Skip to main content
Participant
March 19, 2023
Question

Adobe media encoder using 100% GPU on 3D while there's nothing thats 3D and barely using CPU

  • March 19, 2023
  • 3 replies
  • 3457 views

Hey guys, 

 

I've been editing 4K (Sony A7SIII) Footage on my 3rd gen i5 and gtx 1060 3GB. It's not as bad 

as many of you will think haha. Anyways, today I encountered with the problem I edited video

with simple transitions and few segments where i used 4K assets from Acidbite such as (dust, scratches and 16mm gate overlay). Render was looking fine when system was balanced with around 60% CPU and 20% GPU, when render passed the "difficult" segment it became faster, because other than that whole video is just zoom in and zoom outs with sound effects. so since it became faster after that heavy segment i thought it would finish in hour or so and after some time im seeing this(image provided) this type of segment where there is nothing going on just clips laying around and have little zoom in & outs never had problem before.
Does any of you have idea what can i do in this situation?!

Premiere pro vers. 23.1.0 (build 86)
Encoder vers. 23.0.1 (build 1)

CPU - Intel(R) Core(TM) i5-3570S CPU @ 3.10GHz 3.10 GHz

GPU - ASUS Dual GTX 1060 3GB 

RAM - Corsair vengence PRO 16GB 2400MHz

This topic has been closed for replies.

3 replies

EckiAME
Community Manager
Community Manager
October 22, 2024

I guess you could need a bit more RAM. Can you open task manager and see how the app behaves when rendering the project?

Known Participant
October 20, 2023

I am facing the same problem. I describe my configuration below and attach screenshot
The clips are from a Sony A6400 in 4k-30p. Any way to undo the bottleneck or is it inevitable?

 

MSI B550 MB
RTX 3060 TI 8Gb
Ryzen 7 3700x
32Gb DDR4 3200
1tb m.2 W11 System + apps
2tb m.2 raw files
256 SSD SATA cache + some common resources files

both.works
Inspiring
October 25, 2023

Exact same issue - Using all hardware encoding/acceleration, NVIDIA configured...the whole 9+ yards.
CPU i9 - 11th Gen
GPU NV GeForce RTX 3080 Ti
128GB DDR4 3000+
...and NvME drives.


Participant
October 21, 2024

Same problem here. RTX 4070, I-9, 128GB NVE

5 minute 4K RAW footage with Lumetei with transitions is maxxing out 3D GPU and GPU RAM, but ME has been sitting on the same frame for 9 hours and says 28 hiurs to go.  Remaining time increments and I can pause it so I know the program isn't hung, but also doean't apoear ro ne incrementing, and I have no idea what to do, as my bet is starting over will just end the same way, only 9 hours later, but also not sure if this thing will finish before my career is over.  Over projects with same exact camera, footage, effects, etc. work fine, but not this one. 

Has anyone ever solved this? 

Inspiring
March 21, 2023

Thanks for posting that screenshot, it was really helpful. You can see in the bottom of AME that you're using Hardware Encoding and the Nvidia codec. This means that the encoding is happening on the GPU and your Task Manager is showing that your GPU is the bottleneck. GPU encoding is almost always faster than CPU encoding and it was a big deal when this was added to Premiere and AME a couple years ago. You can do a test by changing the encoding method under the Video tab to Software Encoding, but UHD footage takes a lot of horsepower to process and I'm going to guess that your GPU can process it faster than your 10-year-old CPU, but hey, give it a test on a short segment and see.

 

Video encoding is an incredibly complex topic, so simply looking at utilization value in the Task Manager isn't going to paint a whole picture. Someone else could post those same numbers and ask the opposite question of "why isn't my GPU being used more?" In fact, this is the dominant question that has been asked for the last 10 years with regard to rendering. When your CPU is being used more than your GPU and you're using Hardware Encoding, that means the frames being processed were likely not optimized to render on the GPU. Your transitions might not be GPU-accelerated, and the CPU still has to be used to decode other footage before it can be encoded to H.264. So what you're describing actually makes perfect sense, and as soon as AME could throw all of its processing onto the GPU, it did and encoding sped up.