NVIDIA Ada Lovelace media capabilities

Ada Lovelace media implementations include AV1 encoding on the specified products, with codec tools, engine counts and input/output constraints belonging to their exact matrix rows.

Ada Lovelace adds a documented NVIDIA AV1 output-encoding path on the specified products, alongside AV1 reconstruction.

That changes the set of possible source/target combinations, provided the application and receiver can use the selected representation.

A named product gives concrete feature scope

Selected GeForce RTX 4090 matrix entries include:

Property Declaration
NVENC generation / physical count Eighth / two
NVDEC generation / physical count Fifth / one
AV1 4:2:0 encoding Yes
AV1 8/10-bit decoding Yes
HEVC ten-bit and B-frame encoding Yes
Listed H.264/HEVC 4:2:2 encoding columns No

The absence in the selected 4:2:2 columns is distinct from ordinary 4:2:0 support. The requested sample configuration should remain visible in the claim.

AV1 output needs a compatible receiver

An original Emby conversion exercise takes HEVC Main 10 input and prepares AV1 output. A supported source decoder, picture route and AV1 encoder establish the server-side operations.

The receiving app still needs the chosen AV1 profile, packaging and presentation path. A server encoder capability does not configure the client's decoder.

The reverse direction—AV1 input to H.264 output—selects another encoder while retaining AV1 reconstruction.

Two encoders describe physical work resources

The RTX 4090 count of two NVENC units is useful for aggregate processing capacity under the documented workload and driver behavior.

It does not mean two clients are the only permitted workload, or that a twelve-session system policy becomes twenty-four. Physical units, permitted active sessions and sustainable media throughput are different quantities.

Concurrent sessions develops that distinction with a checked original workload example.

Preserve the coding choices in a quality comparison

AV1, HEVC and H.264 use different representations and tool sets. Comparing outputs needs the same prepared source interval, target dimensions/cadence, achieved rate and declared encoder configuration.

See Also

References