Not convinced by the use case section. Very few servers will allow you a full 25gbit download, let alone anything more than a 1gbit (and often less). And if you own the server on the other end, that sort of bandwidth comes at a cost.
I think beyond 1gbit, the benefit become super marginal and the hardware expensive.
I have the same internet provider and package. While 25Gbps is indeed basically unattainable to anything other than Init7's speedtest server, it's easy to exceed 1Gbps.
Software and driver updates exceed 1Gbps all the time, as do game updates/downloads through Steam.
Piracy also works really well. Downloading copyrighted media is perfectly legal in Switzerland and I was able to get ~7Gbps real-world speeds from Usenet without too much hassle.
It's also really handy for things like backups. As of writing, bandwidth to a Hetzner cloud server is ~5Gbps up/down with iPerf3.
I do agree that 25Gbps is more overkill/bragging rights than it is real utility but I think 10Gbps is an easy sell.
Keep in mind that there is no change at all to the monthly price of your internet by choosing these higher speeds. 10Gbps has the same monthly and setup costs as 1Gbps, you only have to pay a bit extra for hardware capable of dealing with 10Gbps, which is pretty affordable.
Also with init7, very happy with my 1Gbps line. I'd probably even take a 300Mbps if it was half the price. Even with gigabit LAN I don't saturate the line more than a few seconds a day on things like updates, and anything from my RPi NAS isn't going to saturate it anyways.
Please note too, piracy is not legal in Switzerland. What is illegal is the spying and tracking which would be necessary to build a case and prosecute pirates. Technically you could still self-incriminate if you documented all of your piracy, with proof, and published it online.
Downloading is legal in Switzerland. For a nice authoritative source, here's the Swiss Federal Institute for Intellectual Property (part of the Department of Justice and Police)[0]: "Downloading copyright-protected works for private use is permitted in Switzerland (Art. 19 CopA)."
The mentioned legal basis is [1].
Note that this applies strictly to downloading though. Participating in a torrent swarm (where uploading is also happening) is not permitted. That's where the technicality you mention comes in: until recently, it was illegal to monitor internet users for copyright enforcement purposes, which meant it was illegal to monitor a torrent swarm, which meant you could somewhat safely seed torrents, despite it being illegal.
Even where the ports are available, 25 Gbit from a single address is well into the realms of looking like attack traffic in a wide variety of scenarios.
Past even 500 Mbit I'm way more interested in latency considerations than raw bandwidth, and practical matters like how to use that bandwidth from my laptop (good luck doing 500 mbit wireless reliably, never mind 25 Gbit!)
Most routers and wifi adapters are crap. Buyers do rarely go beyond "wifi 5" or "wifi 6", and do not realize that there's much more.
The older Apple Macbook Pros (pre-2019) came with 3x3 MIMO ac adapters. If you had capable AP on the other side, you could reliably do gigabit with them. The newer ones have only 2x2 MIMO, just like the rest of the laptop market, so you will get only 600-700 Mbps (out of theoretical 866 Mbps).
If you are getting 500 Mbps and there's not a concrete wall between your client and the AP, something is quite wrong. Misconfigured AP, your client cannot do multiple streams (yes, there were adapters like that sold on the market), or just older/pre-ac AP or client.
Basically, how fast you can transfer over wireless is determined by:
1) how effectively you can pack data into a channel, determined by the modulation scheme.
In your wifi properties, you can see the modulation scheme used as 'MCS index'. You will see an integer between 1 and 11, the higher, the better. This is negotiated between client and AP, depending on the signal strength (antennas, number of walls/wall material between them, etc). With wifi 5, it is realistic to have MCS 8 or 9, if your client and AP are in the same room.
2) how wide is your channel - the "basic" is 20 MHz wide, newer standards can combine multiples, into 40, 80, 160 MHz wide one. Wifi 7 will bring 320 MHz wide channels.
2,4 GHz band has only 3 non-overlapping 20 MHz wide channels available. If you use more than one, your neighbors will hate you. The situation is better with 5 GHz band, unless you run into the DFS (radar detection), then you can have intermittent outages. This is improved by Wifi 6e which brings more channels in the 6 GHz band, but the client and AP support is not there yet. Currently, running an 80 MHz wide channel bellow DFS frequencies is probably the best.
