DeNoise 3.7.0

Thanks fotr the link. Your video clearly shows the differemces on your computer system. However, I have nowhere near the lag you have. The difference beween Gigapixel and DeNoise/Sharpen on my system, while preceptible, is not objectionable as it is about two seconds. On my PC the DeNoise and Sharpen zoom in does slow up a slight bit in the last20%zoom in and as first 20% zoom out where Gigapixel zooms at a constant rate.

Overall, I don’t consider this a “bug” in DeNoise or Sharpen. The speed difference compared to Gigapixel is so minor (about 2 second) I can’t object.

What ? What do you call lag here ?

My mouse wheel settings are the default in Windows. Maybe yours are at a higher speed ?

Also yes, we can clearly see that in Gigapixel the zoom is linear across all the range, while in DeNoise and Sharpen it starts much faster then becomes slower and slower as the zoom level increases, I’m glad you noticed that too.

I have never said that was a bug, here it clearly is a design flaw. In any case, the differences are obvious, and even a difference of two seconds is enough to notice it. The zoom curves are simply not the same in DeNoise and Sharpen, and I strongly believe that it should rather behave the same as with Gigapixel.
The missing “zoom to pointer” alone bothers me badly.

DeNoise and Sharpen look like they are not “up to date” to match Gigapixel, but maybe that is just a design choice ? But then why ? Only a developer could answer this question anyway.

Calm down Louis, I didn’t say you claimed a bug. As to lag, it is what I saw in your video, the cause doesn’t matter to me. Perhaps a dev will answer your concerns.

Hi,

today I had a problem with DeNoise AI.

When I saved a series of Pictures to DNG-Files I got 2 of 82 Pictures with black boxes in the middle of the pictures. The pictures came from Original-RAWs of my Canon 1D X and Canon 5DsR and the noise reduction were calculated with the RAW-Mode in Denoise AI 3.7.0.

What can I do to prevent from this Issue or is it a bug?

Thanks for help,

Mario


Firstly go to Help, Graphics info, press Copy and paste that info here.

Here it is:

Application & Version: Topaz DeNoise AI Version 3.7.0

Operating System: macOS 10.15

Graphics Hardware: NVIDIA GeForce GTX 680MX OpenGL Engine

OpenGL Driver: 4.1 NVIDIA-14.0.32 355.11.11.10.10.143

CPU RAM: 32768 MB

Video RAM: 2047 MB

Preview Limit: 4492 Pixels

I tried it again and again and every time the result is different…



Yesterday night after midnight (right time to dissect PC) I’ve upgraded from GT1030 to GTX1060 and updated drivers to latest Studio via GeForce Experience clean install (516.93)

As I’m Affinity Photo user quick test was done with plugins - disaster with Denoise, sometimes Clear passes, every other model crash at the end of processing just before return to AP.

Today tested standalone operation - all OK!

Just installed 3.7.1 RC1 and tested each model once as a plugin - success, everything was OK (GPU was forced in preferences) !!!

I hope release will be even better :smiley:

Graphic info:

Application & Version: Topaz DeNoise AI Version 3.7.1-rc1

Operating System: Windows 10 Version 2009

Graphics Hardware: NVIDIA GeForce GTX 1060 6GB/PCIe/SSE2

OpenGL Driver: 3.3.0 NVIDIA 516.93

CPU RAM: 16317 MB

Video RAM: 6144 MB

Preview Limit: 6612 Pixels

Try output to TIFF and see if that corrects the problem.

It doesn’t :upside_down_face:

Today I made a Test and saved a picture in each format for 10 times.

DNG: 2 of 10 had black boxes
TIFF: 10 of 10 had black boxes

It’s equal to use the CPU or GPU for calculation.

It seems that the “black” boxes are not only black like in this picture:

Ok, seeing you are using 3.7.1-rc1 you will need to go back to support and report the error to the technician that sent you the installer.

Note: I assume that you are using the studio driver.

