AMD GPU Hardware Accelerated Decoding works w/ console-only but not when run as a service

[tallguydirk]

I have just installed Blue Iris for the first time on this hardware and having trouble getting Hardware Accelerated Decoding to work when Blue Iris is run as a service.

My setup:

• CPU: Ryzen 7 3800XT (8c/16t)
• MOBO: MSI MAG B550 Tomahawk
• RAM: 64GB
• GPU: (AMD) XFX Radeon RX 580 DirectX 12 RX-580P8DFD6 GTS XXX Edition 8GB
• GPU Driver: 30.0.15019.1005 (Radeon Software Adrenalin 22.3.2)
• OS: Windows 10 Pro (64bit) Version 21H1 Build 19043.1645
• Blue Iris Version: I have tried 5.5.5.13 and 5.5.6.11 w/ same results
• Blue Iris Settings: Global Hardware Accelerated Decode set to 'DirectX VA2' ('Direct3D11 VA' didn't work at all for me). All cameras set to 'Default' for HA.

Console only (no service):

• In the Blue Iris Status window, in the 'HA' column, I see "DX" for all 6 of my cameras.
• I can see video decoding activity from Blue Iris in Task Manager and in HWiNFO64

Run as a Service (and then reboot the PC):

• In the Blue Iris Status window, in the 'HA' column, I see "DX" for all 6 of my cameras.
• There is no video decoding activity from Blue Iris in Task Manager or in HWiNFO64.

Is this a known limitation of running as a service w/ AMD GPUs? Or is this a bug? Any ideas for how to fix it?

 

[wittaj]

Lot's of threads on the topic of Hardware Acceleration.

With the substreams option now in use (if you turn it on), HA is no longer needed for BI.

In fact, now with the substreams in many instances the CPU% needed to offload the video to the GPU is more than the savings of CPU%.

Further, since DeepStack has been introduced, HA is has been problematic for many people.

GPU is only needed now if you are using DeepStack and want to improve the return times (but has to be NVIDIA GPU).

Turn off HA and use substreams.

 

[tallguydirk]

Lot's of threads on the topic of Hardware Acceleration.

With the substreams option now in use (if you turn it on), HA is no longer needed for BI.

In fact, now with the substreams in many instances the CPU% needed to offload the video to the GPU is more than the savings of CPU%.

Further, since DeepStack has been introduced, HA is has been problematic for many people.

GPU is only needed now if you are using DeepStack and want to improve the return times (but has to be NVIDIA GPU).

Turn off HA and use substreams.

That is always an fallback option, but I'd really like to get it working if possible...hoping someone could tell me whether this is a known limitation, a bug, or some configuration issue

 

[wittaj]

It really isn't a fallback if the system is more stable and less resource intensive than using HA LOL.

Not many here run AMD, but hopefully someone can let you know if it is a known limitation.