3) how many many simultaneous streams you are able to transmit/receive - MIMO (one client using multiple streams), MU-MIMO (multiple users using multiple streams, so they do not stomp on each other and sidestep next point).
Most clients are capable of 2x2 MIMO, older Apple MBPs (pre-2019, -ac based) are capable of 3x3. Most APs are also 2x2. The nicer APs can do 4x4 MIMO, some gamer APs by companies like ASUS can do 8x8.
4) how effectively can multiple clients share the same channel.
Here the problem is, that once an older client connects, it is downgrade for all clients, they must be compatible. If you can, keep your 2,4 GHz network -n and higher, and 5 GHz -ac or higher.
Once you will be able to have a band -ax only, you could theoretically be able to use OFDMA, which does improve the performance for multiple clients.
You can have a look at mcsindex.com: pick the modulation scheme (row), number of MIMO streams (group of rows), how wide is your channel (column) and OFDM/OFDMA (group of columns) and there you can see the theoretical bandwidth. Note, that it is a shared bandwidth, for all clients.
There are some minor points:
5) if you have IPTV or another application that uses multicast, do not run it over wifi! Multicast by definition does not have ACK, so what can AP do to ensure every subscribed client gets the data? Slow everything down, for everyone. So do not let it do it.
6) multiple SSID - possible, but do not overdo it; they also have performance impact (the SSID advertising takes time, at the slowest/most compatible rate allowed; the more of them, the more time it takes). Ubiqiti limits them to 4 per radio.
7) at higher frequencies, you won't have a such range than at lower ones. This is both good and bad thing. The bad, obviously, that your range is lower. The good is, that the range of your neighbors is also lower, so you could use the same channel without disturbing each other.
8) It does not help to crank the tx power to the max. The clients may receive, but when they answer, the answer could never arrive. Clients do not have antennas or tx power like APs do. Use the tx power reasonably; if you have multiple APs in your house, adjust it so clients will let go the weaker signal and connect to the stronger one (which AP connect to is fully managed by clients, unfortunately, and they often prefer to stick to the one they are already connected, even if different one with stronger signal is available. At most, some APs can kick the client once the signal is below threshold you specify).
9) some cheap dual-band APs have only single radio, capable of transmitting at both frequencies. Not at the same time, obviously, they time share. Make sure your APs have separate radios for each band they support.
10) some APs do support roaming (802.11-k/r/v); for home (or WPAx-Personal) that's not that important. It is enough if your SSIDs are named the same and have the same password. Roaming support helps with WPAx-Enterprise, which is much more heavy-weight.
If I was revamping my home setup today, I would pick something from Ubiquiti that can do 4x4 MIMO, probably U6 Pro; if the top speed was not really an issue or there would not be two users trying do download the internets, then probably U6 Lite (I'm running nanoHD today, and even that can utilize the 1 Gbit ethernet uplink to its full capacity). Mikrotik is supposedly also working on Wifi 6, but they have nothing available right now. If you have a better budget, then there are brands like Ruckus. The choice is affected by about how big is your home, what is the disposition, what materials are used, where do you want to place your APs and how you need, are your neighbors noisy (in wifi), what you expect from the wifi and what is your budget.
I have a bandwidth in the low hundreds and it definitely feels excessive to me. But changing from wireless to wired and getting rid of those occasional latency spikes - very noticeable while playing games. A static ip is also something I would want.
> I think beyond 1gbit, the benefit become super marginal and the hardware expensive
Is it expensive though? We spend how much of our lives online. How much do you think this all-spanning life upgrade costs? What is the price of never ever dealing with buffer-bloat? Take a guess.
I calculate it as a one time $356 cost plus labor. You might have paid more for your wifi system. 128 port 25gbe switches are around $20k ($156/port). Transcievers are under $100 and you need one on both ends. For a lot of already deployed fiber, this is a drop in replacement. This is absurdly cheap. Given how cheap this is it's an obvious & enthusiastic heck yes. Who wouldnt throw down $356 right now to get 25Gbe for life?
