Hi,

So I had a performance issue that when binding a 4k by 4k texture caused a few seconds of low frame rates (60 to 30 fps, vsync turned on). After a few seconds, the fps jumped back to a consistent 60. What's happening here? Is there some kind of internal repetition-based optimization by the gpu? Would be great if there was some explanation to this.

I'm using a GTX650 gfx card btw, and gpu-profiling shows that binding the 4k texture took the most time.

Some corrections: Drawing a huge polysoup covering much of the render target with 4k by 4k textures binded and having a single sample from each of them causes the profile timing to bloat, then return to 60fps timing levels after a few seconds. Clearing large rendertargets also causes this as well.

Are you sure your GPU has enough memory? System will silently transfer data between GRAM and RAM.

Your explanation reminds me of situation I had in a game I played: I had 60 FPS with static scene but whenever I turned camera it dropped to 10 FPS only because PC didn't have enough GRAM to fit all textures at once.

Are you sure your GPU has enough memory? System will silently transfer data between GRAM and RAM.

Your explanation reminds me of situation I had in a game I played: I had 60 FPS with static scene but whenever I turned camera it dropped to 10 FPS only because PC didn't have enough GRAM to fit all textures at once.

It should have enough, gtx650 has 1gb of dedicated ram and I have very few assets in gpu(~20-30 textures, most of them 256x256). But to confirm, is there a way to find out how much memory is left in GRAM?

4k x 4k texture is about 67 MB (depends on format), so it's not exactly small. My GPU has 2 GB dedicated and video player eats 1.5 GB. Most software does wonders at memory consumption. Especially Chrome.

You can use this tool but I'm not sure how accurate it is: https://technet.microsoft.com/en-us/sysinternals/bb896653.aspx