It’s not me with beta-Version 3.7.1-rc1 :wink:

Hey Mario, I’m the product specialist for DeNoise AI. It looks like what’s happening here is that your NVIDIA GPU video RAM is reaching it’s limit. A few indicators for this: 2GB VRAM, black boxes are usually due to a memory conflict, the larger file sizes for TIFF & DNG.

A simple fix for this would be going to your Processing menu and changing the Allowed Memory Allocation to Low.

Please let me know if that helps! If you are still seeing memory issues (black boxes) with the Processor set to NVIDIA, I would recommend setting the Processor to CPU as @ivkuzmin suggested.

2 Likes

Hi ivkuzmin, hi lingyu.kong,

at first I have to say a very warm thank you for your great support! :ok_hand:

To your questions:

The artifacts will be created in other RAW Files that has been created correctly before and otherwise.

One of my planed fixes for this problem is to switch to a newer Mac (maybe a Mac Pro or newer iMac), because aside of this problem the time for calculating the pictures is much to long.
Until I get a newer Mac I think I can solve this issue by using the GPU and Switch the Allowed Memory Allocation to Low. It seems it works for me.

This evening I made several tests with 10 RAW-files with different sizes, some out of my 18.1 Pixel Canon 1D X (Filesize around 24 MB) and some files out of my 50.6 MPixel Canon 5DsR (Filesize around 62 MB).
At the test I used a tool to watch the VRAM-Usage.
When the computer works at the RAW-Files in DeNoise AI I used some other programs like Photoshop, Apple Safari, PDF Expert, …

When I use the GPU with Allowed Memory Allocation High it’s relatively fast (around 1:15 min for an 50.6 MPixel-Image), but sometimes I get the artifacts in the pictures, not always the same pictures. The VRAM-Usage increases to around 90-94%, but never over 95%.

When I use the GPU with Allowed Memory Allocation Low it’s not as fast as with the preference to High (around 1:30 min for an 50.6 MPixel-Image), but it seems that the pictures has no artifacts. The VRAM-Usage increases to around 80-83%, but never over 85%.

When I use the CPU it’s very slowly (a bit more than 5 min for an 50.6 MPixel-Image) and the last results shows no artifacts. The VRAM-Memoryusage is always at around 61%.
I have to make some further tests, because I think in some tests before I tried it with the CPU and got some pictures with artifacts, too. :woozy_face:

I’m a big fan of using the newest software that I can get all the time, so for me it’s no solution to use an older version of DeNoise AI. The better way is to upgrade the hardware. :wink:

Btw: Every time when I drag and drop a Canon RAW-File to Denoise AI I get a message that the Fileformat is unsupported, but then it will be loaded in Denoise AI.

Bildschirmfoto 2022-07-29 um 00.40.44

My apologies

1 Like

Ok, seeing you are using 3.7.1-rc1 you will need to go back to support and report the error to the technician that sent you the installer.

Note: I assume that you are using the studio driver.

When using a raw file in Denoise AI lens corrections are not done for my m4/3, APS-C, and FF lenses. Lightroom, DXO, darktable, ACDSee, etc. do them though.

Many software packages use the Lensfun database that has correction data on most lenses. ACDSee, On1, Affinity, Aftershot Pro, darktable, etc. use it. Much easier than trying to reinvent it at each company or org. Here is a list of software that uses it (may not be complete):

Here is the location for the current release:

I shoot RAW only have always used DeNoise as a Photoshop plugin but after some testing I noticed that it does a better job on RAW files when used as a standalone. However, when used as a standalone it also creates a really annoying color shift. Out of curiosity I downloaded and tested DxO PureRAW 2 demo version which creates almost no color shift. Any fix for this?
Here are two examples as displayed in Adobe Bridge.
First the original unprocessed RAW image, then DxO and finally Topaz DeNoise.
Canon EOS R:


Canon EOS R7

i am having trouble with this program crashing…as i will try to use it and it closes out and i am getting frustrated with it

Does your PC meet the minimum technical requirements and is your GPU driver up to date?

Please state your PC OS, RAM, GPU and dedicated vRAM along with the driver version.