6 ms·

In 2021, if you’re running network in the walls, run fiber to switches with SFP adapters. We maxed out twisted pair copper years ago (you can squeeze more over

In 2021, if you’re running network in the walls, run fiber to switches with SFP adapters. We maxed out twisted pair copper years ago (you can squeeze more over shorter runs, but that hits diminishing returns quickly) and home internet speeds are going above 1gbps. If you install in-wall cat6 it’ll probably be obsolete by the end of the decade.

MikeKusold 6y ago
Cat6a supports 10gbps. That should be good enough for the foreseeable future. Fiber is significantly more difficult to pull through the wall since it is more fragile.
- yardie 6y ago
  Fiber is not significantly harder than Cat6 and the turns can be quite a bit tighter. But terminating can be very expensive; the best price I have for a commercial fiber tech is $100/end.
  - throwanem 6y ago
    The turns in fiber can be tighter than in Cat 6, really? Granted my experience pulling any kind of cable is most of two decades stale by this point, and my one experience with installing fiber considerably older still - I'm still surprised to hear fiber could be easier to pull.
    - walrus01 6y ago
      g.657.b3 ultra bend loss insensitive can take a remarkable amount of abuse, it's designed for difficult FTTH installs. https://www.youtube.com/watch?v=UBt00CVvMBA https://www.youtube.com/watch?v=UBt00CVvMBA
      - throwanem 6y ago
        Thanks for linking the video. That is wild to me - if I'd tried that staple gun trick with Cat 6 on a worksite, my boss would've kicked my ass all the way back to the office, and rightly so. I'd have never dreamed of seeing fiber that could hold up to it!
  - walrus01 6y ago
    If you're doing a basic project where it makes economic sense and are willing to teach yourself, it takes a $950 fusion splicer and about $300-400 of hand tools/supplies to terminate single mode these days. It's not super hard to learn how to do as an amateur if you can watch some youtube videos and look at reference documents. This particular model which can be found for 900-950 from China is popular for FTTH last mile work: https://toolboom.com/en/fusion-splicer-kit-signalfire-ai-9/ https://toolboom.com/en/fusion-splicer-kit-signalfire-ai-9/ It's not something I'd use to splice a very important cable carrying long haul DWDM circuits, but more than good enough for its purpose. As to whether a house needs fiber to each room? I'm not really sure, at the loop lengths involved, recent cat6 cable has a high chance of working successfully at 2.5GBaseT and 5GBaseT speeds, even if it doesn't qualify and test successfully for 10GBaseT. If you have a really high end 802.11ax 4x4 dual band AP with 2.5 or 5GBaseT interface on it and the switch to support it, in real world use it's unlikely you'll ever get much beyond 1000BaseT speeds to it with real wifi traffic.
  - michaelt 6y ago
    Yes, if you ignore the difficulty of terminating fiber, the more expensive switches, the fact no motherboard, laptop or TV accepts fiber, and the lack of PoE there's really no reason not to use fiber.
    - xbar 6y ago
      Fiber it is!
wffurr 6y ago
100 Mbps ethernet is unimaginable luxury for people struggling with congested 2.4 GHz networks. Let’s not let the perfect be the enemy of the good. The vast majority of homes, at least in the US, are nowhere near 1 Gbps internet service, and of those, many are coping with poor WiFi.
- seniorivn 6y ago
  imagine calling a country with majority of homes nowhere near 1 Gbps a first world country in 2021... P.S. us citizens need to acknowledge that the reason they don't have such a luxury is a lack of competition in ISP space, make that issue nonpartisan, discuss it with everyone who cares, find out why other countries are in a better state? what are the structural differences? are there laws preventing competition? etc
maxsilver 6y ago
> If you install in-wall cat6 it’ll probably be obsolete by the end of the decade. Would it? "Regular" Cat5e has been around for 20 years or so now, and even at "just" 100mbps to 1 gigabit, would be a ridiculously huge upgrade for any home I've ever been inside of, and is still reliably faster than any WiFi device invented thus far. Unless you have some really esoteric requirements, most homes could just run CAT6a to every room without thinking about it, and sleep soundly knowing they'll reliably get 2 to 10 gigabit (depending on distance) into each room for the next 30 years or so.
