Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluation of Hardware-based Video Encoders on Modern GPUs for UHD Live-Streaming

Kasidis Arunruangsirilert, Jiro Katto

TL;DR

The paper investigates hardware-accelerated GPU encoders across NVIDIA, Intel, and Qualcomm platforms for UHD live streaming, comparing rate–distortion performance, throughput, and power against software encoders using $PSNR$, $SSIM$, and $VMAF$ across H.264/AVC, HEVC, VP9, AV1, and VVC. It employs two challenging datasets (ITE 4K/8K and Twitch 1080p upscaled to 4K) with real-time-oriented encoding parameters to reflect live streaming conditions, and benchmarks multiple generations and codecs. Key contributions include a comprehensive RD, throughput, and power analysis; codec- and generation-dependent insights; and practical bitrate guidance to match YouTube transcoding quality for streamers and VTubers. The findings show hardware encoders provide real-time performance with superior energy efficiency, while RD gains are driven more by adopting newer codecs than by hardware generation upgrades, yielding actionable guidance for selecting codecs and bitrates in UHD live streaming.

Abstract

Many GPUs have incorporated hardware-accelerated video encoders, which allow video encoding tasks to be offloaded from the main CPU and provide higher power efficiency. Over the years, many new video codecs such as H.265/HEVC, VP9, and AV1 were added to the latest GPU boards. Recently, the rise of live video content such as VTuber, game live-streaming, and live event broadcasts, drives the demand for high-efficiency hardware encoders in the GPUs to tackle these real-time video encoding tasks, especially at higher resolutions such as 4K/8K UHD. In this paper, RD performance, encoding speed, as well as power consumption of hardware encoders in several generations of NVIDIA, Intel GPUs as well as Qualcomm Snapdragon Mobile SoCs were evaluated and compared to the software counterparts, including the latest H.266/VVC codec, using several metrics including PSNR, SSIM, and machine-learning based VMAF. The results show that modern GPU hardware encoders can match the RD performance of software encoders in real-time encoding scenarios, and while encoding speed increased in newer hardware, there is mostly negligible RD performance improvement between hardware generations. Finally, the bitrate required for each hardware encoder to match YouTube transcoding quality was also calculated.

Evaluation of Hardware-based Video Encoders on Modern GPUs for UHD Live-Streaming

TL;DR

The paper investigates hardware-accelerated GPU encoders across NVIDIA, Intel, and Qualcomm platforms for UHD live streaming, comparing rate–distortion performance, throughput, and power against software encoders using , , and across H.264/AVC, HEVC, VP9, AV1, and VVC. It employs two challenging datasets (ITE 4K/8K and Twitch 1080p upscaled to 4K) with real-time-oriented encoding parameters to reflect live streaming conditions, and benchmarks multiple generations and codecs. Key contributions include a comprehensive RD, throughput, and power analysis; codec- and generation-dependent insights; and practical bitrate guidance to match YouTube transcoding quality for streamers and VTubers. The findings show hardware encoders provide real-time performance with superior energy efficiency, while RD gains are driven more by adopting newer codecs than by hardware generation upgrades, yielding actionable guidance for selecting codecs and bitrates in UHD live streaming.

Abstract

Many GPUs have incorporated hardware-accelerated video encoders, which allow video encoding tasks to be offloaded from the main CPU and provide higher power efficiency. Over the years, many new video codecs such as H.265/HEVC, VP9, and AV1 were added to the latest GPU boards. Recently, the rise of live video content such as VTuber, game live-streaming, and live event broadcasts, drives the demand for high-efficiency hardware encoders in the GPUs to tackle these real-time video encoding tasks, especially at higher resolutions such as 4K/8K UHD. In this paper, RD performance, encoding speed, as well as power consumption of hardware encoders in several generations of NVIDIA, Intel GPUs as well as Qualcomm Snapdragon Mobile SoCs were evaluated and compared to the software counterparts, including the latest H.266/VVC codec, using several metrics including PSNR, SSIM, and machine-learning based VMAF. The results show that modern GPU hardware encoders can match the RD performance of software encoders in real-time encoding scenarios, and while encoding speed increased in newer hardware, there is mostly negligible RD performance improvement between hardware generations. Finally, the bitrate required for each hardware encoder to match YouTube transcoding quality was also calculated.

Paper Structure

This paper contains 14 sections, 4 figures, 14 tables.

Table of Contents

  1. Introduction
  2. Experiment Setup
  3. Encoders
  4. Encoding Parameters
  5. Dataset and Resolution Scaling
  6. Objective Evaluation Metrics
  7. Comparison of the Dataset to Real-world contents
  8. Results and Analysis
  9. Encoding Throughput
  10. Power Consumption
  11. Rate-Distortion Performance
  12. Bitrate for YouTube Live-streaming
  13. Conclusions
  14. Future Work

Figures (4)

  • Figure 1: Pipeline used to obtain YouTube transcoding quality.
  • Figure 2: RD Curve of various contents when being encoded by Intel QuickSync on Arc A770 GPU to AV1 codec.
  • Figure 3: Rate-Distortion Curves based on VMAF score of various hardware encoders compared to software encoders and YouTube transcoding when encoding ITE 4K and Twitch dataset at 1080p and 2160p, and ITE 8K dataset at 4320p.
  • Figure 4: Rate-Distortion Curves of three different codecs: H.264/AVC, H.265/HEVC, and AV1 as encoded by Intel and NVIDIA encoders when performing encoding on various datasets at various resolutions.