Transcoding GPU Benchmark

[CyklonDX]

GitHub - SpaceinvaderOne/transcoding-gpu-benchmark: How many simultaneous 4K transcodes can your GPU sustain? Multi-vendor (Intel/AMD/NVIDIA/CPU) benchmark for Unraid with HDR tone-mapping, subtitle burn-in, power efficiency and a community leaderboard.

How many simultaneous 4K transcodes can your GPU sustain? Multi-vendor (Intel/AMD/NVIDIA/CPU) benchmark for Unraid with HDR tone-mapping, subtitle burn-in, power efficiency and a community leaderbo...

github.com

(my B60 7th place, 4080 etc is already there.)

Leaderboards:

Transcoding GPU Benchmark — Leaderboard

Top 10 4K HEVC to 1080p H.264

�

	GPU	Streams	Best	Watt/Stream	Runs
1	▸NVIDIA RTX PRO 6000 Blackwell Workstation Edition · 96 GB locked nvidia	52 (1 clean run) provisional	52	3.12	1
2	▸NVIDIA GeForce RTX 5090 · 32 GB unlocked nvidia	22 (1 clean run) provisional	22	4.84	1
3	▸NVIDIA RTX PRO 4000 Blackwell SFF Edition · 24 GB locked nvidia	21 (1 clean run) provisional	21	1.87 most efficient	1
4	▸NVIDIA RTX PRO 4000 Blackwell · 24 GB locked nvidia	21 (1 clean run) provisional	21	3.35	1
5	▸NVIDIA RTX 4000 SFF Ada Generation · 20 GB unlocked nvidia	18 (1 clean run) provisional	18	2.57	1
6	▸NVIDIA RTX 4000 SFF Ada Generation · 20 GB locked nvidia	18 (3 clean runs) (17–18)	18	2.66	3
7	▸Battlemage G21 [Intel Graphics] · 24 GB intel	17 (1 clean run) provisional	17	3.65	1
8	▸Battlemage G21 [Intel Graphics]VRAM not recorded intel	17 (10 clean runs)	17	—	10
9	▸Battlemage G21 [Intel Graphics] · 16 GB intel	17 (4 clean runs) (14–17)	17	2.6	4
10	▸Battlemage G21 [Arc B580] · 12 GB intel	17 (13 clean runs) (16–17)	17	3.57	15

 

[thulle]

Might wanna label the columns

 

[CyklonDX]

fixed

 

[BoredSysadmin]

What's "best"? Most streams?

 

[CyklonDX]

What's "best"? Most streams?

Thats right most streams.

 

[jode]

I am shocked, SHOCKED!, that it doesn't run on my Nvidia 710

Spoiler: Nvidia 710

root@linux:~docker run --rm --gpus 'all,"capabilities=compute,utility"' \ \
    nvidia/cuda:11.6.2-base-ubuntu20.04 nvidia-smi
Unable to find image 'nvidia/cuda:11.6.2-base-ubuntu20.04' locally
11.6.2-base-ubuntu20.04: Pulling from nvidia/cuda
96d54c3075c9: Pull complete
a3d20efe6db8: Pull complete
bfdf8ce43b67: Pull complete
ad14f66bfcf9: Pull complete
1056ff735c59: Pull complete
Digest: sha256:a0dd581afdbf82ea9887dd077aebf9723aba58b51ae89acb4c58b8705b74179b
Status: Downloaded newer image for nvidia/cuda:11.6.2-base-ubuntu20.04
Wed Aug 12 01:22:43 2026       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.256.02   Driver Version: 470.256.02   CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| 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 GeForce ...  Off  | 00000000:05:00.0 N/A |                  N/A |
| 40%   45C    P8    N/A /  N/A |      1MiB /   978MiB |     N/A      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                              
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+
root@linux:~docker run --rm -p 8088:8088   --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=all -e NVIDIA_DRIVER_CAPABILITIES=all --tmpfs /ramdisk:rw,size=512m   -v tgb-config:/config   -v /etc/os-release:/os-release:ro   spaceinvaderone/transcoding-gpu-benchmark:latest
[scoreboard] serving on :8088 (detected GPUs: NVIDIA GeForce GT 710, AMD Ryzen 9 5900X 12-Core Processor)
^Croot@linux:~#

 

[CyklonDX]

Video Encode and Decode Support Matrix

Find the related video encoding and decoding support for all NVIDIA GPU products.

developer.nvidia.com

yeah its not even here.

 

[jode]

yeah its not even here.

Too old (as you can see in the legacy driver version).

I still like these a lot because I can run them in the PCIe 1x slot of my AM4 builds and keep the large slots open for real expansion (networking, storage).

I thought the results would be worth a giggle but oh well ...

My Ryzen 5900X manages 5 streams (compared to the consensus 4) probably because I run them a little hot and with quick DDR4-3600 RAM.
Battlemage results are in line with the existing reports.

All good.

 

[jode]

GitHub - SpaceinvaderOne/transcoding-gpu-benchmark: How many simultaneous 4K transcodes can your GPU sustain? Multi-vendor (Intel/AMD/NVIDIA/CPU) benchmark for Unraid with HDR tone-mapping, subtitle burn-in, power efficiency and a community leaderboard.

How many simultaneous 4K transcodes can your GPU sustain? Multi-vendor (Intel/AMD/NVIDIA/CPU) benchmark for Unraid with HDR tone-mapping, subtitle burn-in, power efficiency and a community leaderbo...

github.com

Leaderboards:

Transcoding GPU Benchmark — Leaderboard

Top 10 4K HEVC to 1080p H.264

Just posted #2 spot besting the 5090 with a CPU only run. My 64-core Genoa consumed about as much power as the 5090, but supported ~40% more transcodes. Who says GPUs are best at this task?