Intel Adds Hardware Acceleration for AV1
Intel announced its new discrete Arc GPU for notebook use now and with desktop and workstation implementations coming during this year. Among features that could impact 8K are:
- The chip is said to be the first to support hardware encoding and decoding of the AV1 codec. The encode feature is said to be 50X faster than software implementations and Intel claims to have a way to ‘gang up’ the video processors to bring them to bear on an encoding process.
- The Xe Media Engine is said to be optimised for 8K workloads with AV1 up to 8K with 10 bit HDR. It also supports AVC, HEVC and VP9
- It has new supersampling that supports AI-enhanced upscaling. There’s a video here and Intel said that the new GPU is 2.35X faster at upscaling than its previous Iris Xe integrated graphics processor (using Topaz Labs Video Enhance AI)
- The Arc Xe display engine can support four UltraHD 120Hz displays over HDMI 2.0b or DisplayPort 1.4a or dual 8K 60Hz displays. 1080P and 1440P displays can be driven up to 360Hz. Alternatively, when products are available in the market, it will support DisplayPort 2.0
- Intel sees AV1 as being able to support game streaming. It has less distortion than H.264 at the same bit rate.
- It has a feature called Smooth Sync used when there the display frame rates do not match the frame rate of the GPU to minimise the ‘tearing’ effect. Jon Peddie said that the system blurs the edges of the new overlapping frame using a dithering filter.
We had a bit of a dig into the AV1 encoding aspects of this news. While each generation of codecs gets more efficient at improving quality or reducing bitrates, the trade-off is that the complexity of encoding, especially, gets more complex. For example, Fraunhofer’s VVC codec is said to be able to deliver similar quality to HEVC at half the bit-rate. However, it comes at the expense of a 20-fold increase in compression time.
The firm has a new technology that it calls Intel DeepLink Hyper Encode. The technology can split the workload of video compression between a GPU that is integrated with the CPU and one that is in a discrete graphics card. You can find a video showing the difference in DaVinci Resolve which is up to 60% faster here https://www.intel.com/content/www/us/en/products/docs/arc-discrete-graphics/creator.html
The Xe media engines in the Arc GPU are able to add hardware acceleration to video encoding and decoding. The link above includes a video showing how the media engine will accelerate the AV1 codec and also compares AV1 with the H.264 codec. Intel highlighted that the codec could be used to stream games from the cloud. The demo shows the differences in image quality between the two codecs.
While in the cloud, you can exploit heavyweight processors such as Intel’s Xeon platform to speed up compression of AV1, the hardware acceleration in the Arc GPU means that supporting this kind of advanced codec is still practical on the client side and allowing usable workflows.