Safety of using SFP Transceivers
-
@stephenw10 I was, my biggest concern was the transceiver burning out the 10g SFP port on the pfsense box. The 6100 was a bit of a splurge to future proof for 10g and worried about anything that might reduce the lifespan.
-
Hmm, well I've not heard of that happening on a 6100. But I couldn't vouch for all transceivers.
-
We use lots of different 10G SFP+ at work. They do certainly vary, but rarely have over heating issues as long as they are in a high end switch/router such as Juniper, Cisco, etc.
We did have quite bad over-heating issues in a TP-Link SX3008 switch. The problem was the SFP cages, so we modified the switch adding cheap self-adhesive heatsinks. Attached pictures show the main board before & after. Approx 20degC SFP temperature reduction.
-
@stephenw10 I too had been using a 10GTek SFP+ module in my 6100 but was concerned about the very hot temps. The GUI reported module temps exceeding 70C which is nuts.
I researched and found these new FS.com "low power" (1.5W) modules and ordered one. It arrived yesterday from China and I was excited to see if it ran cool. Maybe it does run cooler, but it's still too hot to touch for more than 2 seconds, and the module does not self-report temps like the other one did, so I honestly have no idea. My cheap IR temp gun says it's 52C right now, but it feels as hot as the 10GTek one did.
-
Mmm, well 52 is quite a lot less than 70. But those things just run hot. They are designed to run hot and mostly work fine like that. I wouldn't want my phone running that hot but....
-
As I said above, it depends a lot on the SFP cage mechanics. I don't know what the internals of the 6100 are, so cannot comment on that.
For example, here's the vendor types & temperatures for a variety of SFP+ in a Juniper EX-4600. Ambient in the room is about 20degC.
PIC port information:
Fiber Xcvr vendor Wave- Xcvr JNPR
Port Cable type type Xcvr vendor part number length Firmware Rev
0 10GBASE ER SM Arista Networks SFP-10G-ER 1550 nm 0.0 2
1 10GBASE LR SM FS SFP-10GLR-31 1310 nm 0.0
2 10GBASE ER SM MENARA NETWORKS 5ER0P2U-1310 1310 nm 0.0
3 10GBASE LR SM FINISAR CORP. FTLX1370W4BTL-E7 1310 nm 0.0
4 SFP28-25G-BASE-CU-5M n/a JUNIPER-1P1 1P153A n/a 0.0 REV 01
5 SFP28-25G-BASE-CU-5M n/a JUNIPER-1P1 1P153A n/a 0.0 REV 01
6 GIGE 1000T n/a FINISAR CORP. FCLF8521P2BTL-TR n/a 0.0
8 10GBASE USR MM TIBITCOM TXM-MPOLT-01EES 1577 nm 0.0
9 10GBASE LR SM SumitomoElectric SPP5300LR-J6-M 1310 nm 0.0 REV 01
10 GIGE 1000T n/a OEM GLC-T n/a 0.0
11 10GBASE ER SM HI-OPTEL ECI HSFP101541LECI 1550 nm 0.0
12 10GBASE ER SM PROLABS ER-SFP10G-47-LGI 1470 nm 0.0
13 10GBASE LR SM LG1 SFP-10GE-LR 1310 nm 0.0
14 10GBASE LR SM Arista Networks SFP-10G-LR 1310 nm 0.0 7
15 GIGE 1000T n/a OEM GLC-T n/a 0.0 ^x
21 10GBASE LR SM FINISAR CORP. FTLX1475D3BNL-TR 1310 nm 0.0
22 10GBASE SR MM LG1 SFP-10GE-SR 850 nm 0.0
23 GIGE 1000T n/a WORLDWIDEPACKETS XCVR-000CRJ n/a 0.0{master:0}
reading-sw-01> show interfaces diagnostics optics | match temp | except high | except low
Module temperature : 40 degrees C / 105 degrees F
Module temperature : 38 degrees C / 101 degrees F
Module temperature : 36 degrees C / 96 degrees F
Module temperature : 40 degrees C / 104 degrees F
Module temperature : 44 degrees C / 111 degrees F
Module temperature : 34 degrees C / 94 degrees F
Module temperature : 41 degrees C / 106 degrees F
Module temperature : 34 degrees C / 94 degrees F
Module temperature : 30 degrees C / 87 degrees F
Module temperature : 35 degrees C / 95 degrees F
Module temperature : 34 degrees C / 92 degrees F
Module temperature : 29 degrees C / 84 degrees F
Module temperature : 33 degrees C / 91 degrees F
