Because (consumer) ethernet is low bandwidth compared to whats needed for video. DisplayPort 2.0 provide a 80Gbps link. HDMI is 48Gbps. Both already use DSC which is a mezzanine compression as this is already saturated for immediate needs.
Ethernet is absolutely used for media transport outside of the consumer space. Particularly where there's a need to distribute signals further than what DP or HDMI can provide. The challenge is this always comes with a trade off. Any compression will result in some mix of a loss of image quality, increase to latency, and cost. What that mix looks like differs across applications.
The current Ethernet spec supports up to 1000 Gbps, but I think 40 Gbps is about as fast as you can find on the market right now. They could build multi Gbps wifi routers too, it's just not common in the consumer market.
Even 10Gb ethernet isn't suitable for consumers. 40Gb is practically only fiber which has never been successful in homes since it requires more careful handling.
As soon as video data is in packets, it should be routable over a network like any other data.
Give up the custom connectors, custom cables, etc.
Need a wireless display? No worries, we can route the data over wifi too.
Need 5 displays? Thats what an ethernet switch is for. Screen mirroring? We have multicast.
Need a KVM? Well you can probably write a few scripts to change which computer your screen gets it data feed from.
I believe this hasn't happened simply because audio/video people like going to conferences to design their own standards every year, keep their licensing royalties, keep their closed club - a software-only solution over any IP network isn't going to fly.
They write that DP uses packets, but it's still an isochronous connection with guaranteed BW allocation.
For video timings with high pixel clocks, the bandwidth used to transfer a signal is very close to what's theoretically available. There's no way you'd be able to do that reliably over something like an Ethernet cable.
There's nothing sinister about how video standards are designed. From DP1.0 to DP1.4, which spans more than 10 years, all changes have been incremental and most just a matter increasing data rates while the coding method stayed the same.
They need a system that's very high and guaranteed bandwidth, high utilization, very low bit error rate, and cheap.
Even today, a 10Gbps Ethernet card will set you back $90. And that will carry less than half the data that can be transported by DP 1.4 which is old school by now.
Have you tried VNC over an Ethernet cable at all? What you are proposing is basically having a VNC-like protocol over IP over Ethernet. Now try it out and see how well it works.
To be even more reductive, a thought experiment is to consider the classic USB. Why did people even invent USB in the first place in the 1990s? The first USB had only 12Mbps, not much better than the first Ethernet at 10Mbps which had existed since 1980! Why didn't people who invented USB simply use Ethernet?
The main delay in VNC is the software stack encoding and decoding the data.
A video-over-ethernet solution would likely have the GPU directly packetize the data into IP packets and route them direct to the network hardware without the CPU touching anything. The network would have QoS to guarantee the necessary bandwidth and guarantee a fixed latency and no packet loss (ie. there is no chance of a packet being dropped due to a queue overflowing, because we have already arranged a fixed bandwidth allocation for the whole network route).
Naw it's because >=40G Ethernet never saw much consumer adoption. QSFP DACs and optics are also kinda bulky by consumer standards so a lot of people would want a new connector. Even LC connectors would be too large for laptops.
If this were true, then people would be running datacenter networking over HDMI.
Data is data - there is no benefit to designing two entirely different transport mechanisms for video vs other data.
You could argue that video data has a deadline to meet - but ethernet already has lots of mechanisms for QoS and bandwidth reservations to make sure that someones torrents don't interrupt something latency sensitive. Sure, they aren't widely supported over the internet, but between your PC and your display - they can be engineered to work properly.
> If this were true, then people would be running datacenter networking over HDMI.
Why would they? If they need more bandwidth than what Ethernet provides, they'd probably just bite the bullet and go with fiber.
...which punts the previous question to why the industry hasn't gone all-in with fiber optic display connectors, like it has with TOSLINK for S/PDIF in the audio world.
Not entirely true. My house was built with an A/V closet that's not in the living room. To put analog equipment (like a turntable) in the living room where it's used, I convert the analog signals into digital with a $100 A/D converter that sends them over CAT-6 under the house to a decoder that connects to the receiver with Toslink. It works great, and I bought this in the last three years.
I was curious about that; because I couldn't be bothered to do the math myself, I looked up the max bitrate of multichannel PCM. Found this at https://www.videohelp.com/hd :
"Linear PCM up to 9 channels (Max 27.648 Mbit/s)"
Cat5 or higher (especially only 20 feet) would easily handle that.
There’s nothing magic about DP cables, they’re just typically a short-ish length, which is why they can be high bandwidth and relatively cheap. Once you get to a decent length, the cables get really expensive because you have to put optical transceivers on either side. I have a 25 foot DP cable which cost over $100, for example.
So you could do a short length, high bandwidth Ethernet cable, I’m sure. But the reason we don’t is probably because differentiation between connectors is desired —- consumers are dumb, frankly. They’ll think any old Ethernet cable will work. Just look at what kind of a mess exists with USBC.
The commonality of connector is important. Ethernet has stopped using twisted pair cables for higher speeds. The SFP transceiver means can use short DAC copper cable or fiber optics for longer distances. The DAC cables that would compete with DisplayPort are pretty cheap.
Also, a big portion of the cost of networking is in the switch that can handle high bandwidths. My guess is that 25G DisplayPort switch would be just as expensive as 25G SFP one.
Yeah, my original comment was based on the "100x more expensive" assertion. a 5m DAC cable is $40 on Amazon, and 25GBe cards can be had for under $200. I think it's probably more than $2.40 for just a 5m DP cable, so 100x was a gross exaggeration.
