Definitely agree that it’s generally not needed (>1Gbps) but game updates these days run 100GiB+ in some cases (particularly for AAA titles) and it’s nice to have a little more speed sometimes. Individual clients are usually 80MHz 2x2 so best case will get ~1.2Gbps [2], but the radio can handle a couple clients doing things at once (MU-MIMO).
For external connectivity, Comcast is now selling 1.2Gbps services in many US markets (they also overprovision the profile so speedtest.net will often see 1.4Gbps), and consequently cable modems have had 2.5GE ports for a couple years now. When combined with APs that have >1Gbps uplinks this can actually get used! Some markers can now get Comcast with 2000/200 (down/up) [3], which is a huge improvement over 1200/35 — the previous maximum.
So, it’s nice to have APs with more uplink, and it’s nice to have >1Gbps Comcast, but I’d definitely not call it mainstream yet.
[1] https://www.arubanetworks.com/assets/ds/DS_AP550Series.pdf
[3] https://corporate.comcast.com/press/releases/comcast-increas...
Edit: Just checked and for 10 G it would be about 10 euros and for 2.5 G - 9 euros, that's a very good upgrade, probably they assume that most users won't take advantage of the full bandwidth.
I definitely don't have any way to take advantage of 10g, that's for sure. I guess WiFi 6 might be able to make some use of 2.5G. I'd upgrade for 1 euro more if I was you.
https://www.purexbox.com/news/2022/11/modern-warfare-2s-late...
Lots of prebuilt PC systems are coming with 2.5GE NICs - “build your own” and it feels like most motherboards are now 2.5GE? Upgrading existing to 2.5GE is pretty cheap (~$30 USD), even via USB adapters [3].
To me, I think it’s fine with NBase-T that the switches be 10GE and the clients negotiate at 2.5GE or 5GE, although I’d note that both Mac Mini, iMac and Mac Studio can now be OEM specified with 10GE NIC for I think $100 additional.
[1] https://dl.ubnt.com/ds/usw-flex-xg_ds
[2] https://dl.ubnt.com/ds/us-xg-6poe_ds
[3] Plugable 2.5G -https://a.co/d/hjV8LTd
€300 for 5 is a fortune for me :(
There are significant tradeoffs that you may not tolerate:
(1) They are very noisy (vs. the Flex XG which is fanless) so if your "Tech" is anywhere near your living space, this may be unworkable;
(2) They consume lots of power and also don't do 802.3{af,at,bt} for PoE either, you can solve PoE using an injector (for a few ports), a midspan (for more ports) but note that injectors/midspans usually don't support NBase-T and if you want the combination of 2.5/5/10GE and PoE you might be back to buying a specific switch which can do that;
(3) Learning curve is steeper than unmanaged or most semi-managed switches and you likely need to be comfortable with serial consoles to get them configured.
With all those and likely more tradeoffs considered, they are ~$10/port!
If you're not going to get direct-attach cables, then check out the stuff that 10gtek sells. In the 6->8+ months that I've been using them, I've not had any trouble with their short-range multi-mode SFP+ modules. Also, you can buy them for a little more than 10USD per module if you buy them in 2->10-packs from Newegg.
(And yeah, 10gtek is totally one of those "Direct from Shenzhen" companies that one might worry about buying from. I figured that 10gbit transceivers have been around for quite a while now, so there's no reason they should cost 80USD per, rolled the dice on two separate cheapo 10g transceiver manufacturers, and got lucky when the second (10gtek) turned out to ship a solid product that, like, actually worked and seems to work well.)
A 1GbE WiFi signal only needs to have 1GbE throughput.
What home users need is a 10GbE hub. 10GbE connections between devices but really only one at a time.
Switches are overbuilt for home user purposes. I don't think anyone will notice that Alice takes up the channel and that Bob has to wait his turn.
In fact, that's how WiFi works and most people don't even notice.
It needs 16 Gb switching capacity; it is duplex.
> What home users need is a 10GbE hub. 10GbE connections between devices but really only one at a time. Switches are overbuilt for home user purposes.
Not really. While it is true, that most users do not have active multiple devices at the same time, they would pay for this in energy usage. Ethernet, which is what most people would end up with, has pretty high consumption at 10 GbE, to the tune 3-5W per port (times 2, because you need two ends for each cable). It is better to keep them idle when possible; that means switching.
I don't think this is the correct terminology. "Switching capacity", as far as I can tell, is the simple addition of the speeds of all the ports. So 16 Gbps switching capacity isn't something it needs, it's something it has simply from the nature of having those 8x 1 GbE full-duplex ports.
What we're talking about here that we'd need to not bottleneck, would be called either "non-blocking throughput" or "backplane bandwidth".
