Gigapixel v5.1.6

Download: Windows, Mac
Released September 9, 2020.


  • Multiple stability improvements
  • Exit dialog now matches DeNoise’s and Sharpen’s


  • Program should no longer crash on machines with NVidia 20-series cards.
  • Program should no longer crash on NVidia pre 20-series cards with outdated drivers
  • Removed need to switch to Game Ready NVidia driver
  • Added warning dialog if Face Refinement or CPU processing fails to process
  • Scrolling in the right panel should no longer cut off the “Remove Blur” slider

What is it with the Game Ready driver? According to NVidia the drivers are the same. Are they not?

With the NVIDIA 1650 Max Q it works perfectly again.
My PC crashed after clicking Auto-Adjust (with 5.15)

1 Like

Seems to be a difference in processing speeds, especially with TensorFlow:

5.1.5 …
GPU processing speed = 16.87 sec/megapixel
CPU TF processing speed = 4.66 sec/megapixel
CPU VINO processing speed = 10.76 sec/megapixel
GPU VINO processing speed = 12.47 sec/megapixel

GPU processing speed = 10.46 sec/megapixel
CPU TF processing speed = 45.89 sec/megapixel
CPU VINO processing speed = 11.04 sec/megapixel
GPU VINO processing speed = 11.76 sec/megapixel

Processing 2400x1600 2x
GPU … 2.53
GPU-OpenVINO … 1.04 (calibrate choice, same as previous version … not 5.1.5)

1 Like

Gigapixel 5.1.6 works correct on RTX 2080Ti and 9700K.


Processing will be slower in this version, unfortunately. It’s a trade-off between stability and the processing speed upgrade from the new Tensorflow version 5.1.5 was using.

1 Like

If you’re still experiencing crashes, specifically on Windows, please post here with how it’s crashing and your system specs.

Why does new TensorFlow crash? Is is NV fault?

Nvidia also offers Studio drivers as well as game Ready. Are not the Studio drivers more for 2D & 3D graphic applications? At least that is my understanding based on the 2D & 3D forums I frequent. Care to elaborate?



It was a red herring - to my knowledge it didn’t actually make a difference besides forcing a flush of VRAM memory usage, which would ultimately make Gigapixel more stable for a limited time. We had told a few people to switch to the Game Ready driver to fix the issue (which it didn’t), so wanted to make sure I explained in the changelog that that’s not necessary.

1 Like

This is true. The last version caused some confusion on which would be ideal for Gigapixel, but it’s no longer the case now. You’re free to use whatever driver type you’ve used in the past.

1 Like

Thanks kindly. I’m sticking to the Studio driver as that is what is preferred by the 3D rendering/modeling apps I use. Thanks for the reply…

Thank you for the insights you provide this community! It’s a big improvement over where we were left on our own to speculate about issues in the past.

Why does new TensorFlow crash? Is is NV fault?

It’s a problem with Tensorflow in particular, and my fault for thinking the upgrade would go as smoothly as it did in DeNoise / Sharpen. Both went through a Tensorflow dependency upgrade with no changes to the user experience. Gigapixel is much more complicated on the backend when it comes to it’s dependencies, something we weren’t aware of until issues arose in the last version.

In the coming months, we’ll be getting rid of Tensorflow entirely and replacing it with a more sophisticated AI inference back-end (the same one that VEAI is beta testing currently). It causes the majority of crashes on both Windows and Mac, due to unbound VRAM usage and lack of error checking in the library itself.


Whoa, thank you for such detailed explanation! I’ll be sure to use this knowledge in my experiments with this library. :slight_smile:
The dready (for me with Win7 back then) OpenCV libraries in the past also didn’t fare well I reckon?

I noticed that OpenVINO option set as OFF gives sharper results with this known Viking warrior lady:

With OpenVINO on:

BTW, these banding artifacts are still here in GPU mode - but I’ll wait patiently for a fix. :slight_smile:

I was the one of that persons with driver issues. On my RTX 2060 Super everything is OK now. That was horrible for few days. Thans Emily for your patience.

1 Like

Performance in 5.1.5 on CPU was fantastic - much faster than GPU (AMD 3950 / 2080TI). Is there any chance for this kind of performance in the future versions ?

1 Like

With my standard test image at best quality and using high memory the fastest processing (11 sec) in 5.1.5 was with CPU only, GPU and OpenVINO Off.

The same set-up in 5.1.6 is fastest (9 sec) with GPU On, down from 14 sec in 5.1.5. The CPU time went the other way from 11 sec to 59 sec!

Since I pretty much exclusively use the GPU/OpenVINO On combination I’m happy with the 35% improvement.

Ran 5.1.5 to open an image, then it froze and I forced it to close. Reopened 5.1.5 again (to my knowledge) and started the update. Got the error posted above and cancelled the update. As a result my parallel installation of 4.2.2 (for more than 6x magnification) was removed.

Hi Taylor, @taylor.bishop

There seems to be a big difference between Natural & Man Made, for example here is the same image processed in 2 different ways:

Natural with Face refinement

Man-made without Face refinement

The second image Man-made is not as soft or dull as the Natural with Face refinement and shadows are muted in the Natural processing model.

Note I have reverted to the latest Studio driver and there are no longer the previous problems reported. I have also seen, in other threads, the RTX problem has also gone away. My environment is: