> The SFP+s use the same SC connector used in PON networks, but otherwise use Active Ethernet. I’m not sure if they’re using IEEE standards or are proprietary on the glass, but I’ll assume the latter.
It's actually standards compliant (unless they screwed something up). IEEE 802.3 clause 158, 10GBase-BRx. x = 20 in this case, at least the ones on the Amazon link say 20km.
You can get them with LC connectors too, which I'd in fact recommend because those are smaller, a bit more precise, and waaaaaay more common these days. https://www.fs.com/eu-en/products/36351.html (note you need a pair of "23" and "32")
As to the author's core complaint (CAT5/6 degrading to 100mbps), I've found that the issue is almost always the termination at the jacks. Reseating the wires with a punchdown tool, or failing that, cutting a bit more off the ends and ensuring that the pair twists run all the way to the punchdowns has been enough to restore 1000mbps negotiation the few times I've had this issue.
This is how my apartment complex retrofitted all 300ish units with AT&T fiber. Fiber runs from your building's telecom room to a small hole they drilled above everyone's front door, then along baseboards and around door jambs until it reaches ~the center of the unit (concrete walls...) in the living room.
It is genuinely invisible until you get up close. You're far more likely to notice the spots where someone got sloppy with the glue in some corner than the fiber itself. With proper attention to detail during installation (and assuming you have white door jambs and baseboards) it would be basically invisible.
If your CAT6 is down to 100 Mbps, check that all the conductors have been properly crimped into the connector -- or consider re-crimping. It's possible the building has shifted and damaged the fished cable, but it's more likely the problem is with the connector(s).
I had a lot of trouble with this when having my contractors run cat6 in my house during a recent renovation. Their electrician kept saying "it's done!" and then I'd test it and find that I was getting only 100mbps. Then they'd re-terminate the ends and a few more of the runs would get full speed when I re-tested. Repeated this cycle a few times until they all worked.
That would 100% been the right move. Some friends of mine with networking and ops backgrounds told me, "never trust an electrician to do low-voltage work right." After my experience with this one, I think I should have listened!
Network engineer of 20+ years here, but I'm struggling with this 'fished' terminology. Does it mean when a cable is pulled through on a piece of string?
In my case, when I upgraded from Cat5 to Cat6, I taped on to the existing 5 and used it to pull the 6 through the attic and down the wall. In the attic, you usually use a string, with a weight of some kind attached (if it's a new install), to drop down the wall seal plate, which you've drilled a hole into, down to the hole you've already carved out for your cable box near the baseboard area.
> Does it mean when a cable is pulled through on a piece of string?
That is often exactly what is meant. The problem of running Ethernet through existing structures has spawned a bunch of tools, including glow-in-the-dark "fish tape."
The process is often brutal, and looks like tug-of-war when it takes a couple men to pull the bundle through a couple 90 degree turns. As such, we have "plenum" cable, which is much tougher than typical patch cable, because the latter would just get shredded inside the structure, or snap when pulled with a few hundred lbs of force.
> The problem of running Ethernet through existing structures has spawned a bunch of tools, including glow-in-the-dark "fish tape."
None of these tools are new (well, the glow fish sticks I think are a recent innovation) or unique to ethernet. Electrical wire is installed the same way, eg. in conduit. There's even a special lube for making your wire slide through the conduit easier.
I had to buy a new crimping tool a while ago because I couldn’t find the one I had from a few years before.
I ended up buying a small kit containing both crimping tool and a couple of additional tools.
One of the tools included is an Ethernet cable tester. It has two parts; one for each end of the cable. The part on one side is powered by 9V battery (same one used in smoke detectors etc).
You connect the parts to the RJ45 plugs of each side. And it has individual LEDs for each of the eight leads in the cable. They are lit up in sequence. Any that don’t light up means that that pin is not properly connected, or the cable is broken.
And using that tool is how I found out the reason one of the router ports report 100M instead of 1000M connection. Half of the leads on the cable connected to that port are damaged or not properly connected! And the four that are working are the ones needed for 100M.
I remember that when I originally crimped that cable years ago I was struggling for quite a while, getting no connection at first until after a couple of attempts I got it connected. I didn’t notice at the time, nor for a long time after, that it was only a 100M connection. And it took until I had this tool to figure out the why of it being only 100M.
You also might have crimped it for 100Mbps specifically. 100Mbps ethernet only uses 2 wire pairs, not all 4. It was common, and intended, to carry 2 separate connections through a 4-wire cable. Ethernet and up to two landlines (or a 2-pair landline, which existed) was also a good combination.
