Not clear on which GPU Pr/Ae use

[mhump]

I am fairly new to Adobe Pr and Ae, and would like to better understand how it is utilizing the GPUs in my computer.

I have a laptop with Intel Integrated HD Graphics (8GB Shared Memory), and a discrete nVidia GeForce GTX 1650 Ti with Max-Q and 4GB RAM.

In Premiere, under Project Settings - General, I have selected Mercury Playback Engine GPU Accleraction (CUDA), and in After Effects, under Preferences - Previews, clicking the GPU Information button lists the NVidia card.

For the most part, when I am playing back or rendering, I can see activity on the NVidia GPU.  Playback is smooth, and rendering is fast.

Occasionally, when I have added certain effects, such as transition effects, I notice that it is playing along smoothly, until after it passes the transition effect.  At which point, the NVidia GPU activity drops to zero, and the Intel increases significantly.  Playback is choppy (or won't play at all).  I've also noticed that when i added Grain Removal in Ae, and then rendered the clip back in Pr, it is using the Intel chip to render, and it takes a very long time.

So my question is, why isn't the NVidia GPU being used all the time for these operations, which it is clearly much better at?  Why does Pr switch between the two GPUs?

Thanks,

-M

[Andy 1968]

Keep in mind the iGPU should only encode and decode H.264/265. The graphics card GPU should handle all the effects.

[R Neil Haugen]

Premiere uses the CPU as the primary worker-bee, and the GPU for certain things that (if an adequate GPU is available). Basic decoding/playback of most media does not involve the GPU in especially H.264  media. When the CPU 'sees' something that it is told to use the GPU for, it sends that bit to the GPU as it gets to it.

They are changing some of that mix of CPU/GPU, and over the next year we'll probably see more use of the GPU in general.

Neil

[salvo34]

I have a laptop like you ( hp omen gaming computer) with an Nvidia chip. It claims to have Nvidia cuda and so on. But it's only the chip there on the mobo and not a real full discreet 'graphics card'. I have an editing computer (custom) with an Nvidia graphics card plugged into a pci x16 express port of mobo. I also have other stuff, but it's irrevalent ( like SDI card for hero monitor ).

Anyway, when you boot the laptop you HAVE to use the bios internal graphics just to see your screen on boot.

You can't disable it and go through the Nvidia chip only.. in bios, it's baked in.

My fully discreet graphics card ( in desktop ) has fan on it and it's HUGE. It works well.

But on laptop I only have the chip from Nvidia.

I never try to use that like a real editing computer.. but do put stuff into laptop to scrub through footage sometimes ( even h264 ). It works OK for that stuff. But to really 'edit' stuff you need a full sized graphics card on a bigger computer. A full sized graphics card would be 4X the thickness of your laptop. It simply can't fit in there.

For temperature reasons and all the manufacturers trying to provide cool stuff ( like full editing capability on a laptop ) there are front end bus type compromises made .... and that's what you see. It's a matter of fact physically ( physics ).  Try to just be patient and fix what you can re: settings ( software only, whatever ) and live with the time factor for using cpu vs. gpu.

[Christian.Z]

Your Nvidia GPU will kick in only when you apply effects that require GPU accelerastion