nonameiguess 6y ago
At a certain points (generally 6 gbs), you're getting network that is faster than internal SATA, so unless you're using pure NVME storage, your network is faster than your filesystem, which means your network isn't limiting you. That happens well before we max out speeds that can be accomplished with twisted pair copper. Even with pure NVME, PCIe as far as I know still maxes out under 100 gbps and no actual storage devices support read or write at full PCIe speed anyway. I mean, not consumer grade at least.
Vvector 6y ago
If you install in-wall cat6 it’ll probably be obsolete by the end of the decade. No. Most devices will still work just fine on Cat 5e for another 10-20 years, running at 1gbps. Cat 6a, running at 10gbps, will be fine for residential for another 25-30 years. Putting fiber in the walls is an expensive overkill. If you are worried about future proofing, just install conduit with cat 6a.
015a 6y ago
Fiber can solidly hit 10x the price of installation over CAT6a/7, between the more expensive cabling, ethernet conversion on the room terminals (ok, maybe you have one computer with a PCI-E fiber adapter? nothing else does), and the networking switch in a closet/basement (price a switch with more than 4 SFP+/fiber channels. they approach five figures. so, you'll probably have to convert back to ethernet at the source as well). And the benefit is tenuous. CAT6a/7 can hit 10Gbps, as long as the run length isn't insane. Even the 11th gen Intel NUCs ship with 2.5Gbps ethernet LAN ports, on-board; outfitting your endpoint devices to breach 1Gbps is far cheaper, especially considering most won't ever breach 1Gbps due to hardware limitations (PS5/Xbox? Ikea Tradfri Gateway?). Even in the "local network upload/download" case; you've got a server, and you want 40Gbps to that server. Building a file server capable of sustained 40Gbps transfer rates is... insane. Its not easy, nor cheap. It requires multiple PCI-E attached NVME drives in RAID-0, on the latest-gen TR/EPYC platform (for their PCI-E lane count, maybe Xeon is good enough nowadays as well). In 2021, this is still in the realm of "something Linus Tech Tips does as a showcase, with all the parts donated by advertisers, and it still sucks to get going because Linux itself becomes the bottleneck". Remember: A Samsung 980 Pro NVME Gen4 ($200/TB) can sustain somewhere around 6Gbps read; you'd need 6-8 of them, in a single RAID-0. And, realistically, you'd want 12-16 of them in RAID 0+1. A server capable of this is easily in the mid-five-figures. (and, fun fact, even after you build a server capable of this; Windows Explorer literally cannot transfer files that fast. you have to use a third party program.) If you're a millionaire outfitting your mansion, then sure, maybe fiber makes sense (due to both upfront cost and length of the cable runs, where sustaining 10Gbps on CAT6a/7 is more tenuous). But I think the assertion that Cat6a/7 will be "obsolete" by 2030 is pretty crazy. Yes, technology will continue to get cheaper and more accessible, and I do think we'll see more fiber providers in tier 1 and 2 metro US areas offer wider 2Gbs and 5Gbs connections, but CAT6a/7 is perfectly capable of saturating this. Just ask yourself: Do you really predict that the PlayStation 6, maybe 2028, will have a duplex fiber port on the back, instead of ethernet? Its 2021, and Microsoft's Xbox download servers can't even download game data at gigabit speeds; they rarely breach 250-500Mbps. Given the niche that fiber lives in, even taking the position that "its just dual-channel light, one up, one down, nothing can travel faster than light, its the perfect future-proof tech" is tenuous. Whos to say that, in the next twenty years, a consumer standard for fiber is developed which runs quadplex (2 up 2 down)? Or simplex (because its "good enough")? Or the connectors are totally different (which would be the easiest to switch because it may not need new cable runs. maybe). Oh, also: PoE! PoE is freakin fantastic for prosumer setups. and only available on copper. You can run copper to areas around your house where you want security cameras or other smart devices, and not have to worry about also running power.