I'd be curious to see what the actual link speed is between the two endpoints, rather than a speed test out to the ISP. Are you able to do a speed test with something like iperf3 locally?
I confess, I've read this twice and I don't really get what's going on here; and I have both MoCA and Powerline going on in my house as well, so I'm like -- why not one of those?
Fiber is much nicer. You can get as much bandwidth as you'd ever want.
10G is trivially doable on cheap hardware. You can also go 100G, 200G or 400G over the exact same cable, it just costs a lot more on the endpoints.
If you run anything modern you probably can benefit from 10G at least. A NVME does 8 GB/s easily, it's a real pity to limit speeds to 1 gigabit. And Ethernet past 1 Gbps quickly gets pretty terrible to deal with, so might as well go with fiber then.
10G on twisted pair isn't too bad either. Used enterprise NICs are cheap on eBay. New low cost cards are available, too (still costs more than used gear, but if you want lower power or multigig, it's worth a look). Switches aren't so cheap, but IMHO, 10g-baseT is comparable to sfp+ and a transceiver on each end.
Obviously, if something is wrong with your existing cabling, 10g-base-t is probably not going to work either. 10G is probably the limit for twisted pair, but I think it'll hold up as a reasonable speed for a long time.
10G on twisted pair is terrible, it's power hungry and the SFP modules have a tendency to overheat. You need decent ventilation and not have 10G twisted pair connectors next to each other. Also only goes to 30M or so.
IMO, if you want 10G just go with fiber, it's less of a pain and more reliable. With twisted pair you may find out that the module overheats at times and the link drops out.
I don't use sfp+ modules for twisted pair... It didn't cost anymore to get nics and switches with the ports I need vs sfp+ and transceivers; although there's a lot fewer switches to choose from. 30 meters feels like a lot for most homes. 10g-base-T spec calls for 100m on cat6a and 55m on cat6. Running on cat5(e) isn't to spec, but largely works in a home environment where you tend not to have to deal with a lot of crosstalk from dense conduit.
And the benefit is I didn't have to run new cable. Lots of homes have some twisted pair in the wall, very few have fiber in the wall. Homes from the late 90s on probably have cat5 for phone lines.
Sure, 10GBASE-T works. It's just not very good overall and you should plan to put your 10G on fiber any time you can. It's not particularly expensive or complicated. If you can't, then sure, use -T.
Until everything moves to SFP it's gonna be hard to make that transition for anything other than core runs. Especially if you need PoE - I've doubled my cable runs and lost a ton of flexibility.
Powerline struggles to reach gigabit and has all sorts of interference issues, and MoCA is expensive and also not full duplex. Bandwidth goes down as you attach it to more nodes.
I recently switched to "invisible" fiber after having used various generations of MoCA/G.hn/Powerline ethernet over the years. The best part is that you can get such cabling down to 0.6mm thin (akin to fishing line) and get it pre-terminated, meaning you don't have to drill holes anywhere as it can easily fit through doorways and supports very tight bend radii without breaking (exceeding G.657.B3). Also, pairing it with bidirectional transceivers means you only need to run one cable to each endpoint.
GoCoax/MoCA is a fantastic non-destructive option if your copper can handle it. Just make sure you install a filter at your ingress point to prevent sending data to neighbors (MoCA does have encryption, but why trust math when you can trust physics and math?).
You can fish standard fiber very well too. It's thinner than pretty much all network cable.
The biggest part is the LC connector, and you can just take that apart and put it back together -- just remember which was the left and which was the right wire.
German here, living in single digit distance to military areas.... The shit they buried under argon et al decades ago, man ... tap in if you have the patience ... gives you seconds around the globe
Last time I cabled my house I laid out two cables between the switch and every spot, precisely for this reason. I think now I’d lay fiber as well, for future use.
I've seen technology change far too much to do anything for future use. I remember running RG58 coax between rooms for ethernet. Then cat 3 for 10baseT. I came close to running fiber at one point but it would have been the wrong size fiber for anything modern (62uM still works but is out, and it would have been terminated wrong. If you want a future proof, you run yourself conduit from each room to a central data closet and put a string in there so you can pull whatever cable you want when you want it.
Article really offers no explanation of what "invisible fiver" is. Its just fiber with a thin, white jacket meant to be glued along baseboards where it is discreet.
