Hacker News Reader

Comment by jassyr | Hacker News Reader

What is the use case today for 224 GB/s wifi? 8k+ video? Genuinely curious.

"LiFi" does seem pretty niche. Anywhere you're okay with a 10-15m range, have full line-of-sight, somehow can't run a wire or fiber run, and don't want regular WiFi for whatever reason.

Only thing I can think of is full fat uncompressed VR streaming to a wireless headset.

Could be useful for modular autonomous robotic devices that act as a single organism, like ants or cells. They would each do their own thing but communicate using this (or something else fast, short-range, and error-resilient) so as to work towards a unified goal without a lot of redundancy.

I can think of two uses:

Open floor plan offices. (Sigh.)

Datacenters or server rooms. This could give quite nice data rates within a rack.

Any place you have this in a point-to-point rig, you could have fiber cables guiding the light and creating a private collision domain, improving overall bandwidth, reliability and security.

The cost of datacenters is not currently constrained by fiber.

Fiber is not free to run. In an open floor plan office with 20 workstations on an overgrown table, you could run 20 fiber pairs (or cat5e cables or cat6a cables), and you could terminate them and connect them, or you could set up one LiFi access point and 20 client devices. The latter is a lot less cable bundle and a lot less labor.

This can be done now with wifi. The question I answered was about datacenters.

There is wifi in datacenters -- usually for the benefit of visiting techs. Not for inside a rack.

Incidentally, if asked to set up 20 workstations on an overgrown table, and it's not a very temporary thing while the office is in turmoil, I recommend finding a different employer.

That is like saying, "handing a 2tb hard drive to someone is 1500x times the download speed of wifi".

"Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway."

https://archive.org/details/computernetworks02tane/page/56/m...

Maybe this is the solution to my niche issue.

Better network speeds in a room surrounded by Faraday cages with MR scanners in them.

Optical is an option but a right pita to run without holing the cage or circumnavigating it.

Niche also means it will be expensive.

For you home network, they're are likely few or no use card to warrant a new tech like this.

But for offices, schools, conference halls, and other place with many people together it could be useful as a very fast zero-wire solution. Even then it'll be niche: most office workers or conference attendees don't need 100Mbit let alone more than 1Gbit. But some will: maybe said office is full of video editors and such.

Remember that the bandwidth is a shared resource like WiFi: that isn't "up to 224Gbit/s for every device in the room" it is "up to 224Gbit/s for every device in total". And that "up to" value is for ideal circumstances: there will be some environmental interference (though LOS limitations will work in this techs favour there) and, more significantly, the more devices you have in a given collision domain the more they will interfere slightly with each other too because they can't share the group resource with complete efficiency (and this difficulty grows exponentially after a certain threshold). Get a couple of machines transferring arbitrary data as far as they can via WiFi, take the total of transfer speeds they are getting and now add more machines doing the same. The total might scale well for an irritation or two of this test, but there will be a pint when it significantly won't and the total bandwidth will actually fall (and latency will shoot up).

The main push for 5G mobile networks wasn't better max bandwidth per device, but better performance (latency and throughout) for every device when you have a lot of them sharing the same collision domain. The headline figure of 224G bit is (while not at all dishonest) just that: a figure to gain interest via headlines. Those who actually have a need or want for such tech will be looking much deeper into it than that.

home networks in high density condo towers where WiFi network channels are often already occupied by neighbors on the same and adjacent floors. The 5GHz band might have a "limited distance" but it's still enough to interfere between neighbors.

Good point. I was thinking about the bandwidth within the local network partition and neglected to consider that the line-of-sight nature is helpful in compact housing because it significantly limits what can interfere with that local partition.

high fidelity AR/VR without cabling

It's "Data transmission tests have reached speeds of up to 224 Gbps" which might translate to something like 50 Gbps, or 6 GBytes/s of reliable goodput. Divide by the number of users. Wireless high res VR/AR gear might indeed be one application but also many normal "access my data over the network" applications on the laptop currently need need this class of bandwidth. Note that on normal ethernet you get the 40-100 Gbit/s BW per end user, since it's wired & switched. But this will be shared media.

Synchronizing LLMs, streaming AR textures, downloading Linux ISOs. If you give someone bandwidth, they will fill it.

The use case can be found here: https://lifi.co/lifi-applications/

I can only think of useful for conference where everywhere there's light.

10x 22.4GB/s download from CDN or 1x 112GB/s Li-Fi to Li-Fi local transfer before inevitable massive derating

Niche: video editing, interacting with AI/ML training clusters, ...?

iPhones that still have USB 2.0 ports /s

Wireless hdmi, 4k webcams would be nice.

You can do 4K with way less that a hundred megabits, some PoE cameras don’t even put a gigabit port anymore since it doesn’t need it.

Only with compression. Uncompressed 4k60 is 12.54 Gbit/s, so high data rates are definitely useful for wireless HDMI.

4K can mean a lot of things but some of the low bitrate iterations look fairly average.

Yeah but OP mentioned “webcams”

Reply on news.ycombinator.com