URL: https://www.progressiverobot.com/minimum-gpu-benchmark-for-video-editing-on-gpu-droplets/

Introduction

With the launch of GPU Droplets powered by NVIDIA H100 GPUs, the cloud provider provides an ideal solution for high-performance video editing. The NVIDIA H100, equipped with 640 Tensor Cores and 128 RT Cores, supports faster data processing, enabling high-resolution video scaling and encoding tasks.

This tutorial will walk you through setting up a GPU Droplet for video editing, utilizing FFmpeg with CUDA support to scale and encode a sample video file from 720p to 4K resolution. This tutorial is tailored for video editors and developers looking for an efficient way to handle video processing on cloud infrastructure.

Prerequisites

Before you start, ensure you have:

A the cloud provider Cloud account.
A GPU Droplet deployed and running.
Familiarity with the Linux command line. To learn more, you can visit this guide on Linux command line primer.

Understanding GPU Benchmarks for Video Editing

Effective video editing requires GPUs with specific capabilities:

VRAM: With 80GB HBM2e, the NVIDIA H100 can handle 4K and 8K videos.
CUDA Cores: 18,432 CUDA cores for high-speed processing and encoding.
Tensor Cores: 640 Tensor Cores to support AI-enhanced tasks, such as noise reduction.
RT Cores: 128 RT Cores for real-time processing and visual effects.
Memory Bandwidth: Up to 2 TB/s, allowing for smooth playback and fast data transfers.

For video editing, these specifications translate to faster processing, efficient scaling, and real-time effects rendering.

Minimum Benchmark Goals

Here are benchmark targets based on resolution:

Resolution	VRAM Required	CUDA Cores	Tensor Cores	Memory Bandwidth	RT Cores
1080p	8GB	2,000+	100+	300+ GB/s	20+
4K	16GB	4,000+	200+	500+ GB/s	40+
8K	32GB+	8,000+	400+	1+ TB/s	80+

The NVIDIA H100 meets and exceeds these benchmarks for 4K and 8K video editing, making GPU Droplet an excellent choice for advanced video projects.

GPU cloud servers: NVIDIA H100 Specifications

The NVIDIA H100 offers extensive CUDA, Tensor, and RT cores, providing the necessary resources for high-resolution video editing workloads.

GPU Model	VRAM	CUDA Cores	Tensor Cores	Memory Bandwidth	RT Cores
NVIDIA H100	80GB	18,432	640	2TB/s	128

The H100 can handle even the most intensive editing tasks with minimal lag and high-speed processing.

Setting Up Video Editing Workload on GPU cloud servers

In this section, let's set up and deploy a video editing workload on a GPU cloud servers.

Step 1 - Set Up the GPU Droplet

1.Create a New Project – You will need to create a new project from the cloud control panel and tie it to a GPU Droplet.

2.Create a GPU Droplet – Log into your cloud account, create a new GPU Droplet, and choose AI/ML Ready as the OS. This OS image installs all the necessary NVIDIA GPU Drivers. You can refer to our official documentation on how to create a GPU Droplet.

3.Add an SSH Key for authentication – An SSH key is required to authenticate with the GPU Droplet and by adding the SSH key, you can login to the GPU Droplet from your terminal.

4.Finalize and Create the GPU Droplet – Once all of the above steps are completed, finalize and create a new GPU Droplet.

Step 2 - Install Dependencies

Once the GPU Droplet is ready and deployed. You can SSH to the GPU Droplet from your terminal.

				
					ssh root@&lt;your-droplet-ip&gt;

Ensure your Ubuntu-based GPU Droplet is up to date:

				
					sudo apt update &amp;&amp; sudo apt upgrade -y

Next, please reboot the GPU Droplet using the below command and wait for it to come online. Rebooting after running sudo apt update && sudo apt upgrade—y is often necessary to ensure that any updated system components, especially the kernel and hardware drivers, are fully loaded and applied.

Note: On most Linux systems, you can check if a reboot is needed by running:

				
					[ -f /var/run/reboot-required ] &amp;&amp; echo "Reboot is required"

If a reboot is required after a kernel and hardware drivers update, you will observe the following output of the above command:

				
					[secondary_label Output]
Reboot is required

Now,let's verify the NVIDIA driver and CUDA version with the following:

				
					nvidia-smi

This command should display the details of your NVIDIA GPU and the driver version.

				
					[secondary_label Output]
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.183.06 Driver Version: 535.183.06 &lt;^&gt;CUDA Version: 12.2&lt;^&gt; |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA H100 80GB HBM3 Off | 00000000:00:09.0 Off | 0 |
| N/A 28C P0 67W / 700W | 0MiB / 81559MiB | 0% Default |
| | | Disabled |
+-----------------------------------------+----------------------+----------------------+
 
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| No running processes found |
+---------------------------------------------------------------------------------------+

Verify the installation of CUDA:

				
					nvcc --version

You should see information about the installed version of CUDA.

