Exit
  • Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
  • 한국 커뮤니티
0

How do I get the GPU to do more of the heavy lifting?

Explorer ,
Sep 20, 2023 Sep 20, 2023

Good day Adobe Community,

 

Even with CUDA GPU acceleration on, we have noticed that the Nvidia RTX A4000 is taking to many breaks and offloading a lot of video lifting to the CPU (Xeon-W). Some examples:

 

When working in Premiere Pro and rendering an intense preview, consisting of detailed stabilization analysis and coloring effects, the CPU takes on nearly 100% of the processing. The GPU is only spiking a few times in workload. How can we flip this?

 

When exporting with Media Encoder, the CPU is processing at 50%, but the GPU is only spiking in processing power, including only making use of 3 to 4 GB of its 16 GB memory.

 

When using the heavy duty aspects of Premiere Pro, and I astonished the GPU fan is not considerably kicking (I only hear the CPU fan), as the A4000 is enterprise grade, designed to run 24/7.

 

For a non-Premiere Pro example, when we render with Topaz, generally upfitting and fine tuning from FHD to UHD, Topaz makes considerable use of the GPU (and CPU) at the same time. I have seen Topaz easily and consistly top 80% plus on the GPU when the PC is doing nothing else.

 

So unless the video card thinks were coin mining in Premier Pro and restricting its processing power, the GPU is significantly underutilized when using Premiere Pro (and Media Encoder). What are we missing? How can we get Premiere Pro to wake up the GPU and make it soar!?

 

Thank you for any help.

TOPICS
Hardware or GPU , How to , Performance
308
Translate
Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Sep 20, 2023 Sep 20, 2023

GPU Accelerated Rendering and Hardware Encoding

 

Premiere is not Topaz: quite different under the hood.

Translate
Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
LEGEND ,
Sep 20, 2023 Sep 20, 2023
LATEST

Every one of these apps is built different, partly just from being different people, but actually ... largely because of intensive design considerations on the part of the devs. They have to look at the more common workfows, including the media, effects, and hardware used. Then figure out under each different scenario what different hardware coding would bring what results.

 

So it's a balancing act as to what is piped only to CPU, or only to GPU, or at times shared.

 

As apps like Topaz, AfterEffects, and such have an easy assumption that their users probably have higher-end GPUs, they can code more to the GPU. Premiere users don't tend to have (as a percentage thing) nearly as powerful GPUs. Ergo, more has to be ported to the CPU.

 

Over the last five years, quite a few more things have been ported to the GPU than before. But that GPU Accelerated List Ann linked to is the current setup. When there's a GPU coded effect, as the CPU needs the work, the GPU is brought up.

 

In Warp for instance, all the first step (analyzing) is a CPU process. The actual "building" of the Warp, the second step, is mostly GPU if you have one worth using.

 

Color tonal/chroma correction and changes are nearly all GPU things if you have them.

 

And of course, H.264/5/HEVC encoding is a completely different "hardware/software" encoding process. A lot of users get the two hardware/software rendering confused because it is confusing to use exactly the same naming for two entirely different processes.

 

The H.264/5 thing is totally based on the hardware your system has ... or doesn't ... for long-GOP decoding/encoding. A completely different hardware issue.

Translate
Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines