M6 with OS 27 Golden Gate | Slow rendering issues

Topaz Photo Bugs and Issues
1

Topaz Photo – Denoise extremely slow / apparent processing hang on Mac mini M6

Subject: Topaz Photo 1.5.1 – Denoise extremely slow on Mac mini M6 / macOS 27 – GPU/CPU barely utilized

Dear Topaz Labs Support Team,

I am experiencing a serious performance problem with Topaz Photo on a new Mac mini. I would appreciate your help in determining whether this is a compatibility issue with macOS 27, the Apple M6 processor/CoreML, or Topaz Photo itself.

System

  • Computer: Mac mini
  • Processor: Apple M6
  • RAM: 24 GBmacOS: 27.0 (build 26A428)
  • Topaz Photo: Version 1.5.1
  • Previous computer: MacBook Air with Apple M1, 16 GB RAM
  • Main application: Adobe Lightroom Classic
  • Security software: Bitdefender

The same type of processing was previously significantly faster on my M1 MacBook Air with only 16 GB RAM.

Original problem

Initially, when sending an image from Lightroom Classic to Topaz Photo, Topaz Photo would hang during the “Analyzing image” stage.

The problem occurred both when using Topaz Photo from Lightroom and when using Topaz Photo directly.

I therefore performed a number of tests to isolate the problem.

Troubleshooting already performed

1. Lightroom / plugin ruled out

The problem initially appeared when sending images from Lightroom Classic to Topaz Photo.

I subsequently tested Topaz Photo directly as a standalone application, using a local JPEG file.

The same problem occurred.

Therefore, the problem does not appear to be caused by the Lightroom plugin or Lightroom itself.

2. Local JPEG test

I tested with a JPEG stored locally on the Mac rather than on my NAS or another network location.

The JPEG was approximately:

  • 7 MB
  • 32 megapixels

The same extremely slow behaviour occurred.

Therefore, this does not appear to be related to RAW processing, Lightroom, NAS storage or network access.

3. Network test

I also tested the system using an iPhone hotspot instead of my normal network connection.

There was no improvement.

This makes a network/router-related problem unlikely.

4. Bitdefender

Bitdefender was temporarily disabled during testing.

The problem remained.

5. Topaz model cache

I used the Topaz function to clear the model cache.

There was no improvement.

6. Topaz configuration reset

I removed the previous Topaz configuration and application support data.

I also removed old Topaz applications/configuration files that were no longer required, including previous versions of:

  • Topaz Photo AI
  • Topaz DeNoise AI
  • Topaz Sharpen AI
  • Topaz Gigapixel AI

Topaz Photo was then installed again.

7. Topaz Photo 1.5.1

I subsequently installed Topaz Photo 1.5.1 for Apple Silicon.

The current installation is therefore not an upgrade over the previously installed version; it was installed after the previous Topaz configuration data had been removed.

Current situation

There has been an important change:

Image analysis now works very quickly.

When opening the same JPEG, the Analyze stage is now completed almost immediately.

However, when I start Denoise, the processing begins normally but becomes extremely slow during approximately the final quarter of the progress.

The process can remain there for several minutes.

The selected Denoise model is:

Denoise – Normal

It is not Denoise Max.

Very important observation: almost no CPU/GPU utilization

While Denoise was apparently still processing, I checked macOS Activity Monitor.

The result was very unusual:

  • Topaz Photo CPU: approximately 0.2–0.3%
  • Topaz Photo GPU: approximately 0–0.5%
  • CPU time: approximately 32 seconds
  • GPU time: approximately 15 seconds
  • Overall system CPU utilization remained very low
  • The Mac was approximately 88% idle

In other words, while Topaz Photo appeared to be processing the image, the M6 was not being significantly utilized.

This makes it look less like normal “slow processing” and more like the application is waiting for something or has encountered a problem during the final stage of the Denoise operation.

I can provide screenshots of Activity Monitor showing these values.

AI Processor setting

In Topaz Photo Preferences → Application → AI Processor, the available options are:

  • Auto
  • CPU
  • Apple M6

I have tested the configuration with Apple M6 selected explicitly rather than Auto.

The problem nevertheless remains.

Model/log information

I also inspected the Topaz Photo log output.

The log shows that the models are successfully loaded.

For example, the log contains messages such as:

  • Model loaded successfully
  • Selecting backend ... coreml
  • TargetDevices: Apple
  • Using CoreML CPU/ANE mode

The log therefore appears to indicate that the Apple/CoreML backend is detected and that the models themselves can be loaded successfully.

There are some warnings in the log, including:

  • Could not load camera profile
  • Could not open DCP file
  • several Qt/QML binding warnings

but I have not been able to determine whether these are relevant to the Denoise problem.

There is also a neuroserver.json file showing:

  • "enabled": 1
  • models: "wonder-2", "wonder-3"

I would be interested to know whether the Neuroserver/CoreML processing path is involved in the Denoise problem.

Relevant log excerpt

The log shows the following sequence during image processing:

  • Image is successfully loaded
  • Image dimensions are detected
  • CoreML backend is selected
  • Target device is Apple
  • Model files are loaded
  • Model loaded successfully
  • Subsequent image-analysis/model operations complete

The analysis itself is now very fast.

The problem appears to occur after this stage, during the actual Denoise processing.

Comparison with previous Mac

The same general workflow was previously used on my:

MacBook Air – Apple M1 – 16 GB RAM

and Topaz Photo/AI processing was noticeably faster.

The new machine is:

Mac mini – Apple M6 – 24 GB RAM

Therefore, I would not expect the new system to be dramatically slower.

My questions to Topaz Labs

Could you please investigate whether there is currently a known issue involving:

  1. Topaz Photo 1.5.1 and macOS 27
  2. Apple M6 / CoreML
  3. Denoise processing on Apple M-series systems
  4. CoreML / Neural Engine / GPU processing becoming inactive during Denoise
  5. Neuroserver and the Denoise processing pipeline
  6. A known problem where Topaz Photo reaches approximately the final 20–25% of Denoise processing and then becomes effectively idle.

In particular, the Activity Monitor results seem significant to me. If Denoise were actually calculating on the M6, I would expect substantially higher CPU/GPU activity.

Could you please advise:

  • whether this is a known issue,
  • whether there is a recommended workaround,
  • whether a newer/older Topaz Photo version should be tested,
  • whether a specific AI Processor setting is recommended for the M6,
  • and which log files or diagnostic information you need from my system.

Thank you very much for investigating this issue.

Kind regards,

Martin Voss

2

Thanks for flagging this, and for all the details on what you’ve already tested.

Our development team is actively investigating the slow/stalled processing we’re seeing on new M6 machines running macOS 27. Since both the M6 and macOS 27 are new, compatibility issues can occur while we work toward full support for this new hardware and OS combination.

At this time, there isn’t a local processing workaround we can recommend for the M6. For any processes: Make sure to keep the AI Processor set to Auto rather than forcing M6 or CPU.

For now, if you have another supported machine available, you can use Topaz Photo there. If your account includes access to our web apps, you can also access them from your Apps Page while M6/macOS 27 compatibility is being worked on.

We’ll update this thread as we have compatibility updates or a release that addresses the issue.

If you notice any issues other than the slow/stalled processing, please let us know as well. We can relay those findings to the development team as they work on full compatibility.

1 Like