[HW Accel Support]: Is Synology DS1621+ supported ?

[speaking-frame]

Describe the problem you are having

Apologies if this is not the correct category to ask this.

I have a Synology DS1621+ (that has an AMD Ryzen V1500B 4-core 2.2 GHz CPU with no integrated GPU as far as I can tell).
On it, I have a VM running Home Assistant.
I've installed 32GB on the NAS (and the Home Assistant VM gets all the CPU & RAM it needs).
I've installed go2rtc add-on (which is not running, as I don't see it is needed) and frigate as add-ons in the Home Assistant.
In Frigate - I've added a single Google Nest Wired camera (see attached config).
I do see the camera appear in Frigate's UI (not sure how it works given that the go2rtc add-on is stopped in HA), but the video feed is very sluggish, the video is almost unusable, and the CPU indicated within Frigate appears to be anywhere between 40%-70%. At some point Frigate add-on crashes, and the logs appear to indicate it killed itself, I presume due to high CPU utilization.

I've attached screenshots of the 'General' and 'Camera' tabs in the system metrics.
I'm not sure why "CPU" shows 49%, including a warning at the bottom right corner "Nest entrance has high FFmpeg CPU usage (41%)" when the "Process Memory Usage" graph shows "just" 9% CPU used.

I've spent hours trying to figure out what's going on, given that I did not turn on detection, so I'm not sure what all the processing is going on (is it just ffmpeg for rendering the video ?), but I figured I might as well ask here if this type of NAS processor even supports something like Frigate or am I out of luck ?

and if not - will purchasing a Coral TPU USB Stick, plugging it to the Synology NAS, exposing it to HA and using it in HA be any help in allowing me to use Frigate ? or does the Coral TPU only help for reducing the detection load part, but given that I'm not even there and struggling with just rendering the camera - than Coral TPU won't help me at all ?

Thanks for everyone willing to pitch in.

Version

0.16.1-e664cb2

Frigate config file

go2rtc:
  streams:
    entrance:
      - nest:?client_id=REDACTED&client_secret=REDACTED&device_id=REDACTED&project_id=REDACTED&protocols=WEB_RTC&refresh_token=REDACTED&device_id=REDACTED

birdseye:
  enabled: false

mqtt:
  enabled: false

cameras:
  nest_entrance:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/entrance
          roles: [detect]
    detect:
      enabled: false
    record:
      enabled: false

version: 0.16-0

docker-compose file or Docker CLI command

N/A (using Frigate Add-On in Home Assistant VM running on Synology DS1621+)

Relevant Frigate log output

2025-10-08 07:32:53.247999198  [INFO] Preparing Frigate...
2025-10-08 07:32:54.180736573  [INFO] Starting Frigate...
2025-10-08 07:33:06.735066797  [2025-10-08 10:33:06] frigate.util.config            INFO    : Checking if frigate config needs migration...
2025-10-08 07:33:06.752744496  [2025-10-08 10:33:06] frigate.util.config            INFO    : frigate config does not need migration...
2025-10-08 07:33:06.927153531  [2025-10-08 10:33:06] frigate.util.services          WARNING : Did not detect hwaccel, using a GPU for accelerated video decoding is highly recommended
2025-10-08 07:33:11.360640020  [2025-10-08 10:33:11] frigate.app                    INFO    : Starting Frigate (0.16.1-e664cb2)
2025-10-08 07:33:11.367066672  [2025-10-08 10:33:11] frigate.app                    INFO    : Creating directory: /tmp/cache
2025-10-08 07:33:11.393581901  [2025-10-08 10:33:11] peewee_migrate.logs            INFO    : Starting migrations
2025-10-08 07:33:11.398913620  [2025-10-08 10:33:11] peewee_migrate.logs            INFO    : There is nothing to migrate
2025-10-08 07:33:11.444735403  [2025-10-08 10:33:11] frigate.app                    INFO    : Recording process started: 342
2025-10-08 07:33:11.444741043  [2025-10-08 10:33:11] frigate.app                    INFO    : Review process started: 344
2025-10-08 07:33:11.470932431  [2025-10-08 10:33:11] frigate.app                    INFO    : go2rtc process pid: 127
2025-10-08 07:33:11.525704065  [2025-10-08 10:33:11] detector.cpu                   INFO    : Starting detection process: 367
2025-10-08 07:33:11.534753825  [2025-10-08 10:33:11] frigate.detectors              WARNING : CPU detectors are not recommended and should only be used for testing or for trial purposes.
2025-10-08 07:33:11.581296081  INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
2025-10-08 07:33:11.584861254  [2025-10-08 10:33:11] frigate.app                    INFO    : Output process started: 383
2025-10-08 07:33:11.713724357  [2025-10-08 10:33:11] frigate.app                    INFO    : Camera processor started for nest_entrance: 416
2025-10-08 07:33:11.754548784  [2025-10-08 10:33:11] frigate.app                    INFO    : Capture process started for nest_entrance: 418
2025-10-08 07:33:12.155270030  [2025-10-08 10:33:12] frigate.api.fastapi_app        INFO    : Starting FastAPI app
2025-10-08 07:33:12.678222977  [2025-10-08 10:33:12] frigate.api.fastapi_app        INFO    : FastAPI started

