GPU-accelerated rendering twice as slow as without GPU acceleration (r9 290x 4GB)

  • 1
  • Question
  • Updated 4 years ago
1st is rendering is significantly much slower in Camtasia 8.4 compared to before.
2nd GPU acceleration is on by default. I've timed GPU-accelerated rendering and non-gpu-accelerated rendering. GPU-accelerated rendering takes TWICE as long. I have an AMD r9 290x with 4GB of ram.

Can anything be done to improve performance?
Photo of vbguy2013

vbguy2013

  • 3 Posts
  • 0 Reply Likes

Posted 4 years ago

  • 1
Photo of Dave O'Rourke

Dave O'Rourke, Senior Software Engineer

  • 1438 Posts
  • 408 Reply Likes
Most of the time, GPU accelerated rendering is faster, but not always. It does depend on factors like scaling, video fx, animations, etc. As you would expect, these are faster when done by the GPU. If you're just producing a video and you have the canvas dimensions set to the same dimensions as the source video, the GPU doesn't actually help much, and the cost of going across the bus from system ram to GPU memory and back again outweighs any performance gain you get by using this path. So yes, there are cases where disabling the GPU will outperform the alternate path. In most cases, we find that users are producing at smaller dimensions than they record, or doing composition, rotation, translation, etc. that is costly to do on the CPU and takes away cycles from the CPU that could be spent on compressing the video. That's why the default was chosen to use the GPU enabled path.

There are certainly things you can do to improve performance. Here's a few:
1. Keep all source files on a local disk with fast I/O. Avoid storing your source files on a network drive, or in a drive that's cloud sync'd. This will improve the read speed, which can become a bottleneck causing your render times to be longer than necessary.
2. Reduce the amount of scaling that needs to be done. You can do this by recording, editing, and producing at the same dimensions, for example.
3. Use MP4, AVI or TREC source files on the timeline. These tend to decode faster than other sources (e.g. MOV).
4. Use source files that are 30fps or lower. The timeline currently tops out at 30fps. Using higher framerate source files causes heavy load on the CPU, because it still has to decode these extra frames, only to throw them out.
5. Produce to MP4. MOV output is slow due to the fact that QuickTime is single threaded, and slow on Windows.
6. If you have 2 local hard drives, store your source files (the files on the timeline) on one, and change the temp folder to point at the other. During rendering, CS writes to the temp folder, and then moves (not copies) the file to the final location you specify as the last step. So it's not the output folder you choose in the production wizard, but rather the temp folder that determines the speed of the writes during production. If reads (from the source files) are on a different hard drive than the writes, you may see the overall throughput increase.

Here's another thread with some additional discussion about rendering performance:
https://feedback.techsmith.com/techsm...

Hope this helps.

This conversation is no longer open for comments or replies.