I am new to this forum and I am currently trying out PL8.
My PC is rather slow compared to most of the numbers in the thread but … it actually works !!! only requires some patience.
WIN 11, AMD Ryzen 7 5700U
16 GB RAM, 3 GB dedicated to GPU in BIOS
I have also exported “some” photos taken with my Panasonic G9 (rw2 files).
The files had to be processed with Adobes Panasonic G9 dcp colour profile to actually get the DeepPrime XD/XD2s running reliable without blotches and other artefacts showing. I also had to dedicate 3 GB ram to the GPU in a BIOS setting.
But then it managed to process 257 files in 3h 59m 11s
i.e. an average of 56 seconds per photo (DeepPrime XD/XD2s)
DeepPRIME and XD2S processing acceleration is all about the GPU. Are you certain that PhotoLab is even using your GPU? It automatically defaults to your computer’s CPU when the GPU does not meet it’s expectations. In any case, if you’re getting results with XD2S in around a minute per image, while very slow, at least it’s usable especially If you’re prepared to walk away from your computer for a while.
Thanks for the input, I checked and got a factor of 3 longer times when using Auto setting (or CPU). So no great performance by forcing the usage of GPU but at least I can get the export done on a long lunchbreak !
I also reduced the number of simultaneously processed images to 1. This helped to increase stability with only 3 GB dedicated GPU memory. Everything is a balancing act when trying to use older HW and latest and greatest SW.
You will find that many, if not most, of the popular post processing programs available today have one or more features that, like DeepPRIME and XD2S, are AI based and require a powerful graphics card’s GPU to accelerate processing. Some of these features may not even work at all without an acceptable card.
The new Apple Silicon M4 processor series’ Neural Engine more than doubles performance, with 38 trillion operations per second, compared to 18 trillion for the earlier M-series chips. How does this affect DeepPRIME processing speed? I’m wondering how an M4 Mac mini would compare to my M1 Max Mac Studio.
This should double performance on DeepPrime and DeepPrime XD exports. M1 Max and M1 Ultra exports are already so fast (as were Radeon VII exports under Big Sur) that export performance is not a bottleneck. I see times around ten seconds per image with Nikon 45MP files.
What matters more to me is fluidity of the interface. For performance, I still do better to apply noise reduction after all the other editing. Even though NR preview on the whole image is only High Quality[1], it’s still slow. Moreover, NR preview on whole image is inaccurate so better to work without a misleading preview.
I prefer DeepPrime over XD (fewer artifacts, more natural). HQ at low settings can still be useful, especially on black & white as an alternative to adding grain later (instead of smoothing the image with DeepPrime and then adding gain, just use HQ). ↩︎
As an event pro, I shoot a lot of high-ISO (12,800-25,600) images. I basically can’t work on my images without at least applying High Quality NR, because without any NR some of them look like rainbow sprinkles on black paper and I cannot judge whether they’re usable.
PhotoLab runs fine with NR turned on, but most sliders no longer work in absolute real time, there’s a small delay. I prefer instantaneous feedback with sliders that move freely back and forth with changes to the slider instantly reflected in the image.
The well-known French high-tech news and hardware testing site “Les Numériques” recently included DxO Photolab among the software used in its test protocol to compare GPU performance.
Its latest test shows the following graph:
The processing time in seconds is given for a series of “10 photos of 200 Mb under DXO Photolab with PrimeX2D denoising”. This can give you a good idea of the benefits of choosing one GPU over another, and the performance you can expect.
For DxO Photolab, we can already see that the AMD Radeon RX 9070 offers good value for money, better than its Nvidia RTX 5070 competitor. This was not the case with the previous generation of GPU (AMD RX 7700 XT vs. Nvidia RTX 4070 super).
If this site continues to maintain this measurement in the future, it will be interesting to watch for new products.
