Performance of PL 9.4 super bad - hardware proposal - advice and recommendation welcome!

I take photos with a Fujifilm X-E5 with the latest 40 MP sensor. I am currently testing Photolab 9.4 and am very satisfied with the quality of the results and how PL supports the Fuji-World.

However, the performance on my 2-year-old WIN 11 office notebook (16 GB RAM | Intel Arc Graphix onboard | 1 TB SSD | Intel Core i7) is not acceptable. Exporting with Deep Prime XD2/XD3 takes 2-4 minutes per image. When AI masks are also used, I experience frequent crashes. Headache!!!

I have now found a laptop that I hope will give me reasonable export times (I expect 2-10 seconds per image). It is a Lenovo Legion 7 with 32 GB RAM | SSD 1 TB | GPU GeForce RTX5070 with 8 GB VRAM | CPU Intel Core Ultra 9 275HX. On top it looks not so bad:-)

Does anyone have experience with this hardware and what performance can be expected? Any advice and recommendation more than welcome and highly appreciated?

The problem is probably that your GPU-unit has too little VRAM. At least 12 ort preferrably 16 GB is in fact the real system demands to get it to run as expected.

There are two different maskibg methods

De first is to select a premade masking preset in the drop down list. The problem with that is that it is very demending due to pretty unoptimized local AI-models that consumes at least the doubble systems resources as is for example Lightroom and Capture One.

So test instead to hoover over the areá you want to select and then click. That method works fine if you just apply masks to one image att the time and it use to be five times faster to export too. Usually Deep Prime 3 is a little faster too.

I have a desktop computer with 5070 Ti and 16 GB VRAM and have no problems what so ever either with the masking or exporting but I also had problems with an older card (3060 Ti) with 8GB. We just can hope DXO will optimize their masking models - but don´t hope of too much before the next version that might get released next September ort October.

I have a very similarly spec laptop from Asus, Zephyrus G16 with the same RAM and GPU.

With AI Preset masks it takes just over 30 seconds to export with NR XD2s and about 15-20 seconds with non-preset AI masks (manual AI masks) and same NR.

YOU DO NEED TO SET PL TO USE NVIDIA CARD IN THE NVIDIA CONTROL PANEL UNDER 3D SETTINGS TO PREVENT EXPORT CRASHES.

I find that AI Preset masks are no better than manually setting the standard AI masks with either clicking an area or drawing a box around an object. The export is also much quicker.

Learn to use the mask type that best fits your image and use the AI Preset masks as a last resort.

Many thanks Stenis. I find the same behaviour here. Your positive experience with the 5070Ti is a good indication for me and helps me to make a good decision.

Thank you Keith. The ASUS Zephyrus G16 seems to be a good decision. Even if the export with AI masks lasts more than my expected 2-10 sec. If I will buy such a laptop I will consider your advice regarding the setup of the Nvidia card. On my actual configuration every export using AI masks, regardless the automatic or the manual procedure crashes :-(. By the way: I hope really that the DxO developers work hard on this features to improve them soon.

As Stenis describe, GPU VRAM amount is the key point.

PL9 has minimum requirement’s, like 6GB GPU VRAM, etc.
I like to note: even system memory (16GB) in your system is in the minimum requirement.

Whatever your ‘Intel Arc Graphix onboard’ VRAM amount (ohh, exactly what in your system amount?), may this helps on you:
DxO 9.2.1 + Manual AI mask + DP3 NR Export can work even with 4GB GPU VRAM!
In the nutshell: DP3 and Manual AI masking.

Exporting with Deep Prime XD2/XD3 takes 2-4 minutes per image.

AI Acceleration in GPU mode more-or-less its just depend on your GPU performance (and of course overall system performance) - Like in Export. Note: I suggest not use ‘Auto’ in the Preferences-Performance-AI acceleration but choose manually the GPU. OpenCL also need to be enabled.

The 2-4 minute suggest to me, its runs on ‘CPU only’ mode or OpenCL not enabled, etc.

Thank Andras for your assessment. I agree and know very well based on my recent experience my configuration is very much ‘borderline’. I did several tests regarding ‘CPU only’ or dedicated GPU for processing. The actual low performance graphic card provides the shortest export times compared with the CPU mode. Additional activation of the OpenCL does not bring more performance.

Therefore I want to upgrade to a way better hardware. Since my measurements show a heavy GPU load while exporting using the DeepPrime XD3 X-Trans denoising feature I have a special eye on the GPU performance of my new hardware.

Do you have experience with a 5070Ti oder 5080 graphic card?

1 Like

Personally not, check the forum.

I like to note this DxO support article about nVidia driver version and xx60 series (WinML issue): I’m experiencing issues with AI Masks in PhotoLab 9 on Windows using the latest NVIDIA drivers. What should I do?

Also about some ‘Compute’ or whatever settings can be good (some post a few weeks ago)

1 Like

Update 23.2.2026:

The answer of KeithRJ brought me finally to the decision to renew my hardware. In the meantime I installed a new ASUS ROG Zephyrus G14 - Laptop with the following specs:

  • CPU: AMD Ryzen AI 9 HX 370 w/ Radeon 890M
  • RAM: 32 GB
  • GPU: Nvidia GeForce RTX 5070Ti Laptop with 12 GB VRAM
  • SSD: 1 TB

I reinstalled Photolab, now in Version 9.5 (newest edition) and checked it out carefully. My performance values are now VERY GOOD and the software runs smoothly and fast. Great progress!

My export times for an image, taken with Fujifilm X-E5 (40 MP X-Trans V Sensor) with AI masking, retouche and other adjustments , using DeepPrime XD2/XD3 Denoising lasts now 3-6 seconds instead of 3–4 minutes! Now I feel comfortable to test DxO Photolab in all details. The enhanced GPU is responsible for this huge progress.

Many thx to all of you for your helpful advice!

1 Like