Media (Intel® Video Processing Library, Intel Media SDK)
Access community support with transcoding, decoding, and encoding in applications using media tools like Intel® oneAPI Video Processing Library and Intel® Media SDK
The Intel Media SDK project is no longer active. For continued support and access to new features, Intel Media SDK users are encouraged to read the transition guide on upgrading from Intel® Media SDK to Intel® Video Processing Library (VPL), and to move to VPL as soon as possible.
For more information, see the VPL website.

What is low latency according to Intel for QSV H.264 encoding?


When reviewing marketing and other literature Intel consistently is referring to QSV and their H.264 encoder as having "lower latency or low latency. While I understand that itdoes have lower latency in comparison to a software encoder, what are Intel's actual latency results when compared to say software only, Sandy Bridge, and now Ivy Bridge? How is Intel defining low latency? is it 2ms? 10ms? .5ms?

Are there published benchmarks or case studies that take different sized video streams, say 480p 30fps, 720p 30fps, 1080p 30fps, etc. and measure the encoding latency on each different system using a live 1080p HD or 480p stream? What would be nice is a general matrix like the below with a latency number filled in to the blanks.

Core2 software only

Core i7 Sandy Bridge 2600k

Core i7 Ivy Bridge xxxxk (relevant to SNB)

320x240 at 30fps

640x480 at 30fps

480p at 30fps

720p at 30fps

1080p at 30fps

With any benchmark there are standard configurations used for memory/HD/etc so I know these are factors in measuring. Im not looking for a discussion on that topic :) Latency should be the time from source video in to encoded video out

0 Kudos
3 Replies


I am a bit confused by the exact measurements you are looking for.

Your table shows stream information of width x Height @30fps, but if you are looking for time to recieve first frame, the "frames per second" is not relevant, but the bitrate/quality of the encoding would make a significant difference. Also iun the Ivy Bridge column you mention "relevant to SNB", but in seems you wanted a static benchmark.

The marketing claim of "low latency" is not a performance improvement claim, but is instead an additional feature (software interface) that allows new variables to be set for new usage models (for example, video conferencing). The new software interface allows your application to request that priority be given to providing the frames quickly (as opposed to optimization for encoding or trans-coding an entire file). This feature was added to Media SDK for API version 1.3.

Our literature referring to "low latency" should be referring to our Media SDK 2012 product and the hardware that supports Media SDK 2012.

For information about the new features, please see


0 Kudos

Hi,I have read the article, but  After 40 frames have been to encoder, I got the first encoded data, I hope to  Get encoded data immediately,

What should  I do 

0 Kudos
New Contributor III

What is the value of mfxVideoParam::AsyncDepth parameter in your configuration? Try setting it to 1.

0 Kudos