				
					[secondary_label Output]
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Mon_Apr__3_17:16:06_PDT_2023
Cuda compilation tools, release 12.1, V12.1.105
Build cuda_12.1.r12.1/compiler.32688072_0

Note: From the above outputs, please ensure that the CUDA driver version from the output of the command nvidia-smi and the command nvcc -v matches. If there is a version mismatch between the output of both commands, you must re-install both and again reboot the GPU Droplet.

Next, you will install FFmpeg. FFmpeg enables GPU-accelerated video processing, which you will use to scale the video.

				
					sudo apt install -y ffmpeg

Run the following to check if FFmpeg detects CUDA support.

				
					ffmpeg -hwaccels

This command should list cuda, vdpau, and vappi, indicating that GPU acceleration is available.

				
					[secondary_label Output]
ffmpeg version 4.4.2-0ubuntu0.22.04.1 Copyright (c) 2000-2021 the FFmpeg developers
 built with gcc 11 (Ubuntu 11.2.0-19ubuntu1)
 configuration: --prefix=/usr --extra-version=0ubuntu0.22.04.1 --toolchain=hardened --libdir=/usr/lib/x86_64-linux-gnu --incdir=/usr/include/x86_64-linux-gnu --arch=amd64 --enable-gpl --disable-stripping --enable-gnutls --enable-ladspa --enable-libaom --enable-libass --enable-libbluray --enable-libbs2b --enable-libcaca --enable-libcdio --enable-libcodec2 --enable-libdav1d --enable-libflite --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-libgme --enable-libgsm --enable-libjack --enable-libmp3lame --enable-libmysofa --enable-libopenjpeg --enable-libopenmpt --enable-libopus --enable-libpulse --enable-librabbitmq --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libsrt --enable-libssh --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx265 --enable-libxml2 --enable-libxvid --enable-libzimg --enable-libzmq --enable-libzvbi --enable-lv2 --enable-omx --enable-openal --enable-opencl --enable-opengl --enable-sdl2 --enable-pocketsphinx --enable-librsvg --enable-libmfx --enable-libdc1394 --enable-libdrm --enable-libiec61883 --enable-chromaprint --enable-frei0r --enable-libx264 --enable-shared
 libavutil 56. 70.100 / 56. 70.100
 libavcodec 58.134.100 / 58.134.100
 libavformat 58. 76.100 / 58. 76.100
 libavdevice 58. 13.100 / 58. 13.100
 libavfilter 7.110.100 / 7.110.100
 libswscale 5. 9.100 / 5. 9.100
 libswresample 3. 9.100 / 3. 9.100
 libpostproc 55. 9.100 / 55. 9.100
Hardware acceleration methods:
&lt;^&gt;vdpau&lt;^&gt;
&lt;^&gt;cuda&lt;^&gt;
&lt;^&gt;vaapi&lt;^&gt;
qsv
drm
opencl

Step 3 - Download Sample Video Data for Testing

For this tutorial, you will use a sample video from the Blender Foundation to demonstrate GPU-accelerated video processing. Blender is a free and open-source 3D creation suite that supports the entirety of the 3D pipeline—modeling, rigging, animation, simulation, rendering, etc.

Download it with the below command:

				
					wget https://download.blender.org/demo/movies/ToS/tears_of_steel_720p.mov

Step 4 - Upscale the Video to 4K with FFmpeg

You can now process the video with FFmpeg installed and the GPU configured.

The basic syntax of the ffmpeg command is:

				
					ffmpeg -i &lt;input_file&gt; -vf "scale=width:height" -c:v &lt;codec_name&gt; -preset &lt;encoding_preset&gt; -b:v &lt;bitrate&gt; &lt;output_file&gt;

Here is what each of the parameters mean:

input_file: The name of the input video file.
scale=width:height: The scaling filter, where width and height are the desired dimensions for resizing.
codec_name: The codec to be used for encoding the video (e.g., libx264 for H.264).
encoding_preset: The speed vs. compression efficiency preset for encoding (e.g., fast, medium, slow).
bitrate: The target video bitrate (e.g., 10M for 10 Mbps).
output_file: The name of the output file, including format and extension.

In this example, you will upscale the Video from 720p to 4K resolution.

