PL 9.7. black screen when I activate DeepPrime

I get a black screen when I activate DeepPrime.

Nvidia 1660 Super

Can you revert to the previous version you were using and if so does this problem go away?

I fear the root of the problem might be this now well below minimum spec GPU.

1 Like

I tested with DeepPRIME 3 AND XD3 on an old PC with a GTX1660 Super and no problems.
The export is slow but it works.

I have a GeForce 1660, and after updating to v970 build 643, I’m having two issues:

1. When I set the processing to CPU and GPU, I get a black preview with the magnifying glass during denoising and a black image when exporting.

2. If I set it to CPU only, the preview image takes about 5 minutes during denoising, and saving takes another 5 minutes.

Everything was fine up until the last version of DxO. Downgrading the NVIDIA driver didn’t help.

That’s a real shame. Without DxO’s denoising, I’d switch to Adobe.

Is there any help here?

Best regards,
Rainer

I have the same problem. Interestingly, the export is marked as successful! - a green checkmark on the thumbnail. Version 9.7 seems to exclude this GPU.

Hi, when OpenCL is active I do not get a black screen and it works, both with GPU and CPU.

1 Like

Hi, when OpenCL is active I do not get a black screen and it works, both with GPU and CPU.

Great! Good finding!

At general I suggest the followings if someone experience issue when DeepPrime rendering enabled (under Preferences → Display) or Loupe issues:

  • Enable OpenCL (if not enabled) - as its helps for @chp. I think OpenCL anyhow necessary for speed up the things.
  • Try with DP3 (and not DP XD3) - seems DP3 use a bit less GPU VRAM than DP XD3. Note: ‘DeepPrime rendering’ use what is setup in the Denoising section: If Standard then its Display with ‘Standard’ NR, if its DP3 than its display with DP3 NR, if its DP XD3 than its display with DP XD3 NR. Loupe is the same.
  • Try to close other applications (free up system GPU VRAM usage). Even small amounts can count (especially for 6GB, 4GB GPU VRAM), so seems good idea to close web browsers, etc. Even the cutie ChatGPT client use a few hundred MB of GPU VRAM.
  • Try with ‘Compatibility mode’ under Performance → Windows ML acceleration.
  • Try to avoid ‘Pre-Defined’ AI masks (Keyword like: People, Sky, etc.). May do double check without any AI mask.
  • Do editing BEFORE Exporting. Once you does Export, quit from PL, start it and continue editing.

Note: Of course, quit from PL before each try. CPU mode expected to works (if you have like 32GB system memory (may 20-24GB also can be fine). Some of the points based on my observations (i have 4GB GPU, so i try many cases to see what’s can be works out with 4GB).

Unfortunately, that doesn’t work for me.

1 Like

After reading this thread – particularly Stuck’s post – I contacted DxO support and asked for help. They told me that my graphics card (GeForce 1660) wasn’t powerful enough. So I bought a new graphics card (12GB Asus GeForce RTX 3060 Dual OC V2 Active PCIe 4.0 x16) and everything now works perfectly in a matter of seconds.

Many thanks to the forum and DxO support!

Thanks you for your advice. Now it works better than ever.