I went in there after and removed that pin header feeding that chip and I've seen no ill effects.