Run the following FFmpeg command to upscale the video to 4K (3840×2160 resolution):

				
					ffmpeg -i tears_of_steel_720p.mov -vf "scale=3840:2160" -c:v libx264 -preset fast -b:v 10M tears_of_steel_4k.mov

Once the processing finishes, you should observe the following output:

				
					[secondary_label Output]
Output #0, mov, to 'tears_of_steel_4k.mov':
 Metadata:
 major_brand : qt 
 minor_version : 512
 compatible_brands: qt 
 encoder : Lavf58.76.100
 Stream #0:0(eng): Video: h264 (avc1 / 0x31637661), yuv420p(tv, progressive), 3840x2160 [SAR 120:89 DAR 640:267], q=2-31, 10000 kb/s, 24 fps, 12288 tbn (default)
 Metadata:
 handler_name : VideoHandler
 vendor_id : FFMP
 encoder : Lavc58.134.100 libx264
 Side data:
 cpb: bitrate max/min/avg: 0/0/10000000 buffer size: 0 vbv_delay: N/A
 Stream #0:1(eng): Audio: aac (LC) (.mov / 0x6134706D), 44100 Hz, stereo, fltp, 128 kb/s (default)
 Metadata:
 handler_name : SoundHandler
 vendor_id : [0][0][0][0]
 encoder : Lavc58.134.100 aac
frame=17620 fps= 69 q=-1.0 Lsize= 949201kB time=00:12:14.07 bitrate=10592.7kbits/s speed=2.88x 
video:937125kB audio:11532kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.057354%
[libx264 @ 0x564504ee7f40] frame I:223 Avg QP:22.07 size:180259
[libx264 @ 0x564504ee7f40] frame P:5973 Avg QP:25.97 size: 92140
[libx264 @ 0x564504ee7f40] frame B:11424 Avg QP:27.21 size: 32306
[libx264 @ 0x564504ee7f40] consecutive B-frames: 8.2% 14.4% 5.4% 72.0%
[libx264 @ 0x564504ee7f40] mb I I16..4: 20.9% 75.8% 3.2%
[libx264 @ 0x564504ee7f40] mb P I16..4: 9.4% 21.2% 1.1% P16..4: 29.5% 4.3% 1.4% 0.0% 0.0% skip:33.0%
[libx264 @ 0x564504ee7f40] mb B I16..4: 0.9% 1.4% 0.1% B16..8: 25.0% 1.3% 0.2% direct: 1.8% skip:69.3% L0:44.6% L1:53.4% BI: 2.0%
[libx264 @ 0x564504ee7f40] final ratefactor: 25.73
[libx264 @ 0x564504ee7f40] 8x8 transform intra:66.4% inter:91.8%
[libx264 @ 0x564504ee7f40] coded y,uvDC,uvAC intra: 30.5% 41.4% 7.3% inter: 6.3% 10.8% 0.3%
[libx264 @ 0x564504ee7f40] i16 v,h,dc,p: 31% 23% 7% 38%
[libx264 @ 0x564504ee7f40] i8 v,h,dc,ddl,ddr,vr,hd,vl,hu: 34% 16% 23% 3% 5% 6% 4% 5% 3%
[libx264 @ 0x564504ee7f40] i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 37% 17% 10% 4% 8% 9% 6% 6% 4%
[libx264 @ 0x564504ee7f40] i8c dc,h,v,p: 55% 16% 22% 6%
[libx264 @ 0x564504ee7f40] Weighted P-Frames: Y:2.2% UV:1.1%
[libx264 @ 0x564504ee7f40] ref P L0: 62.5% 12.6% 18.5% 6.3% 0.1%
[libx264 @ 0x564504ee7f40] ref B L0: 90.1% 8.2% 1.7%
[libx264 @ 0x564504ee7f40] ref B L1: 96.5% 3.5%
[libx264 @ 0x564504ee7f40] kb/s:10456.65
[aac @ 0x564504f67ec0] Qavg: 259.313

Here is the command's breakdown:

-i tears_of_steel_720p.mov: Specifies the input video file.
-vf "scale=3840:2160": Sets the scale filter to upscale the video to 4K resolution (3840×2160).
-c:v libx264: Uses the libx264 codec to encode the video.
-preset medium: Specifies the encoding speed/quality balance (fast is a good balance).
-b:v 10M: Sets the target video bitrate to 10 Mbps to maintain quality.

To learn more about the FFmpeg command, you can refer to its official documentation.

Step 5 - Download the Processed Video to Your Local System

Once the video is processed, download it from the droplet to your local machine using scp. Replace <your_droplet_ip> with your droplet’s IP address.

				
					scp root@&lt;your_droplet_ip&gt;:~/tears_of_steel_4k.mov ~/Downloads/

This command copies the 4K scaled video file to your Downloads folder on your local desktop.

Conclusion

the cloud provider’s GPU Droplets, powered by NVIDIA H100 GPUs, offer a high-performance environment for video editing. With GPU-accelerated scaling and encoding via FFmpeg, you can achieve significant improvements in processing time, enabling real-time adjustments and fast video exports. This setup is ideal for video editors and developers handling high-resolution workloads.

Minimum GPU Benchmark for Video Editing on GPU Droplets

Table of Contents