> Achieving a robust, end-to-end user visual experience remains the utmost priority for VESA’s DisplayPort specification, whether across a native DisplayPort cable, via DisplayPort Alt Mode (DisplayPort over the USB Type-C connector), or tunneled through the USB4 link.
Wake me when I can put a random Nvidia or AMD card in my PC and connect a random monitor, AND have the monitor's USB ports work with only two cables on the monitor (one for power, one to the PC). Bonus points for one cable. Apple's ecosystem has had this for years, but why does my gaming PC and monitor still need a separate USB cable between them for full functionality? My GPU already handles my sound, can't it handle being a USB host device, too?
I didnt fullu understand but I heard some very sad story in commemts a while back that Thunderbolt was largely impossible to do in an add-in-card fashion.
There are of course some add-in cards, and many people use them successfully, but also I dont know if any of these users have multiple cards, which closer models the potential situation here. But notably these users also dont have a dedicated gpu for their thunderbolt ports!
I 100% think likewise to you- the gpu obviously should have usb4 (+ ideally pcie) ports out thr back. Its also frustrating & silly how few systems have usb-pd out.
VirtualLink on the NV 20xx series was almost there on this.
A USB-C socket on the card and it can host audio/video/usb.
Skipped from 30xx onwards, and I'm not sure of the maximum video resolution, but I use an Apple USB-C to HDMI/USB dongle in mine to connect a rift - I don't think there is any reason it couldn't be a monitor on the other end of the USB-C cable.
The lack of VirtualLink/USB-C on nvidia cards seems like a major loss going forward.
Video cards are realistically the only component in a desktop actually hooked up to enough power to cover the full USB-PD spec(240w currently?) + the rest like display/audio/hub.
I can only hope the upcoming AMD 7000 reference series still has a USB-C port. In fact I would be perfectly happy with a card with nothing but USB-C ports on it.
If I want to plug in a bunch of USB hubs and then plug in 50 screens, I should be able to.
I don't care if the rendering happens on the system GPU or the screens. I am happy for frame rates to drop if all 50 screens have stuff animating on them at once and I have a slow GPU. But it should work.
Why do we have this insistence that video needs to have its own special socket, cable, protocol, etc?
I can happily plug in 50 USB headphones and they just work.
Yeah, but if I plug in too many printers into my USB port, they still work, just might take a couple of seconds longer to print a page.
Display stuff needs to gracefully degrade - and that probably means start sending compressed video. Remember even 10 Mbits is enough for HD video, and USBC can do 40Gbps - so there should be no issues with 4000 screens all playing video right? And make sure to use all the available bandwidth - so if I plug in 50 screens, but 49 of them are showing a static image, I should be able to have full performance on the remaining one.
To actually provide reasonable QoS, the "display" is then gonna need to present as a PCIe video accelerator endpoint.
It implies that bare minimum every "display" is in fact, not "a display" but a media player that implements H.264/FFV1/VP9/AV1 (the most common "somewhat FOSS" codecs, I don't think whoever is penny-pinching the BOM of those "displays" will spend extra for decoder licenses).
As we're that far, might as well add GPU and VRAM (a cheap phone SoC with all those components?) and support Vulkan...
For the longest time all digital video protocols kept trying to pump data at standard VGA/CRT timings despite this technology being obsolete for the last 20 years.
DVI just encodes VGA signal in digital form :( Same for HDMI. Both are still sending pixels at VGA timings, Hsync/Vsync/blanking and all, with clocking locked to PixelClock.
DisplayPort is packetized, but not in a 100% sane way. You would expect that clean slate purely digital interface designers would at least optionally allow sending whole screen update at a time. Alas you are still sending same VGA formatted pixel stream (Hsync/Vsync/blanking), just chopped into TUs (32-64 symbol Transfer Units) with stuffing (aka never used garbage) between them, even when connected to eDP display supporting Panel Self-Refresh (own frame buffer inside display).
I havent followed USB 4.0 closely, but there is a chance the newest spec finally flipped from lane switching (dedicating whole pins to useless DP signaling with all the Hsync/Vsync/blanking+garbage stuffing) to real packetized protocol, but knowing designs by committee its the DP packets still with all the garbage and legacy cruft.
