Denoise AI, Sharpen AI – Which Graphic Options Are Faster

What this is and is not:

These are my test performance numbers for processing a 60Mb TIF photo file in both Topaz Denoise AI and Topaz Sharpen AI on my late 2018 Mac Mini. I tested using the three different settings available in each application: processing with just the CPU; processing using eGPU; and processing using Intel OpenVINO. 

The performance numbers listed below were collected only for my own information.  I tried to record these performance numbers using standard methods on an equal playing field but they are unofficial and unscientific.  These numbers are in no way to be considered as referring to the performance of any mentioned hardware or software as being good or bad.  They are just what I was experiencing at the time of testing using the equipment available.

https://meanderingpassage.com//wp-content/uploads/images/2020/06/EBM-20110916074845.jpg

Testing Resource Image

(See download link below for original TIF file)

https://meanderingpassage.com//wp-content/uploads/images/2020/06/About_This_Mac.jpg

Testing Computer Platform

https://meanderingpassage.com//wp-content/uploads/images/2020/06/Preferences_DeNoise_AI-1.jpg

Topaz Denoise AI Preference Pane

https://meanderingpassage.com//wp-content/uploads/images/2020/06/Preferences_Sharpen_AI.jpg

Topaz Sharped AI Preference Pane

The setup:

Hardware:  Late 2018 Mac Mini, 3.2Ghz 6 core, Intel i7 CPU, 64Gb memory, internal 512Gb SSD & AKiTiO Node Titan Thunderbolt 3 eGPU Enclosure with Radeon RX 580 GPU

Software:  macOS X Catalina v10.15.5, Topaz Denoise AI v2.2.2, Topaz Sharpen AI v2.0.5

Test Image:

The same 60Mb TIF image was used in all cases.  I randomly selected this image because it seemed detailed enough to give both Sharpen and Denoise applications something to process that would take long enough to capture a time measurement.  

How timing was done:

I used the stopwatch on my iPhone to time from the moment the image processing started at the selected settings until it was completed.   Two runs were made at each setting to confirm the numbers recorded were consistent.

Test Results:

Topaz Denoise AI  – For all runs…Sharpen setting was at 15, Allow graphics memory consumption set to “HIGH” (Enable Intel OpenVINO is available and selectable for Denoise AI. You cannot have both eGPU and OpenVINO selected at the same time – see results of each below)

Topaz Sharpen AI – For all runs…Suppress noise=50, Add grain=0, Allow graphics memory consumption set to “HIGH” (Enable Intel OpenVINO is available and selectable for Sharpen AI. You cannot have both eGPU and OpenVINO selected at the same time – see results of each below)

Notes from the test:

• Changing the Denoise and Sharpen levels seemed to make very little difference in the time it took to process the image.  It was close enough that at least some of the differences in time could be attributed to my own reflexes starting and stopping timing.  
• As expected the eGPU made a substantial difference in reducing the processing time.  The Radeon RX 580 GPU used in the test is one of the lower speed GPUs so there could still be additional performance improvements available with a faster GPU card.
• The surprise for me from these tests is in both Denoise AI and Sharpen AI using the Intel OpenVINO is faster on my MacMini than using the eGPU.  In-app info says that this is available with 6th generation Intel CPU’s or later.  OpenVINO with the Intel i7 with six cores and high memory is faster than the 580 eGPU plus the slight lag time back and forth over Thunderbolt 3.  

If you wish to run the same tests yourself using the same test image file I used, you may download the original 60Mb file by clicking the following button.  Feel free to share your results with pertinent details in the comments below. 

Download Test File 

Note:  The concept for this post was gleaned while reading Mark Graf (Grafphoto.com), Topaz Sharpen AI review.