DeNoise 3.4

Adding a little more information. I just ran the same test on that Windows 10 PC with 32GB of RAM, choosing the RAW AI model instead of the Severe Noise AI model. DeNoise AI crashed after processing 53 images.

Here is the error message from the Event Logs:

Windows successfully diagnosed a low virtual memory condition. The following programs consumed the most virtual memory: Topaz DeNoise AI.exe (12376) consumed 33515233280 bytes, vmmem (16216) consumed 629682176 bytes, and MsMpEng.exe (6928) consumed 309280768 bytes.

Adding more data here. After converting the 300 Nikon D850 NEF files to TIF format I was able to have all 300 successfully be processed on all computers. Seem very clear to me that there is something about the NEF files from the Nikon D850 that is not being handled well by DeNoise AI on Mac or Windows.

Even more interesting is that the swap usage on the M1 Macs was now almost nothing, the swap size only got to 1GB and the average swap use during the processing was only 0.6GB. The swap usage on M1 Macs was very high across all of the raw file types I have tested while the TIF format doesn’t seem to need any swap at all.

@adam.mains

btw, I am doing all of this testing of different raw types on different computers in preparation for my next Photo Taco Podcast episode.

Is it possible you have used Nikon Transfer to transfer the images? If you used Nikon Transfer it could of corrupted the files, If that is the case and you don’t still have the images on the memory card for your camera then you will need to use Nikon software to convert the NEF files into TIFF files

Pretty certain the issue is not corrupted images. I use this same set of images in testing of Lightroom and Photoshop. I can convert all of them to DNG and TIF. In fact, as noted above, one of the most recent tests was after converting the Nikon D850 images to TIF format and processing those. All 300 processed without a problem (there was a great side benefit too that the swap usage on M1 Macs wasn’t insane like it is for all of the other raw types).

I am in the midst of testing the images converted to DNG format using the Adobe DNG converter. So far I have seen the same behavior with these files converted to DNG where on an M1 Mac with 16GB of RAM the swap file grows until it reaches 50GB and then MacOS stops the application from continuing and tells me the computer has run out of application memory.

If I process the same files in smaller batches, 100 per patch on Mac and 50 per batch on PC, all of them get processed so long as I close DeNoise AI between batches at which point swap/memory is freed back up to the operating system.

1 Like