MSI Afterburner 3.0.0 Beta 18(2014-01-02)

Good news, RTSS 6.1.0 beta is expected soon. Full changes list includes the following:

• Improved encoding plugins architecture:
o Encoder benchmark results are no longer affected by encoder initialization time
o Now the plugins are allowed to start encoding frames without specifying the header prior to encoding. MKV container implementation has been also improved to provide support for delayed header encoding
• Improved dedicated encoder server service:
o Seriously optimized data exchange path between video capture engine and dedicated encoder. Data exchange related performance overhead is seriously reduced and close to invisible now
• Improved Intel QuickSync H.264 encoder plugin:
o Added wait loop to encoding routine to prevent frame dropping when VPP input queue is full. Such frame dropping caused by queue overflow could cause the benchmark to be stopped with MFX_ERR_MEMORY_ALLOC error
o Increased upper adjustable bitrate limit to 50 Mbps
o Output bitstream is now dynamically resizable to allow encoding video at 40 Mbps and higher bitrates
• Added AMD VCE H.264 encoder plugin:
o New plugin uses AMD Accelerated Parallel Processing SDK and AMD OpenVideo API to access fixed function H.264 encoder unit built into AMD RADEON HD 7xxx and newer AMD GPU families. The plugin is able to encode H.264 video in up to 1080p resolution at up to 60 FPS framerate
o The pluging GUI provides you adjustable encoding bitrate and AMD’s recommended speed, balanced and quality presets. However, each preset is represented by independent CFG file and can be additionally tweaked by experienced users if necessary
• Various protections and improvements for low memory conditions:
o Added exception handling for memory allocation routines
o Now encoder input queue size is limited to 256MB in 32-bit applications
o Improved encoder input queue memory management and reduced memory footprint for VFW and external plugin compression modes
• Framerate calculation fix for Outcast, introduced in the previous version, is no longer hardcoded. Now it is being applied at application profile level to selected executables only to avoid breaking support for other applications handling multiple swap chains differently (e.g. Dragon Age : Origins)
• Added raster 3D On-Screen Display rendering mode support for Rayman Legends and ARMA II rendering engines
• Updated profiles list

And special note for those who hate searching: there will be NO Mantle support until AMD release Mantle specs and SDK

 

Decided to add one more small but useful thing to 6.1.0:

- Extended encoder benchmark settings. Now you can run the benchmark in current display settings just like before, as well as select custom display mode. This feature is useful for comparing peak encoding performance of different hardware H.264 encoders (e.g. Intel QuickSync vs AMD VCE)

 

Holy ****! With all this AMD related coding during the last weeks I totally missed very important note in the release notes for 334.67 drivers:


- Now supports NVENC SDK allowing developers to use Kepler-based GeForce GPUs for video encoding.


NVIDIA finally dedided to open free access to NVENC SDK on GeForce GPUs. Previously it only worked on Quadro/Tesla series cards only and required NVIDIA license key to be used on GeForce series GPU. Now it works on GeForce as well so any software developer can use NVENC SDK on GeForce cards without that licensing nonsens.

Edit: I think it is worth delaying 6.1.0 release a bit to create and add NVENC plugin to this build as well. I expect to get it working within a week or so.

 

NvIFR is already in use for D3D9, NvIFR for D3D1x and NVFBC are locked by NVIDIA for their private usage and GRID systems sadly.

 

Benchmark is implemented at plugin engine level. New plugins don't have to provide any special support for it, benchmarking is available for any plugin including future ones.

 

It was already answered: no.