Module temperature : 32 degrees C / 89 degrees F{master:0}
reading-sw-01> show chassis environment
Class Item Status Measurement
Power FPC 0 Power Supply 0 OK
FPC 0 Power Supply 1 Present
Temp FPC 0 Sensor TopMiddle C OK 47 degrees C / 116 degrees F
FPC 0 Sensor TopRight I OK 31 degrees C / 87 degrees F
FPC 0 Sensor TopLeft C OK 41 degrees C / 105 degrees F
FPC 0 Sensor TopRight C OK 29 degrees C / 84 degrees F
FPC 0 Sensor CPURight C OK 38 degrees C / 100 degrees F
FPC 0 Sensor CPULeft E OK 35 degrees C / 95 degrees F
FPC 0 Sensor CPU Die Temp OK 49 degrees C / 120 degrees F
FPC 0 Sensor QIC0 OK 38 degrees C / 100 degrees F
Fans FPC 0 Fan Tray 0 OK Spinning at normal speed
FPC 0 Fan Tray 1 OK Spinning at normal speed
FPC 0 Fan Tray 2 OK Spinning at normal speed
FPC 0 Fan Tray 3 OK Spinning at normal speed
FPC 0 Fan Tray 4 OK Spinning at normal speed{master:0}
-
@stephenw10 Thanks. I've already somewhat wrecked the SFP cage of my 6100... somehow the old module got stuck in there, and no amount of jiggling would release it. I just kept pulling on it until it eventually came loose. Now the modules no longer lock/click in place they just sort of slip in and out.
The thermals on the CPU itself seem "ok" and tend to hover right around 50C:
-
I just would like to add here, that the thermal design of many switches is ... bad not to say very bad.
The cooling of SFP+ cases should be much much better !! And earflow of fans should be routed along things which have to be cooled and not just blowing towards ... nothing!
I general like MikroTik but it also an example of bad thermal design. Worse it is very hard if not impossible to correct that with extra coal fins.
So complain towards the switch designers and if possible ...
- use fiber
- or SFP+ to 1G utp
- or at least use the latest sfp+ to 10G modules''
Personally:
I migrated all 10G connections to fiber in my server cabinet. And also to 10G connections in other rooms .... not so easy.And replaced the small fans in my Miktotik CRS317 with a 12 inch fan in an home made lid :(
But as said it is ridiculous that that is necessary!!
-
@louis2 Not sure what Mikrotik has to do with this conversation, but yes I know that all 10G SFP+ RJ45 modules run hot. That's why I dropped $150 for what I thought would be the coolest-running of the bunch, since it was a new design, rated at 1.5W vs >2W for most of the cheap-o Amazon stuff like 10Gtek, Hi-Fiber etc.
Also, the 6100 is fanless, and actually has a pretty neat and clever thermal design with that giant metal plate/heatsink as its base...
-
I have been down this road as well. Tried various SFP+ to RJ45 transceivers, including a more expensive lower power consumption one (around 1.8W as I recall). It's interesting to see that module power consumption has been able to be reduced further down to 1.5W since the last time I looked into these. In my case, I ultimately decided just to purchase a 10Gbit RJ45 expansion card since I had an expansion slot available on the firewall and didn't want to worry about any heat or compatibility issues. I used to have fiber inside the rack as well, but have since switched to all DAC cables as the runs are short enough (less parts needed and saves a few more Watts this way without compromising performance).
-
We run quite a few 10G RJ45 SFP in various Juniper & Cisco & Allied Telesys switches in our work lab with no heat issues. Most of them are between 30-40 DegC in a room with 22 ambient.
It's definitely the SFP cages in lower end switches.
You can get refurbished Juniper QFX-5100 or similar on eBay for << £400 these days, and will have no more heat issues.
