I am using Topaz Photo AI 1.5.3 on Windows. System is a water cooled Ryzen 7950x with a water cooled RTX 4090 with 96GB RAM. Running on Windows 11.
I am cleaning up a large number of images.
Only settings I have are Sharpening Standard with a setting of 46 and Subject Only Off.
Also have Recover Faces set to a strength of 80.
Seems painstaking when going thru the images manually. I am converted these and will take months. My GPU is basially dead. Seems to tick up a bit briefly per photo. My CPU is max 4% from Topaz. Memory is at 2.8GB used by Topaz. As mentioned GPU is basically dead even though set to use the GPU.
Any idea what I’m missing? All my drives are NVME PCIe 4.0 and are sleeping.
Even though it still crashes a lot, I tried Sharpener AI and that thing uses the GPU. Not sure Imo if your GPU is basically sleeping with Photo AI like mine even though selected? I find the app goes full screen and just loading a bunch of images there is low utilization but the app is locked until it’s done doing what it needs to. Same activity with Sharpener AI and it’s super responsive still to minimizing or accessing the task bar plus the GPU is pretty highly used.
EDIT: Note I went back to Photo AI 1.4.1 as some say prior versions worked better but same deal. Also if anyone asks, this is a new system so has latest everything for drivers and only Premiere Pro installed besides a couple Topaz apps.
I tested it again right now and loading 300 jpg images took only a millisecond. Processing speed is not great considering the power of the hardware but it could be even worse.
Wouldn’t call that a millisecond however I have figured out what is going on. The less images you load the faster it runs. So if I load 10 images it’s super fast and uses the GPU most of the time. If I load 40,000 images it hangs a lot and each image takes a hot minute and you can see lots of non system resource utilization with occasional spikes in GPU usage for each image. So seems the system is effected by the number of images you pre-load. Now I have zero memory limitations in this case with 96GB and only 2GB used by the application, but clearly an internal app issue with loading that many images. Trying to figure out based on this how to do the same settings I do now thru CLI. Just the CLI doesn’t seem to have a way to do granular settings like thru the GUI.d
Please visit this page, select your operating system (Windows) and click “Find Your System Profile”. Send me the system profile so I can check your system specs. Topaz Photo AI System Requirements
It’s possible that loading in 40,000 images is overloading the application so it’s no longer working as expected.
To my knowledge we have not tested with that many images to know what the behavior would be.
I made a task so my team can check what the expected behavior would be and see if it’s something we can fix.
Thanks for sending this. It seems this is an issue with how Topaz Photo AI handles large batches as your computer is definitely above the requirements.
Is that a typo, or did you really load 40K images? That’s utterly insane and way beyond what any rational QA testing would have emulated. Even 100 at a time is a heavy load. I’ve done maybe 30 at a time.
At the very least Topaz should figure out what the upper limit is and cap the import at that; put it back on the user to do it in batches.
Make a deal with you. if Topaz Video AI could clean up video frames as well as Photo AI I would use it. But it doesn’t. So I am trying to clean up 154,000 frames in batches and then merge them back together again. Not the only one who does this for this reason.
Reality is the way the image loading is handled is flawed and shouldn’t slow down to a crawl encoding of each frame based on how many images are loaded. if it does, load the list in memory but don’t physically load every image if it’s causing issues. Not sure what it’s doing as my utilization is low.
Wow. Ok, I understand the scenario now. You said images, not video frames, so what you’re doing wasn’t apparent. But you’re right, this is something Topaz should address.
Even 100 at a time is a heavy load. I’ve done maybe 30 at a time.