I have a 12 core CPU and an RTX 2060 Super GPU, so state of the art about four years ago. When processing a batch of photos, CPU usage mostly hovers in the 25-35% range, with GPU in the 3-10% range. I conclude that Photo AI could be doing 4 photos in parallel before hitting diminishing returns, and finish almost 4x as fast.
We will be working on improving our efficiency with newer graphics cards soon.
It’s also possible that your VRAM is used up and Topaz Photo AI is not able to bring in more data to process.
Could you open task manager and check your VRAM usage when processing an image in Topaz Photo AI?
I upgraded to a 3090 with 3x as much vram. Still seeing similar behavior – CPU 25-35%, GPU 3-10%. GPU drivers are up to date.
How large are the files you are processing? It seems like the bottleneck here is a CPU step that is happening, either autopilot analyzing the image, import, or export. If your files are small, then the processing step will be quick and the other steps will take longer.
I asked my team and it seems like we could improve the utilization, but it would be very complex and far in the future. For now, this is expected behavior.
They are 24MB Sony compressed raws (24MP) which turn into 100MB tiff files when sending them to Topaz for reasons I don’t fully understand.
To clarify, I am sending them to Topaz from Lightroom using “Open In External.” (I am not using Topaz’s raw processing directly because the results there are still terrible in every camera I have tried. I am baffled that Topaz pushes people to use that by default.)
The TIF file Lightroom creates is uncompressed by default. You can select ZIP compression if you would like to decrease the file size.
Are you using single images or batches of images?
I’m using batches of images.
Same on a Mac M2 Studio from Capture One. I want to process four photos but it is done sequentially. When saving files it processes one file at a time while the others are queued. My machine has 30 GPU cores and 12 CPU cores.