16.3 editing/viewer performance
I am sure there was already post about it, but it has disappeared. With some images (I haven't found any key or clue yet) it takes up to three seconds to reflect any settings like Highlight and so on, that makes the editing almost impossible. The 16.2.5 version on the same hardware didn't have the issue. You can see the animated icon on the toolbar everytime you move a slider or any other settings (crop cursor selected, no AI masking).
It isn't the new AI masking, it happens even for very simple single layer images, but it is progressively worse with more layers (no AI masking again) added. No matter the Hardware Acceleration or Image / Cache settings are. AMD Ryzen 5 3600, 32 GB RAM, GT 1030, Samsung 980 Pro SSD, 4K monitor.
-
Almost everything is significantly slower on 16.3. Not sure how this one was released in such a state.
0 -
Maybe it is related to this? https://support.captureone.com/hc/en-us/articles/14360621972509-Capture-One-Pro-16-3-release-notes#h_01HDGS74MFBWTT9BHZ85ARN1NT
Optimizing the experience by utilizing more hardware, it now feels faster to browse between images, zoom in and out repeatedly, and pan around on an image when zoomed in.
Do we need more powerful GPU to keep the same performance as the old versions? I agree my GT 1030 isn't anything special (I have it because of the passive cooling), but there were no issues in 16.2.5 version.
0 -
Previously imported images need their previews to be updated to the new type manually.
There is an option in C1 Preferences for previews to be optimized for quality or speed.
??0 -
The issue is with the new fresh imported ones.
0 -
I also tried creating fresh previews at various sizes, using the 12.5 and 12.6 engines, same bad performance either way.
Quality, speed, makes no difference. Sometimes it's takes over 3 full seconds for the preview to be updated. I thought it was maybe my hard drive or something, but I tried 3 drives total and it's identical on all of them sadly.
0 -
What about the old "turn off hardware acceleration" option in preferences?
0 -
It is possible that hardware acceleration on GPU needs something much faster than GT 1030. I have RTX 2070 which is about 7 times faster than GT 1030. When I drag sliders quickly (e.g. Highlights), GPU usage goes up to 50-60%. It is possible that new version on C1 *requires* good GPUs, but then I would have expected this to be specified somewhere explicitly. E.g. DxO Photolab 5 specified GTX 1060 or better, Photolab 6/7 says RTX 2060.
Another thing: GT 1030 is really too slow for 4K monitor. My monitor is not 4K but QHD (2560x1440) and C1pro is blazingly fast for all adjustments.
0 -
The 16.2.5 was fast too. The problem is that all the more powerful GPU cards comes with the noisy fans :-(
Yes, it would be nice if the minimum requirements have reflected the updated version.
0 -
It is not too bad - when I do Photo editing, my RTX 2070 has noise level about 29 dBA (the same as idle Ryzen 2700x). It becomes noisy only when I am using Topaz AI (especially for Video AI). 30 dBA is 'whisper' level.
0 -
Here you can download the RAW file that has the issue: https://1drv.ms/u/s!Apg2GA-B4X5p8n9EYOvarAXKPuUo
Import it and perform Auto Adjust. Set the Highlight slider to +100. Move it fast to -90 or so and watch how long it takes to show the change in the viewer. It is several seconds in my case. Task Manager displays it as CPU bound operation, not very GPU intensive. Add some radial gradient mask layers (not the AI masking), it will become even worse.
The 16.2.5 performed well.
EDIT I see a lot of these errors in ImgCore.log.
(ERROR) CTileExecuterOpenCL::Flush queue wait error (0000020F673CCB30, CL_OUT_OF_HOST_MEMORY, CL_SUCCESS) dev 0 RunChain
0 -
- Imported and performed Auto Adjust
- set HL to +100, then fast to -90, no delay
- added three layers, each with its own radial mask (one inverted)
- moving HL and/or Exposure on any of them quickly (from -100 to +100, about 2 times per second) - no delay, GPU usage ~50%, CPU usage about 40%)
1 -
Thank you. It seems as the better GPU is required for the 16.3 version to perform well.
0 -
I'll try upgrade to RTX 4060.
0 -
I have a 4090 and the performance is still largely wacko
1 -
That was a good suggestion, I look at task manager and rotated like a psycho, but nothing was unusual. No components went past 8% usage
1 -
That was a good suggestion, I look at task manager and rotated like a psycho, but nothing was unusual. No components went past 8% usage
I did the same with my Hightlight slider issue and the GPU was below 10% and CPU about 15% with a very short 50% peak. Disk activity below 1% (it is fast SSD). The Windows performance tool is a good suggestion, but not sure if there is anyone interested in posting the data.
It seems as we haven't come to a clear conclusion what the culprit is, probably not the GPU only.
0 -
This could be a timing issue and lock contention. If this is the case, using a more powerful CPU (with more threads) can make the problem even worse. It would be interesting to try two things:
- set power plan to "Best Performance"
- change C1 processor affinity to use less cores
1 -
"Lock contention"?
While I'm still on 16.2.x, I can say that most operations are done on a single thread, so with Ryzen 5 3600 having a single core score at ~130 with current maximum of ~220 on https://cpu.userbenchmark.com/ is at a visible disanvantage, having only about 60% of the performance currently possible.
1 -
I was rather thinking about James Quinn who has Nvidia 4090 and AMD 7950X (16 cores)
1 -
I upgraded the GPU from GT 1030 to RTX 4060 and the performance is much better know. I can no longer reproduce the issue.
0
Please sign in to leave a comment.
Comments
20 comments