You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
I’m curious why 10gbps copper ethernet has taken so long to get “consumer cheap” compared to previous iterations.
We went from 10 mbps being ‘affordable’ in the late 1980s to 100 mbps in the 1990s and 1 gbps in the ~ 2005 timeframe.. yet almost 20 years later we’re still not there with 10 gbps copper.
I *think* the barriers have been:
1. Consumer demand. (= not building enough units to get cost down or justify more engineering) (and low demand for a lot of reasons)
2. PCI Express bandwdith (a lot more expensive when you require 4 or more lanes — only PCIe 4.0 has enough bandwidth to sustain 10 gbps bi-directionally on one lane
3. Power consumption for ports/cards/NICs (it seems to have taken a long time to get 10G down to the power level 1G was at even 10-15 years earlier)
Is there something else beyond this I’m missing? Is the ASIC complexity for communicating at that speed over copper disproportionately harder gen over gen?
Also is there a certain die size and/or power level that faster Copper Ethernet needs to hit for it to become ubiquitous or cheap? How close are we to that today?
Thanks..
P.S. I know pricing is very subjective, my personal stance is 10 gbps is definitely entering the “professional consumer” price range but still above even “premium” commodity pricing.
In my opinion, the primary reason 10GbE client support is a niche is that you can't buy 10Gbps internet connections. I'm not sure how far off high volume 10Gbps internet connections are, but I suspect we're talking several years from now, at least in the US. Comcast revealed a prototype cable modem early this year that got to 4Gbps with DOCSIS 4.0, but that's not 10Gbps:
I’m curious why 10gbps copper ethernet has taken so long to get “consumer cheap” compared to previous iterations.
We went from 10 mbps being ‘affordable’ in the late 1980s to 100 mbps in the 1990s and 1 gbps in the ~ 2005 timeframe.. yet almost 20 years later we’re still not there with 10 gbps copper.
I *think* the barriers have been:
1. Consumer demand. (= not building enough units to get cost down or justify more engineering) (and low demand for a lot of reasons)
2. PCI Express bandwdith (a lot more expensive when you require 4 or more lanes — only PCIe 4.0 has enough bandwidth to sustain 10 gbps bi-directionally on one lane
3. Power consumption for ports/cards/NICs (it seems to have taken a long time to get 10G down to the power level 1G was at even 10-15 years earlier)
Is there something else beyond this I’m missing? Is the ASIC complexity for communicating at that speed over copper disproportionately harder gen over gen?
Also is there a certain die size and/or power level that faster Copper Ethernet needs to hit for it to become ubiquitous or cheap? How close are we to that today?
Thanks..
P.S. I know pricing is very subjective, my personal stance is 10 gbps is definitely entering the “professional consumer” price range but still above even “premium” commodity pricing.
1. For home users and small to medium size business, the advancement of WiFi and the relatively lower cost and easiness to deploy WiFi system make the 10Gbs copper Ethernet less attractive.
2. For corporations, not all but majority of them, the 1 Gbs connection from the wiring closet to each PC is probably good enough for most daily operations. What they are more worried about is the backbone capacity that connects all the wiring closets on each floor and in each building. Then the dropping cost and longer distance and higher bandwidth capabilities make the fiberoptic network more attractive than 10Gbs copper Ethernet.
1. For home users and small to medium size business, the advancement of WiFi and the relatively lower cost and easiness to deploy WiFi system make the 10Gbs copper Ethernet less attractive.
Definitely, but even WiFi 6 in 2x2 mode will only achieve about 2Gbps per stream (2.4Gbps theoretical), and even that's assuming a line of sight access to an AP.
2. For corporations, not all but majority of them, the 1 Gbs connection from the wiring closet to each PC is probably good enough for most daily operations. What they are more worried about is the backbone capacity that connects all the wiring closets on each floor and in each building. Then the dropping cost and longer distance and higher bandwidth capabilities make the fiberoptic network more attractive than 10Gbs copper Ethernet.
I agree; most business apps are lightning fast with 1Gbps of bidirectional throughput. Some aren't, like 8K content creation, medical imaging, data science, some telemetry analysis, but these apps are a small part of the market. Growing faster than standard office apps, but starting from a small base.
Fibre is always better than copper for transmission, but until silicon photonics (or the equivalent) are practical, available fibre connections have most of the power consumption and cost of copper circuits plus the transceivers.
Home 10Gbps internet connections are probably going to be a niche for a long time once available. What fraction of the population will be super-serious gamers and 8K movie watchers? The real business case for 10Gbps internet is edge computing, IMO.