I Fully agree we should have switched to Video being just pure data transfer as soon as first Self Refresh/Dynamic Refresh panels showed up on the market.
11 comments
[ 2.6 ms ] story [ 40.6 ms ] threadWake me when I can put a random Nvidia or AMD card in my PC and connect a random monitor, AND have the monitor's USB ports work with only two cables on the monitor (one for power, one to the PC). Bonus points for one cable. Apple's ecosystem has had this for years, but why does my gaming PC and monitor still need a separate USB cable between them for full functionality? My GPU already handles my sound, can't it handle being a USB host device, too?
There are of course some add-in cards, and many people use them successfully, but also I dont know if any of these users have multiple cards, which closer models the potential situation here. But notably these users also dont have a dedicated gpu for their thunderbolt ports!
I 100% think likewise to you- the gpu obviously should have usb4 (+ ideally pcie) ports out thr back. Its also frustrating & silly how few systems have usb-pd out.
A USB-C socket on the card and it can host audio/video/usb.
Skipped from 30xx onwards, and I'm not sure of the maximum video resolution, but I use an Apple USB-C to HDMI/USB dongle in mine to connect a rift - I don't think there is any reason it couldn't be a monitor on the other end of the USB-C cable.
Video cards are realistically the only component in a desktop actually hooked up to enough power to cover the full USB-PD spec(240w currently?) + the rest like display/audio/hub.
I can only hope the upcoming AMD 7000 reference series still has a USB-C port. In fact I would be perfectly happy with a card with nothing but USB-C ports on it.
If I want to plug in a bunch of USB hubs and then plug in 50 screens, I should be able to.
I don't care if the rendering happens on the system GPU or the screens. I am happy for frame rates to drop if all 50 screens have stuff animating on them at once and I have a slow GPU. But it should work.
Why do we have this insistence that video needs to have its own special socket, cable, protocol, etc?
I can happily plug in 50 USB headphones and they just work.
Displayport 2 is 80Gbps which is a lot of data per second...
Display stuff needs to gracefully degrade - and that probably means start sending compressed video. Remember even 10 Mbits is enough for HD video, and USBC can do 40Gbps - so there should be no issues with 4000 screens all playing video right? And make sure to use all the available bandwidth - so if I plug in 50 screens, but 49 of them are showing a static image, I should be able to have full performance on the remaining one.
To actually provide reasonable QoS, the "display" is then gonna need to present as a PCIe video accelerator endpoint.
It implies that bare minimum every "display" is in fact, not "a display" but a media player that implements H.264/FFV1/VP9/AV1 (the most common "somewhat FOSS" codecs, I don't think whoever is penny-pinching the BOM of those "displays" will spend extra for decoder licenses).
As we're that far, might as well add GPU and VRAM (a cheap phone SoC with all those components?) and support Vulkan...
At the end of the day there are limits to how much data you can pump through a cable, and as seen with thunderbolt cables it gets expensive quick.
99.9% of users don't need more than 2 monitors, for those that do they can plug in a few more usb-c.
using video compression sucks balls as anyone who has to use a displaylink monitor on some sucky office hub can attest.
DVI just encodes VGA signal in digital form :( Same for HDMI. Both are still sending pixels at VGA timings, Hsync/Vsync/blanking and all, with clocking locked to PixelClock.
DisplayPort is packetized, but not in a 100% sane way. You would expect that clean slate purely digital interface designers would at least optionally allow sending whole screen update at a time. Alas you are still sending same VGA formatted pixel stream (Hsync/Vsync/blanking), just chopped into TUs (32-64 symbol Transfer Units) with stuffing (aka never used garbage) between them, even when connected to eDP display supporting Panel Self-Refresh (own frame buffer inside display).
I havent followed USB 4.0 closely, but there is a chance the newest spec finally flipped from lane switching (dedicating whole pins to useless DP signaling with all the Hsync/Vsync/blanking+garbage stuffing) to real packetized protocol, but knowing designs by committee its the DP packets still with all the garbage and legacy cruft.
I Fully agree we should have switched to Video being just pure data transfer as soon as first Self Refresh/Dynamic Refresh panels showed up on the market.