NC State University's Frameless Rendering for Sub-2ms Latency Displays

[Zaphorizhia1]

This could be revolutionary right? and swing the ball back to OLED's favor considering you would only care about gtg after tossing out frames entirely?

Idk how to link yt videos here just search topic title on youtube.

[RealNC]

phpBB [video]

[blurfreeCRTGimp]

Considering how many sample and hold frames current GPUs are capable of rendering IE 90 hz natively or 11 ms easily even on 70 class chips, I feel like you could have an ML model take in a standard rasterized frame (in the context of a run ahead instance on GPU, asynchronus rendering, or unconstrained Vsync off rendering) and convert standard frames to a point cloud or voxels fast enough ( FPGA box in the middle process) so that if a given screen had this group's electronics/driver circuitry, algorithm's etc. you could use a standard OLED or micro LED and a standard GPU signal and still get this method's latency and persistence benefits.

This is insanely cool, but I feel like doing this on micro LED (at least right now) would absolutely hobble any adoption unless they nab A-Wall as a backer.

[RealNC]

I'm not sure I see the point of this for consumer displays. It's way easier to have a 1000Hz display and show normal 60FPS content in it, giving a 1ms scanout time, compared to doing this.

[Zaphorizhia1]

this is motion to photon end to end latency at sub 2ms it's not comparable to scanout time.

@blurfreeCRTGimp that sorta defeats the purpose of it being frameless and not relying on frame buffer in the first place. I do agree that this is borderline impossible for the next decade given the bandwidth required but we could potentially see region of interest frameless implementation sooner i guess.

[Zaphorizhia1]

i'm sorry RealNC your point does make sense I was too quick to rebut it poorly. I concede your point and I think it's more likely 1000hz gets standardized than frameless gets implemented any time soon. However, that gets me thinking just where this sub 2ms latency comes from. the rise time of the diode, the physical click of the mouse (what polling rate they are using), etc. I wonder how the floor of this technology compares to rasterized frames.

[RealNC]

i'm sorry RealNC your point does make sense I was too quick to rebut it poorly. I concede your point and I think it's more likely 1000hz gets standardized than frameless gets implemented any time soon. However, that gets me thinking just where this sub 2ms latency comes from. the rise time of the diode, the physical click of the mouse (what polling rate they are using), etc. I wonder how the floor of this technology compares to rasterized frames.

This whole thing sounds like "frameless" doesn't only apply to the display, but to the video generation as well. Not sure. If true, it means GPUs themselves would need to change from rendering into framebuffers to rendering into... something else?

Way more info is required here than what's on that video