Relevant go2rtc log output

2025-10-08 07:32:54.177735234  [INFO] Preparing new go2rtc config...
2025-10-08 07:32:54.328735409  [INFO] Got IP address from supervisor: <<<REDACTED>>>
2025-10-08 07:32:54.426769908  [INFO] Got WebRTC port from supervisor: 8555
2025-10-08 07:32:55.416722905  [INFO] Starting go2rtc...
2025-10-08 07:32:55.626267587  10:32:55.626 INF go2rtc platform=linux/amd64 revision=fa580c5 version=1.9.9
2025-10-08 07:32:55.626854409  10:32:55.626 INF config path=/dev/shm/go2rtc.yaml
2025-10-08 07:32:55.627453611  10:32:55.627 INF [rtsp] listen addr=:8554
2025-10-08 07:32:55.627975513  10:32:55.627 INF [api] listen addr=:1984
2025-10-08 07:32:55.628733445  10:32:55.628 INF [webrtc] listen addr=:8555
2025-10-08 07:33:01.817335424  [INFO] Starting go2rtc healthcheck service...

FFprobe output from your camera

Not sure how to run this when Frigate is an Add-On in HA

Install method

Home Assistant Add-on

Object Detector

CPU (no coral)

Network connection

Wired

Camera make and model

Google Nest (Wired)

Screenshots of the Frigate UI's System metrics pages

Any other information that may be helpful

No response

[inkeep-ai-assistant[bot]]

The Synology DS1621+ with its AMD Ryzen V1500B CPU can run Frigate, but there are several important considerations for your setup.

Your high CPU usage (40-70%) is expected because your system lacks hardware acceleration. The warning in your logs states "Did not detect hwaccel, using a GPU for accelerated video decoding is highly recommended"(1). Since the AMD Ryzen V1500B doesn't have an integrated GPU, all video decoding is happening on the CPU(2).

For AMD CPUs without integrated graphics, Frigate supports hardware acceleration via VAAPI using the libva-mesa-driver, but this requires adding the environment variable LIBVA_DRIVER_NAME=radeonsi(2). However, this typically works with AMD GPUs rather than CPU-only systems.

Your current configuration shows you're using a CPU detector, which is not recommended for production use(3). The logs confirm this: "CPU detectors are not recommended and should only be used for testing or for trial purposes"(3).

Regarding the Coral TPU USB stick - yes, this would significantly help your situation. The Coral is specifically designed for object detection inference and would handle that workload efficiently(4). However, the Coral only accelerates object detection, not video decoding(4). Since you currently have detection disabled (detect: enabled: false), adding a Coral won't immediately help with your current high CPU usage.

The high CPU usage you're experiencing is primarily from video decoding and streaming through go2rtc, not object detection. Even with detection disabled, Frigate still needs to decode the video stream for live viewing(4).

To improve performance:

1. Add a USB Coral TPU for when you enable object detection(3)
2. Consider using lower resolution streams for live viewing
3. The CPU load should be more manageable once you're not doing CPU-based object detection

The Synology DS1621+ is capable of running Frigate, but without hardware acceleration for video decoding, CPU usage will remain relatively high, especially with multiple high-resolution cameras.

📚 Sources:

1. Frequently Asked Questions
2. Video Decoding
3. Supported Hardware
4. Recommended hardware

---

Have another question? Just tag @inkeep.