High CPU Warnings After Hardware Change – More Cores, Comparable or lower Clock Speed

[vajonam]

I’ve moved my Frigate setup from a 12-core machine with a higher clock speed to a 36-core machine with lower per-core speeds. Frigate now shows “high detect CPU” warnings in the UI. I understand this is because it’s measuring per-core usage, but across 36 cores, I’ve got plenty of capacity.

Everything is stable and performing well, but the warnings seem to be misleading. Is there a way to adjust thresholds or suppress these CPU warnings when overall system capacity is more than sufficient? I’d love to keep the system clean without unnecessary alerts.

Pathway is rotated to portrait mode physically and in the camera settings.

Input #0, rtsp, from 'rtsp://go2rtc.domain.com:8554/pathway_sub':
  Metadata:
    title           : go2rtc/1.9.10
  Duration: N/A, bitrate: N/A
  Stream #0:0: Video: h264 (High), yuv420p(progressive), 360x640, 5 fps, 10 tbr, 90k tbn

Input #0, rtsp, from 'rtsp://go2rtc.domain.com:8554/front_sub':
  Metadata:
    title           : go2rtc/1.9.10
  Duration: N/A, bitrate: N/A
  Stream #0:0: Video: h264 (High), yuv420p(progressive), 640x360, 5 fps, 10 tbr, 90k tbn

Input #0, rtsp, from 'rtsp://go2rtc.domain.com:8554/doorbell_sub':
  Metadata:
    title           : go2rtc/1.9.10
  Duration: N/A, start: 0.000000, bitrate: N/A
  Stream #0:0: Audio: pcm_mulaw, 8000 Hz, mono, s16, 64 kb/s
  Stream #0:1: Video: h264 (Main), yuvj420p(pc, bt709, progressive), 640x360, 5 fps, 4.92 tbr, 90k tbn

All of them are decoded using Nvidia hardware decode. I dont understand why randoming some of them are hight. Pathyway and Front are the same brand of Amcrest cameras.

[inkeep-ai-assistant[bot]]

The "high detect CPU" warnings you're seeing are related to the CPU usage of individual camera processing threads, not overall system capacity(1). Your observation is correct—Frigate measures per-core usage, which can trigger warnings on systems with many lower-clocked cores even when total capacity is sufficient(1).

Understanding the High CPU Warnings

High CPU usage in Frigate typically comes from three main sources(1):

1. Video Decoding - The most CPU-intensive task(1)
2. Motion Detection - Pixel-level analysis before object detection(1)
3. Object Detection - Handled by your detector (Coral, GPU, etc.)(1)

Since you're using Nvidia hardware acceleration for video decoding, that workload should be offloaded from the CPU(1). However, motion detection is not hardware accelerated and runs on CPU(1).

Why Some Cameras Show Higher CPU Usage

Looking at your camera streams, there are key differences that explain the varying CPU usage:

• Pathway: 360x640 (portrait, 230,400 pixels)
• Front: 640x360 (landscape, 230,400 pixels)
• Doorbell: 640x360 (landscape, 230,400 pixels)

Even though the pixel counts are similar, portrait orientation and different camera brands can produce different motion patterns and encoding characteristics that affect motion detection workload(1).

Current Limitations

Based on the available documentation, there isn't a built-in configuration option to adjust or suppress the CPU warning thresholds in Frigate(1). The warnings are designed to alert when individual camera processes are consuming excessive CPU resources, which could indicate:

• Inefficient hardware acceleration setup(1)
• Excessive motion in the scene triggering constant detection(1)
• Suboptimal camera configuration(1)

Recommendations

Since your overall system is stable and performing well, you have a few options:

1. Verify hardware acceleration is working - Confirm your Nvidia GPU is actually decoding the streams (check GPU utilization)(1)
2. Optimize motion detection - Use motion masks to exclude areas with constant motion(1)
3. Monitor actual performance - If you're not seeing skipped frames or detection delays, the warnings may be informational rather than critical(1)

The documentation notes that systems with proper hardware acceleration should see minimal CPU load(1)(2), so if you're seeing high per-core usage despite GPU decoding, there may be room for optimization in your motion detection configuration.

📚 Sources:

1. High CPU Usage
2. [HW Accel Support]: Unsure if hardware acceleration actually working, or working as it should be

---

Have another question? Just tag @inkeep.

[vajonam]

I think in my overzealous efforts to detect skunks and squirrels and squirrels I had turned the motion detection way down.. and it was indeed triggering a log, will ramp it down slowly.