RTSS 6.1.0 beta 2 is expected soon, probably today. Most changes are related to NV12.
Considering that both NVENC and VCE hardware require input video frames to be in YUV (NV12) color format, both NVENC and VCE plugins needed to perform similar RGB->NV12 color space conversion before encoding each frame. So NVENC / VCE benchmark results were affected by both GPU H.264 encoding performance GPU and CPU RGB->NV12 encoding performance.
In order to simplify the NVENC/VCE plugins now RGB->NV12 color space conversion can be performed at RTSS plugin engine level so the plugins no longer need to take care about color space conversion before uploading video frames to encoder. So pure upload+encoding speed will be reflected by pugin benchmark instead of convert+upload+encoding speed before. Due to this reason NVENC/VCE benchmark results in 6.1.0 beta 2 will increase a bit (mostly on high performance presets where CPU conversion could be a part of bottleneck).
Additionally, moving color space conversion task to the plugin engine level gives RTSS a few more bonuses:
- RGB->NV12 conversion inside the plugin engine is effectively multithreaded so it is much more efficient and less CPU consuming now.
- Using NV12 representation instead of RGB24 decreases encoder input queue memory footprint by 2x
- Now you can save raw NV12 video (updated MSI Afterburner beta is required in addition to RTSS update to get this option). This option can be rather useful addition to uncompressed video recording. Raw NV12 video gives fixed 2x compression ratio comparing to uncompressed video recording at the cost of some color info loss caused by RGB to NV12 (YUV 4:2:0) color space conversion.
I have to add that those changes do not affect QuickSync plugin. Intel solution is able to convert RGB->NV12 at hardware level pretty fast.