MikeKusold 6y ago

Cat6a supports 10gbps. That should be good enough for the foreseeable future. Fiber is significantly more difficult to pull through the wall since it is more fragile.

yardie 6y ago
Fiber is not significantly harder than Cat6 and the turns can be quite a bit tighter. But terminating can be very expensive; the best price I have for a commercial fiber tech is $100/end.
- throwanem 6y ago
  The turns in fiber can be tighter than in Cat 6, really? Granted my experience pulling any kind of cable is most of two decades stale by this point, and my one experience with installing fiber considerably older still - I'm still surprised to hear fiber could be easier to pull.
  - walrus01 6y ago
    g.657.b3 ultra bend loss insensitive can take a remarkable amount of abuse, it's designed for difficult FTTH installs. https://www.youtube.com/watch?v=UBt00CVvMBA https://www.youtube.com/watch?v=UBt00CVvMBA
    - throwanem 6y ago
      Thanks for linking the video. That is wild to me - if I'd tried that staple gun trick with Cat 6 on a worksite, my boss would've kicked my ass all the way back to the office, and rightly so. I'd have never dreamed of seeing fiber that could hold up to it!
- walrus01 6y ago
  If you're doing a basic project where it makes economic sense and are willing to teach yourself, it takes a $950 fusion splicer and about $300-400 of hand tools/supplies to terminate single mode these days. It's not super hard to learn how to do as an amateur if you can watch some youtube videos and look at reference documents. This particular model which can be found for 900-950 from China is popular for FTTH last mile work: https://toolboom.com/en/fusion-splicer-kit-signalfire-ai-9/ https://toolboom.com/en/fusion-splicer-kit-signalfire-ai-9/ It's not something I'd use to splice a very important cable carrying long haul DWDM circuits, but more than good enough for its purpose. As to whether a house needs fiber to each room? I'm not really sure, at the loop lengths involved, recent cat6 cable has a high chance of working successfully at 2.5GBaseT and 5GBaseT speeds, even if it doesn't qualify and test successfully for 10GBaseT. If you have a really high end 802.11ax 4x4 dual band AP with 2.5 or 5GBaseT interface on it and the switch to support it, in real world use it's unlikely you'll ever get much beyond 1000BaseT speeds to it with real wifi traffic.
- michaelt 6y ago
  Yes, if you ignore the difficulty of terminating fiber, the more expensive switches, the fact no motherboard, laptop or TV accepts fiber, and the lack of PoE there's really no reason not to use fiber.
  - xbar 6y ago
    Fiber it is!

wffurr 6y ago

100 Mbps ethernet is unimaginable luxury for people struggling with congested 2.4 GHz networks. Let’s not let the perfect be the enemy of the good. The vast majority of homes, at least in the US, are nowhere near 1 Gbps internet service, and of those, many are coping with poor WiFi.

seniorivn 6y ago
imagine calling a country with majority of homes nowhere near 1 Gbps a first world country in 2021... P.S. us citizens need to acknowledge that the reason they don't have such a luxury is a lack of competition in ISP space, make that issue nonpartisan, discuss it with everyone who cares, find out why other countries are in a better state? what are the structural differences? are there laws preventing competition? etc

maxsilver 6y ago

> If you install in-wall cat6 it’ll probably be obsolete by the end of the decade. Would it? "Regular" Cat5e has been around for 20 years or so now, and even at "just" 100mbps to 1 gigabit, would be a ridiculously huge upgrade for any home I've ever been inside of, and is still reliably faster than any WiFi device invented thus far. Unless you have some really esoteric requirements, most homes could just run CAT6a to every room without thinking about it, and sleep soundly knowing they'll reliably get 2 to 10 gigabit (depending on distance) into each room for the next 30 years or so.

nonameiguess 6y ago

At a certain points (generally 6 gbs), you're getting network that is faster than internal SATA, so unless you're using pure NVME storage, your network is faster than your filesystem, which means your network isn't limiting you. That happens well before we max out speeds that can be accomplished with twisted pair copper. Even with pure NVME, PCIe as far as I know still maxes out under 100 gbps and no actual storage devices support read or write at full PCIe speed anyway. I mean, not consumer grade at least.