I’m sure there are some safety issues I’m not considering that make it impractical to pump a high-powered laser into a wall cutout, burn a line-of-sight through all the studs, and install an optical link along the resulting path...
These days you can get a cheap-o chinese fiber splicer kit for like $500-800. They work pretty well and are in many ways simpler than terminating ethernet. Super convenient for wall pulls and so on.
It's actually standards compliant (unless they screwed something up). IEEE 802.3 clause 158, 10GBase-BRx. x = 20 in this case, at least the ones on the Amazon link say 20km.
You can get them with LC connectors too, which I'd in fact recommend because those are smaller, a bit more precise, and waaaaaay more common these days. https://www.fs.com/eu-en/products/36351.html (note you need a pair of "23" and "32")
As to the author's core complaint (CAT5/6 degrading to 100mbps), I've found that the issue is almost always the termination at the jacks. Reseating the wires with a punchdown tool, or failing that, cutting a bit more off the ends and ensuring that the pair twists run all the way to the punchdowns has been enough to restore 1000mbps negotiation the few times I've had this issue.
It is genuinely invisible until you get up close. You're far more likely to notice the spots where someone got sloppy with the glue in some corner than the fiber itself. With proper attention to detail during installation (and assuming you have white door jambs and baseboards) it would be basically invisible.
That is often exactly what is meant. The problem of running Ethernet through existing structures has spawned a bunch of tools, including glow-in-the-dark "fish tape."
The process is often brutal, and looks like tug-of-war when it takes a couple men to pull the bundle through a couple 90 degree turns. As such, we have "plenum" cable, which is much tougher than typical patch cable, because the latter would just get shredded inside the structure, or snap when pulled with a few hundred lbs of force.
None of these tools are new (well, the glow fish sticks I think are a recent innovation) or unique to ethernet. Electrical wire is installed the same way, eg. in conduit. There's even a special lube for making your wire slide through the conduit easier.
Recrimping is always the best first step. Sometimes running 2.5gb instead of gigabit will help too.
I ended up buying a small kit containing both crimping tool and a couple of additional tools.
One of the tools included is an Ethernet cable tester. It has two parts; one for each end of the cable. The part on one side is powered by 9V battery (same one used in smoke detectors etc).
You connect the parts to the RJ45 plugs of each side. And it has individual LEDs for each of the eight leads in the cable. They are lit up in sequence. Any that don’t light up means that that pin is not properly connected, or the cable is broken.
And using that tool is how I found out the reason one of the router ports report 100M instead of 1000M connection. Half of the leads on the cable connected to that port are damaged or not properly connected! And the four that are working are the ones needed for 100M.
I remember that when I originally crimped that cable years ago I was struggling for quite a while, getting no connection at first until after a couple of attempts I got it connected. I didn’t notice at the time, nor for a long time after, that it was only a 100M connection. And it took until I had this tool to figure out the why of it being only 100M.
10G is trivially doable on cheap hardware. You can also go 100G, 200G or 400G over the exact same cable, it just costs a lot more on the endpoints.
If you run anything modern you probably can benefit from 10G at least. A NVME does 8 GB/s easily, it's a real pity to limit speeds to 1 gigabit. And Ethernet past 1 Gbps quickly gets pretty terrible to deal with, so might as well go with fiber then.
Obviously, if something is wrong with your existing cabling, 10g-base-t is probably not going to work either. 10G is probably the limit for twisted pair, but I think it'll hold up as a reasonable speed for a long time.
IMO, if you want 10G just go with fiber, it's less of a pain and more reliable. With twisted pair you may find out that the module overheats at times and the link drops out.
And the benefit is I didn't have to run new cable. Lots of homes have some twisted pair in the wall, very few have fiber in the wall. Homes from the late 90s on probably have cat5 for phone lines.
Quite frankly, people should be doing fiber LANs more, even if the WAN is coax or wireless.
But our townhouse was renovated in a post-cable era so Coax was never put in. If it was a 1-2 decades older it would've had coax.
The biggest part is the LC connector, and you can just take that apart and put it back together -- just remember which was the left and which was the right wire.
That should save the need for special hardware.
I’m sure there are some safety issues I’m not considering that make it impractical to pump a high-powered laser into a wall cutout, burn a line-of-sight through all the studs, and install an optical link along the resulting path...
https://momandpop.network/wp-content/uploads/2026/11/PXL_202...
Zoom in on the black box. There's an orange ethernet cable plugged in, and the other plug is the "invisible" fiber.