Cost gets a bit more complicated when it comes to the POPs & their uplink. Subscribers are going to be way oversubscribed even with some fairly expensive 100Gb uplinks. As you get further from an exchange the difficulty grows geometrically (because pops become.further hops away from the ix). Peering needs to be bigger too, as does transit (but ask whether the net volume of traffic grows elastically or not), which has costs. But I think we need to frame this question a bit better, of whether it's "worth" upgrading. Honestly costs are so low it doesnt make sense not to; the rest of the world is just milking us, bilking us, charging what the market will bear, protecting it's profit centers, and this company init7 is doing what makes financial sense for the consumer. Donwe need all that? Maybe maybe not. Should we settle for less? There's almost no financial case when the hardware is so so so very cheap. This tech sounds magical but 25Gbe is not exotic, not extreme technology; "the future is already here, it's just not evenly distributed yet".
A significant part of my life is online, indeed, but 99% of the time, bandwidth is a nonissue. Latency, very important. Bandwidth, not so much. I haven't experienced any bandwidth related internet problem in the past ten years, and that's moving from apartment to apartment, from hotel to hotel, from airport to airport.
Don't get me wrong. I'm very much in favor of this upgrade. I just don't think it's going to be an everyday quality-of-life improvement for most people. It's more about providing a service for people with special needs, and future-proofing the infrastructure.
> 99% of the time, bandwidth is a nonissue. Latency, very important
Indeed. And I'm speaking as someone who downloads most visual media before I watch it, so bandwidth matters to me. But not that much. With 100-200 Mbit/s I am good.
Cookie prompts, newsletter pop-ups, scrolljacking and ads constitutes the majority of wasted time for me, by a long shot. Latency to sites in other parts of the world can cause problems sometimes, since number of round trips can be quite high with TLS neg + progressively loaded content.
It sounds like you're just taking about cost for the ISP to upgrade.
To actually realize your faster speeds, you need to spend thousands of dollars yourself on new switches and NICs. And then, as mentioned, the benefits are marginal. You would have to be streaming 10+ 4K movies at once to even "need" gigabit, let alone 25Gbps.
Wifi 7 is expected to be capable of 30-40Gbps. A dual port nic can be had under $200. Currently low/medium port count equipment has no demand, but perhaps the wifi7 world or pressure like init7 generates can make more visible & obvious the market demand. For anyone setting up today, do what I did: (used byt plentifully available) 18 port 40Gbps infiniband switch for $150, nics for $100.
I semi agree that I dont think we know what this is for. Never ever having buffer bloat is a tempting first ask. Connectivity is more than the sum of throughputs, as your figures imply- there's questions of availability too.
Being able to access each other's systems at near local speeds sounds quite compelling, could help jumpstart post-Big Social computing. You talk about netflix streams, but those are heavily compressed with the best offline encoding on the planet: if i just want to open Steam Remote Play Together & share realtime 4K with a friend, I'd need a lot more throughput since I have much much much less efficiemt encoding. If i wanted Remote Play Together with 3 friends, well, that figures goes up. If my family member also wants to do the same, now we're using a lot or maybe all the throughput & we're starting to have some contested bandwidth, some rising latencies.
The truth is somewhere between. Rationalizing ourselves down to what sounds sensible today, to me, is a cruel trick, is not just path dependency but an ideology that believes only in what we have & can see now, & refuses exploration & trying. To me the world & tech is spiritually fueled by why not thinking, by deciding to opt for the extra thats within reach.
Forgoing a cheap (still less than the price of a nice tv, by far), available one-time purchase option that vaults us into near-local connevtivity caliber with the world is still a lock in my book.
A router with a 25G uplink and a bunch of 10G sfp slots will set you back $600
There's benefits to 25G (certainly when transporting 4K video around which needs more than a 10G nic), whether that's worthwhile for a typical home is likely "no", so unless you've got hundereds of employees in an office it doesn't feel very useful.
The calculation works only if each customer has their own fiber.
Unfortunately for most (consumer) FTTH deployments, that's not the case. Most of them are GPON, where the initial deployment was more effective, as up to 64 of your customers share single fiber, but then that means all of them have to upgrade all 64 of them at the same time, you cannot do them one at the time (see also the speed of XG-PON upgrades).
Additionally, many providers forced use of their CPEs. If you can send out SFP module and the customers can put it into whatever they want, it is much simpler, as replacing CPEs for all the customers on that fiber.
This +10. For the vast majority of users, having 25Gbit/sec home fiber is like having a 250 MPH-capable supercar. Big, bad bragging rights. Token-at-best use cases.