And even on higher end switches, the switching capacity and backplane bandwidth are usually not equal. I'd be surprised if the same wasn't true for consumer switches (though maybe it is since there's far fewer ports on consumer switches than enterprise switches).
(There's also a packets per second metric that determines how much the switch can handle, but I'm assuming that's only going to be a bottleneck if your packet sizes are too low)
Most enterprise gear on the switching side is also non blocking, but there are exceptions of course (Cisco sucks).
Perhaps so.
Still, how to switch? If the "switch internals" is just a token-ring + ARP table (IE: each port remembers the mac addresses its serving, and token-rings the data around), the switch will have ~10GbE for all-ports, but only a max bandwidth of 10GbE before it starts losing packets. (Ex: Alice <-> Bob <-> Charlie <-> Dave with only 10GbE each link. If Alice talks to Charlie and Bob talks to Dave, then the Bob -> Charlie needs 20GbE, so half the packets will get dropped)
This isn't a full switch classic.
Maybe a "Hub classic" isn't enough, since that keeps the ports needlessly active as you noted. But I don't think the full 80Gbps bidirectional is needed for an 8-port home-consumer (or even many prosumer) environments.
The full 10Gbps switches all get quite hot, probably due to drawing a ton of power. I'm thinking that some simplified architecture is needed to reduce power (by trading off and reducing bandwidth)
Is it noisy?
Will say, as far as I'm concerned not seeing a need to upgrade any time soon for my own needs. Short of another project where I regularly have to xfer a couple hundred gigs of files, where my internet provider is 1gb down, 40mb up.
If you can wait for 7, then it probably makes sense to just get the latest and greatest.
But at $WORK we're moving into a new office in 2023, and given networking gear lead times, we have to order what's available now (6 and 6E) to make sure we get the stuff in time for move-in.
Even with 'only' 6(E), you still get OFDMA, which AFAICT really helps with spectrum efficiency in high-density environments. 6GHz with 6E is also a good thing.
Most Wi-Fi heads are more excited about the use of MLO for roaming improvements in enterprise networks though. The ability of the client to completely associate to another AP on another radio before officially roaming the traffic to it could be interesting.
Overall MLO is going to be an interesting rollout. Just like many had fears of MU-MIMO sounding great on paper but being near useless in the real world on real devices with real drivers it's possible this ends up as more a "you can if you precisely set up ideal conditions and select the devices to see this work really well" for most of the modes of operation. I think the roaming one is relatively safe and the worst case is you just fall back to normal roaming. The more complicated ones where there are a lot of combos about how a client can associate to multiple APs and actively do things feels like a complicated coordination nightmare and that might take a really long time, if not forever, for client devices to get good at.
I am going to disagree a bit and add a few things.
WiFi 6 is basically the 4G version from IEEE with OFDMA. And OFDMA is the biggest reason why 802.11ax was delayed by 2+ years. As anyone jumping from 3G to 4G will immediately notice the difference, lower latency, high bandwidth all due to better frequency allocation and efficiency. Jumping from WiFi 5 to WiFi 6 will see the same immediate benefits.
WiFi 7, or 802.11be is the 5G version, an evolution and improved version of 4G or Wifi 6. But the difference wont be as dramatic as WiFi 5 moving to WiFi 6, given most of the bottleneck we have today are total WiFI capacity constrain, not singular user use case.
And while 320Mhz and 4096 QAM are mostly given. ( They are on some of the WiFi 6E equipment already if I remember correctly ). A lot of the WiFI 7 features, such as Multi Link Operation and Multi RU tends to go missing when they first launch. i.e While they are part of the 802.11be spec, they are not (yet) guarantee to be shipping it. And even if they do you are likely end up with something like the fist gen of OFDMA and MU-MIMO, which requires lots of tuning and fixing and even 2nd generation silicon.
But a proper WiFi 7 is going to fill up the needs for 99% of consumer use case in the foreseeable futures. It sort of hit the end curve of Power law.
One thing I do hope will happen because marketing will have a hard time explaining it ( or at least I will make enough noise on the internet to demand a marketing explanation ) will be WiFi is now faster than the Wired Ethernet. Which will hopefully force them to include at least 2.5Gbps port on their routers. I dont think 10Gbps Ethernet will ever be cheap enough for consumer market, but I hope may be at least 5Gbps / Nbase-T / 802.11bz? You only need CAT6 cable for 5Gbps. ( You could even do it with CAT5e in short distances )
https://en.wikipedia.org/wiki/IEEE_802.11be
The 6Gghz from Wifi 6E will be included but critically Wifi 7 devices can use difference frequencies at the same time alongside the 320Mhz channel widths and 4096 QAM. It will be a bigger jump than Wifi 6 and 6E and might finally start to see genuinely greater than 1gbps performance in practice.