Unusual high usage of GPU in Firefox process.
Unusual high usage of GPU in Firefox process.
I recently faced an unusual problem with Firefox's rendering engine. I think it would be best for the community to tackle this, but this forum seems more lively than Mozilla's discussions. When playing videos on YouTube, my GPU usage spikes above 60%. Deleting that process causes Firefox to restart with much lower GPU consumption, around 20%. Chrome, on the other hand, consistently uses a lot of GPU power, which is why I prefer Firefox. The graphics drivers are currently up to date via Intel's utility and ASUS's update tool. (My device is an ASUS Vivobook notebook with an I5 11300H processor and Nvidia 3050 GPU.) I'm leaving a private YouTube video for you to see the effect. Please share any ideas or let me know if you have contacts at Mozilla. OBS screen recording
It seems it's using either an Nvidia or Intel GPU, but your main GPU is actually an integrated graphics card. You didn't check the performance settings after playing, and I'm not familiar with Firefox Developer Edition unless you need the latest features.
I reviewed it quickly, I know, but it was there. It relies on igpu. The developers edition includes several flags, especially for scrolling features. It's not nightly; it's more of a mid-step beta. You can check the GPU usage in the Firefox technical section. You don't have to revisit the performance tab in Task Manager since I opened the utilisation part in hwinfo, where you can clearly see the large spikes during the first tab reload versus the same page after the process restarts, showing much smaller spikes.
GPU performance metrics aren't always clear-cut. Excluding clock speed makes the numbers less useful. Around 60% utilization between 200-400MHz doesn’t represent much of the device’s peak potential, which you’d expect in gaming scenarios.
) The first image shows the normal behavior, which is the problem, while the second reflects after the process restarted. The minor spikes in the second image are just scrolling events. In the first one, I didn't do anything and it still performed well across charts, usage, temps, and clocks.
Point addressed and acknowledged, but I'm convinced it was moving quickly since the fans got really excited when those spikes showed up. Edit: as I considered it, it matches what we saw, not a good outcome
) The first image shows the normal behavior, which is the problem, while the second reflects after the process restarted. The minor spikes in the second image are just scrolling events. In the first one, I didn't do anything and it still performed well across charts, usage, temps, and clocks.