I am running Krita on a CentOS 7 machine with i7-4790k CPU, 32GB RAM, and an NVIDIA GTX 1080 Ti GPU.

However, when I am painting with large pixel brushes (400px+) on a 32-bit 3840x2160 canvas, the brush lags a lot. I tried with Instant Preview both on and off, and it still lags. With it off, the actual drawing of the strokes lags behind, and with Instant Preview on, the strokes appear quickly, but the actual processing takes time, which, when accumulated, makes Krita unresponsive for a while, until all processing is done.

Any ideas what can I do, to fix this issue?

Check https://docs.krita.org/KritaFAQ#Slow_Brushes -- it's impossible to say at a distance what's up with your system, and whether what you call "lag" would be fine for another person.

Also, you're not specifying which version of Krita you use, what color model and channel depth (32 bit, but is that 32 bit/channel?) your images are, what brush presets you are using (smudge brushes will always be slower than others) -- the only thing that's clear is that with your hardware, performance should be fine.

Hi boudewijn,

Thanks for your reply.. I forgot to mention it is Krita 3.3.3.

I am painting in RGB/Alpha Color Model, and, as I mentioned above, with 32-bit depth, on a 3840x2160 canvas.

For the brushes, I am using pixel brushes. I don't use smudge brushes a lot.

But which pixel brush presets? And 32 bit per channel, or 32 bit per pixel?

How about you try 8bit per channel first? 32bit per channel is 128bits for RGBA. It'd be the first thing I would change when I experience lag issues

@boudewijn
32-bit/channel.

And I am trying multiple of the default pixel brushes such as the Splat_texture, Block_bristles, and Block_fuzzy. Some of them work fine, some not.

@TheraHedwig
I can't go to 8-bit as I am working on 3D textures and images that will need a lot more post processing in a compositing program, so I need as much information as possible.

I tried some of the tips in the Slow Brushes section of the FAQ, and it seems that the Precision and the Spacing improve the performance a lot.

And apart from that, you'll just have to accept that working in 32 bits/channel is challenging for any hardware. You're hitting the memory bandwidth limitations of the CPU.

@boudewijn

I think I can go down to 16-bit float/channel. Saves a bit of processing power without losing a lot of information.

Thank you for your help.