Vvector 6y ago

If you install in-wall cat6 it’ll probably be obsolete by the end of the decade. No. Most devices will still work just fine on Cat 5e for another 10-20 years, running at 1gbps. Cat 6a, running at 10gbps, will be fine for residential for another 25-30 years. Putting fiber in the walls is an expensive overkill. If you are worried about future proofing, just install conduit with cat 6a.

015a 6y ago

Fiber can solidly hit 10x the price of installation over CAT6a/7, between the more expensive cabling, ethernet conversion on the room terminals (ok, maybe you have one computer with a PCI-E fiber adapter? nothing else does), and the networking switch in a closet/basement (price a switch with more than 4 SFP+/fiber channels. they approach five figures. so, you'll probably have to convert back to ethernet at the source as well). And the benefit is tenuous. CAT6a/7 can hit 10Gbps, as long as the run length isn't insane. Even the 11th gen Intel NUCs ship with 2.5Gbps ethernet LAN ports, on-board; outfitting your endpoint devices to breach 1Gbps is far cheaper, especially considering most won't ever breach 1Gbps due to hardware limitations (PS5/Xbox? Ikea Tradfri Gateway?). Even in the "local network upload/download" case; you've got a server, and you want 40Gbps to that server. Building a file server capable of sustained 40Gbps transfer rates is... insane. Its not easy, nor cheap. It requires multiple PCI-E attached NVME drives in RAID-0, on the latest-gen TR/EPYC platform (for their PCI-E lane count, maybe Xeon is good enough nowadays as well). In 2021, this is still in the realm of "something Linus Tech Tips does as a showcase, with all the parts donated by advertisers, and it still sucks to get going because Linux itself becomes the bottleneck". Remember: A Samsung 980 Pro NVME Gen4 ($200/TB) can sustain somewhere around 6Gbps read; you'd need 6-8 of them, in a single RAID-0. And, realistically, you'd want 12-16 of them in RAID 0+1. A server capable of this is easily in the mid-five-figures. (and, fun fact, even after you build a server capable of this; Windows Explorer literally cannot transfer files that fast. you have to use a third party program.) If you're a millionaire outfitting your mansion, then sure, maybe fiber makes sense (due to both upfront cost and length of the cable runs, where sustaining 10Gbps on CAT6a/7 is more tenuous). But I think the assertion that Cat6a/7 will be "obsolete" by 2030 is pretty crazy. Yes, technology will continue to get cheaper and more accessible, and I do think we'll see more fiber providers in tier 1 and 2 metro US areas offer wider 2Gbs and 5Gbs connections, but CAT6a/7 is perfectly capable of saturating this. Just ask yourself: Do you really predict that the PlayStation 6, maybe 2028, will have a duplex fiber port on the back, instead of ethernet? Its 2021, and Microsoft's Xbox download servers can't even download game data at gigabit speeds; they rarely breach 250-500Mbps. Given the niche that fiber lives in, even taking the position that "its just dual-channel light, one up, one down, nothing can travel faster than light, its the perfect future-proof tech" is tenuous. Whos to say that, in the next twenty years, a consumer standard for fiber is developed which runs quadplex (2 up 2 down)? Or simplex (because its "good enough")? Or the connectors are totally different (which would be the easiest to switch because it may not need new cable runs. maybe). Oh, also: PoE! PoE is freakin fantastic for prosumer setups. and only available on copper. You can run copper to areas around your house where you want security cameras or other smart devices, and not have to worry about also running power.