Newshot

Players develop experimental setup to stream GTA6 on PC: using GPU frame generation to boost 30 fps to 60 fps

AI summary

Developer Phosphen announced a special streaming solution that lets players on PC stream the PS5 version of GTA6 and use their GPU to post-process the streamed image, raising it to 60 FPS and improving visuals, though it cannot reduce actual input latency because the PS5 still runs the game at 30 FPS. The setup forces 30 FPS gameplay to be transmitted at 1080p60 to a Windows PC, where the open-source software Veyra removes duplicate frames and frame generation fills in intermediate frames, with experimental DLSS 5 running on an RTX 5080; in a test of about 80 seconds of footage across 14 scenes, mainly from IGN's previously released PS5 Pro GTA 6 gameplay demo, distinct frames rose from 2,351 to 4,700 and average frame rate from 29.5 FPS to 59 FPS. Because the processing happens after rendering and only handles compressed video without access to engine data, Veyra must estimate depth information and motion vectors itself, which can cause visible artifacts in fast-motion scenes, and image quality is further limited by PS5 Remote Play's 1080p lossy stream and by the extra latency of sending inputs to the console.

Translation(model translation, not the source's own text)

Players develop experimental solution for PC streaming GTA6: using PC GPU frame generation to interpolate 30 FPS gameplay to 60 FPS

IT之家, November 20: Developer Phosphen announced that they have developed a special streaming solution for GTA6. When players stream the PS5 version of GTA6 on PC, they can use their GPU to post-process the streamed game image, boosting the game visuals to 60 FPS and also improving visual quality. However, since the PS5 still runs the game at 30 FPS, this solution cannot reduce actual input latency.

According to the introduction, the basic principle of the streaming solution is to force the 30FPS game image to be transmitted to a Windows PC at 1080P60 Hz, then use open-source software Veyra to identify and remove duplicate frames, and use frame generation technology to fill in intermediate frames, ultimately letting the PC monitor display the game at 60 FPS.

In terms of image quality enhancement, the solution also incorporates an experimental DLSS 5 implementation. According to the project description, Veyra uses an NVIDIA-signed neural rendering runtime environment on an RTX 5080 GPU.

Because this solution performs post-processing after the game image has been rendered, it has certain technical limitations. Unlike games with DLSS integrated directly, Veyra cannot read raw data provided by the game engine; it can only process video frames that have already been rendered and compressed, and must estimate depth information and motion vectors on its own. In fast-motion scenes, this processing method may produce noticeable visual artifacts.

Phosphen tested with about 80 seconds of gameplay footage covering 14 scenes, mainly from the PS5 Pro version of GTA 6 gameplay demonstration previously released by IGN. Test results show that the original footage contained 2351 distinct frames; after processing, this increased to 4700 frames, and the average frame rate rose from 29.5 FPS to 59 FPS.

In terms of performance, Phosphen says the RTX 5080 takes about 12 milliseconds to process each frame in real time, with two neural rendering passes taking 4.1 milliseconds each, frame generation taking 1.5 milliseconds, and motion information calculation taking 2.1 milliseconds. GPU usage is about 39%, and the 1% Low frame rate reaches 56.3 FPS.

It should be noted that although the extra generated frames can make the game action on the monitor smoother, players are still limited by the PS5's original 30 FPS during actual operation. Because

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by IT之家
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