CES 2025 Thread (AMD, Intel, Nvidia, and others!)

[MfA]

Huh? Nvidia has been the most vocal and active player tackling rendering pipeline latency in recent memory.

Yes ... after it increasing it by 1 frame.

You're also not really addressing that decreasing it by far more through frameless would kill G-sync ... which is painful.

 

[trinibwoy]

Yes ... after it increasing it by 1 frame.

Not sure what timeline you’re looking at but reflex and ULLM predate framegen by quite a bit.

You're also not really addressing that decreasing it by far more through frameless would kill G-sync ... which is painful.

Frameless? What’s that?

 

[raytracingfan]

Frameless? What’s that?

I'm guessing he means something like Adaptive Frameless Rendering. That paper is two decades old, but now that real-time ray tracing is finally a reality maybe its time has come.

 

[arandomguy]

Yes ... after it increasing it by 1 frame.

Reflex was launched 2 years prior to DLSS3 FG/RTX 40xx in 2020.

Prior to that they gave options in the driver for a "low latency mode" and "ultra low latency mode" at the expense of frame rate output.

 

[MfA]

Or NVIDIA's own paper Temporally Dense Ray Tracing. Really though, it's all in the name. The rendering doesn't create any proper frames any more, it just creates samples ... to be used by some backend to create actual frames, v-synced frames.

This would obviously kill G-Sync if adopted, which is painful.

 

[raytracingfan]

Or NVIDIA's own paper Temporally Dense Ray Tracing. Really though, it's all in the name. The rendering doesn't create any proper frames any more, it just creates samples ... to be used by some backend to create actual frames, v-synced frames.

This would obviously kill G-Sync if adopted, which is painful.

Nvidia could find some other way to sell a special module to display OEMs. Maybe the new module could make the display itself frameless, so whenever the renderer creates samples the corresponding pixels on the display are immediately updated instead of waiting for an entire frame to be ready.

 

[trinibwoy]

This would obviously kill G-Sync if adopted, which is painful.

Why would Nvidia care about Gsync? It’s not doing anything for their bottom line.

 

[DegustatoR]

Or NVIDIA's own paper Temporally Dense Ray Tracing. Really though, it's all in the name. The rendering doesn't create any proper frames any more, it just creates samples ... to be used by some backend to create actual frames, v-synced frames.

This would obviously kill G-Sync if adopted, which is painful.

Why would that kill gsync? You would just set the fps target the engine would be outputting. There is no reason why this target has to be specifically at vsync refresh.

 

[Scott_Arm]

Nvidia could find some other way to sell a special module to display OEMs. Maybe the new module could make the display itself frameless, so whenever the renderer creates samples the corresponding pixels on the display are immediately updated instead of waiting for an entire frame to be ready.

Nvidia already did this. The Nvidia-made gsync module is dead, but MediaTek has the license for it and they're building gsync functions and new gsync tech like Nvidia Pulsar into their display scalers. So it's third party and it's fully integrated into the display scaler which is a processor that every display already needs. The available of gsync should actually grow, not shrink, because you don't need a separate costly module.

Nvidia doesn't look like they're abandoning gsync technology any time soon, and they seem to be growing a big partnership with MediaTek, who is doing all kinds of stuff with them (potentially a future desktop cpu).

From this summer:

Level Up: NVIDIA, MediaTek to Bring G-SYNC Display Technologies to More Gamers

NVIDIA and MediaTek are working together to make the industry’s best gaming display technologies more accessible to gamers globally.

blogs.nvidia.com

 

[MfA]

Why would that kill gsync? You would just set the fps target the engine would be outputting. There is no reason why this target has to be specifically at vsync refresh.

Framegen has nearly deterministic processing time. When the position in time for the generated frames is decoupled from the rendered samples from the game engine, the FPS becomes not a target, but a parameter. If you can create approximate frames at arbitrary points in time reliably, you will create them at fixed intervals and display them v-synced, because it's the most sensible thing to do.

 

[DegustatoR]

Framegen has nearly deterministic processing time. When the position in time for the generated frames is decoupled from the rendered samples from the game engine, the FPS becomes not a target, but a parameter. If you can create approximate frames at arbitrary points in time reliably, you will create them at fixed intervals and display them v-synced, because it's the most sensible thing to do.

Why is that any more sensible than displaying them at whatever refresh the display supports?

 

[MfA]

you will create them at fixed intervals and display them v-synced, because it's the most sensible thing to do.

Is not in opposition to

displaying them at whatever refresh the display supports

What no longer makes sense is VRR (ie. continuously adapting the refresh rate). Making VRR work as well as possible is what G-Sync is for.