Researchers reach throughput of 44.2Tbps over standard fiber optic cables

Published by

Click here to post a comment for Researchers reach throughput of 44.2Tbps over standard fiber optic cables on our message forum
data/avatar/default/avatar01.webp
i wonder how they benchmarked it. 44Tbps is around 6 tera Bytes per second. How do you put up a device that can send and receive 6 tera per second, network cable excluded? Even if you use just a 1ms of transmission and transferred only 6giga byte, you need a super fast buffer. Does anyone know?
https://forums.guru3d.com/data/avatars/m/254/254132.jpg
asturur:

i wonder how they benchmarked it. 44Tbps is around 6 tera Bytes per second. How do you put up a device that can send and receive 6 tera per second, network cable excluded? Even if you use just a 1ms of transmission and transferred only 6giga byte, you need a super fast buffer. Does anyone know?
This might help: https://www.nature.com/articles/s41467-020-16265-x
data/avatar/default/avatar26.webp
Even more interesting is the question of how do you do error checking of that much data without introducing latency?
data/avatar/default/avatar28.webp
The current Australian government will bury the technology unless they are able to guarantee it won't harm Foxtels operations in Australia... in which case they'll sell it off to the chinese
https://forums.guru3d.com/data/avatars/m/273/273678.jpg
ETAxDOA:

The current Australian government will bury the technology
this will be used for international links well before domestic networks.
https://forums.guru3d.com/data/avatars/m/279/279919.jpg
So imagine you just click on the download button in Steam and it morphs into the 'play' button instantly. That's .... massively mindblowing!
https://forums.guru3d.com/data/avatars/m/260/260150.jpg
Dam, and here i am in my rubbish flat in the Uk and the fastest i can get to the property is 40mb, currently get 37.5mb
https://forums.guru3d.com/data/avatars/m/226/226700.jpg
I'm sure that all of us are pretty much amazed at this achievement. And thankfully, it's this kind of research that will ultimately lead to better throughput for all of us. There is of course little doubt that the technology needed to properly manage (both sending and receiving) this magnitude of data will be developed too. Btw, thanks Hilbert for posting this great article!
https://forums.guru3d.com/data/avatars/m/53/53598.jpg
JOHN30011887:

Dam, and here i am in my rubbish flat in the Uk and the fastest i can get to the property is 40mb, currently get 37.5mb
Welcome to the family, although i am only getting 34 meg at full grunt, but on the upside my bro in law is on 200 meg VM but streaming from netflix or iplayer he will get that annoying buffering while wayching stuff, something i never suffer from on my none cable connection.
https://forums.guru3d.com/data/avatars/m/266/266726.jpg
Makes the future of optical computer seem more feasible and desirable.
https://forums.guru3d.com/data/avatars/m/273/273678.jpg
DocStr4ngelove:

So imagine you just click on the download button in Steam and it morphs into the 'play' button instantly. That's .... massively mindblowing!
your disk will max out limiting your download.
https://forums.guru3d.com/data/avatars/m/250/250418.jpg
Meanwhile my ISP is cuckolding me with 120 down and 12 up...
https://forums.guru3d.com/data/avatars/m/121/121558.jpg
And here in Canada I'm using a 'typical' relatively-affordable connection of 50 Mbps Download (translates to about a maximum of 7.5 mb/s, under best conditions), and around 7 Mbps Upload, which is borderline just enough to stream at 1080p 60fps. And that stuff costs $60 per month for the first 2 years (if it's your first time subscribing for that package) and goes up to its "regular price" of $85 per month after that (yeah, in Canadian dollars). So whenever I see claims of high speeds from research, or from currently-existing services out there in various countries all I can think of is if that stuff came here we'd have to pay something like $200 per month for it if we're very lucky. Basically: Not gonna happen anytime soon here for about 95% of the consumers base.
data/avatar/default/avatar09.webp
Astyanax:

your disk will max out limiting your download.
A tthis speed you could just stream straight to RAM. No disc necessary.
https://forums.guru3d.com/data/avatars/m/232/232349.jpg
What they also didn't explain was the distance they tested it at. How far away were they from end to end??
data/avatar/default/avatar22.webp
David3k:

Even more interesting is the question of how do you do error checking of that much data without introducing latency?
Error-checking would be implemented in hardware, in parallel with the transceiver. If you have a design that is fast enough to transmit at that speed, then you have transistors that can switch fast enough for that. You use some of your transistor budget to add some latches, you latch on the last n bits (depending on the error-checking algorithm you're implementing), you get a validity output every n bits, and you feed that back into the rest of your design in order to determine what happens next (ask for retransmission, perform error correction if possible, etc...). You scale horizontally if your implementation is parallel (e.g. 8 transceivers in parallel).
https://forums.guru3d.com/data/avatars/m/279/279919.jpg
Astyanax:

your disk will max out limiting your download.
Well as soon as you've played something it autodeletes itself. No need to store anything longer than you use it because it's just available upon clicking on 'download'.
https://forums.guru3d.com/data/avatars/m/232/232130.jpg
DocStr4ngelove:

Well as soon as you've played something it autodeletes itself. No need to store anything longer than you use it because it's just available upon clicking on 'download'.
F for SSD write cycles 😀