5 people in my house, each using 5Gbit/s, sounds far more like the grandkids playing a "mine's faster than yours" game than like any "real world" usage scenario. Similar for 10 people each using 2.5Gbit/sec, or 25 people each using 1Gbit/sec, or ...
If you run a large scale Plex server I could see the 25Gbps coming in handy. Most 4K movies (Native Blu-ray rips, not the low bitrate renditions you get on streaming services) run about 50-100Mbps. That would let you stream to 200+ people at once, and have overhead for other things.
You forgetting the case where one hosts their own servers in the basement. Assuming that the connection is symmetric and offers static IP (which it does in my case but "only" for 1GBit/s).
Some time ago PC has revolutionized the world by giving access to computer power to general population. This has unleashed a tidal wave of creativity and business.
Giving the ability to host own servers to everyone can open up endless opportunities as well.
I host some of my own servers and benefit from it greatly.
Yeah, I think more home connections going symmetric could open up big personal computing applications that haven’t been super practical before, and make a market for easy-to-use personal home servers. Even just personal media streaming while traveling can benefit greatly from upload being boosted past the 10mbit up that’s somewhat common among cable ISPs. Having a symmetric connection also removes one of the mental constraints that we have at the back of our minds.
It’s my biggest hope for internet re-decentralization.
> Very few servers will allow you a full 25gbit download
As TFA mentions, same was the case with 1 gbit. I can confirm, I'm on the same ISP and was "early" to have 1gbit/s. Lots of servers still only had 100mbit/s links. These days I have zero issues saturating my 1gbit link.
Give it time, now that it starts rolling out, costs will come down and server links will be upgraded. The usual early adopter stuff.
Agreed. A 25gbps download will fill an 8TB HDD in 47 minutes, so how much are you going to actually utilize that full bandwidth? Not that any webhost will ever give you even remotely that much bandwidth. I'm on 1gbps fiber and I've never saturated my connected. That would require 40 simultaneous 4K streams going at the same time in my house.
I'm wondering: is my storage fast enough to write at 25 Gb/s?
SATA drives max at 6 Gb/s, and probably a bit worse than that in practice.
M.2 NVMe SSDs can reach 25+ Gb/s on sequential writes. Maybe I could overclock these to squeeze a bit more out too. Although, an 8TB M.2 drive will cost over $1k for now.
I suppose there's also RAID. Although, I'm not sure what other limits/gotchas I'd run into there.
> But with 25Gbps you can just have multiple connections open without any of them ever affecting any other.
Realistically, this is pretty close to true with 2gbps symmetric, too. My provider seems to give me 2.2gbps in practice.
PS5 downloads/updates are 500-600mbps in practice. Steam is 1.5gbps or so. Most other things-- streaming, video calling, etc, are under 40mbps. So, you know-- if I kick off a PS5 download, and a steam update, and my kids are streaming and video calling... And my machines are backing up to the cloud at 1.5gbps (other direction)... and I decide to do a big apt-get update on a machine, maybe my steam update completes a couple seconds later.
Of course, I want even faster... but I'm hard pressed to say what would be better.
Multiple 4K streams become possible. E.g. when the whole family comes over for the holiday we can all ignore eachother in favor of 4K video streams (which have better resolution, color, and cinematography than the real world ;) )
There aren't many (any?) services that are offering Bluray quality 4K HDR streams but even those are only 140 mbps. I would think 1 gigabit down should cover most families handily, especially when you consider most streaming services are more like 20 mbps max.
The main argument I see is "if you build it they will come" i.e. we won't see higher bitrates until connections like these are more common, but for now gigabit has even fairly extreme use cases for video streaming pretty well covered.
why stream on demand if we have the bandwidth to just send the whole video in the first 100ms? this might actually save power; instead of back & forth back & forth with the service, we can transmitnthe whole thing & be done, the server can now go serve other people.
I can't find it right now but I remember reading a paper by Cerf? (I'm not great with names) that detailed such a concept. If I remember it I will edit and post a link.
It's not 1998 any more, you've got multiple TCP streams, things like QUIC and other UDP based protocols, and of course good old fashioned window scaling which will go upto an 8gbit window, so as long as you've got an rtt under 300ms you'll be fine at 25gbit.
You'll encounter other limitations, like processing 2 million 1500 byte packets a second, which is possible but you need to make sure you have the grunt to process and the right types of options in the kernel
I think beyond 1gbit, the benefit become super marginal and the hardware expensive.