Blue Iris

I'm building a new home PC for BI and a docker setup.

My questions are

Does BI properly use the GPU? This new PC will likely not be used for anyhting else. Should I get a decent video card? Suggestions?
Does BI function better with higher core speeds or more cores at lower speed?

I didn't use a graphics card with BI5 until CPAI. Before CPAI and sub-streams, the Intel cpu graphics was used as an accelerator. It all depends on what you will be doing with BI5, how many cameras you will have etc.

TimG wrote: ↑Sun Jan 28, 2024 10:18 am I didn't use a graphics card with BI5 until CPAI. Before CPAI and sub-streams, the Intel cpu graphics was used as an accelerator. It all depends on what you will be doing with BI5, how many cameras you will have etc.

Good point. I am getting a little old in my experiences. In my aged experience there is practically zero use for GPU if you used a modern intel cpu. A basic machine could handle 10 or more cameras without any issue.

Maybe the newer detection algorithms change that model.

I need to retire from posting.

The GPU will speed up examination time in the BI/CPAI application.
The card needs to be on the supported list and the amount of memory is key. I am using an ancient NVIDIA adapter with only 2GB of memory and have to be careful how many cameras that use AI. Planning to upgrade.

HeneryH wrote: ↑Sun Jan 28, 2024 2:36 pm In my aged experience there is practically zero use for GPU if you used a modern intel cpu. \

Must it be Intel? Would an AMD CPU and a dedicated GPU help?

Homer81 wrote: ↑Mon Jan 29, 2024 10:18 am

HeneryH wrote: ↑Sun Jan 28, 2024 2:36 pm In my aged experience there is practically zero use for GPU if you used a modern intel cpu. \

Must it be Intel? Would an AMD CPU and a dedicated GPU help?

BI5 can use Intel cpu's for acceleration and it is simple to use if you have the right cpu. However the advent of sub-streams made this largely irrelevant, as sub-streams alone saved me more cpu than the Intel cpu acceleration did.

CPAI can use certain Nvidia gpu's for acceleration.

As I said before, it all depends on what you are trying to do here. BI5, or BI5 with CPAI ?

If you plan on using the GPU for AI motion detection, then look into getting a Google Coral Tensor Processing Unit (TPU) instead. It is inexpensive, is better than a high end GPU at AI motion detection, and uses 2 to 4 watts of total power doing so.

douga wrote: ↑Mon Jan 29, 2024 7:03 pm If you plan on using the GPU for AI motion detection, then look into getting a Google Coral Tensor Processing Unit (TPU) instead. It is inexpensive, is better than a high end GPU at AI motion detection, and uses 2 to 4 watts of total power doing so.

If this is accurate (no reason to doubt), it would be one of the most valuable nuggets of insight I have seen posted here.

Thanks for all of the interesting discussion. I'm not using CPAI although just reading about it, I may buy a Google TPU and try

My question is this

If I get a AMD Ryzen 7 with 16 core / 32 thread, will BI be able to properly leverage the multi cores?
if I have 15 camera's set up at 1080p/2k. but at 15 fps, will i need a dedicated GPU? Any suggestions if I do need it?

BI will certainly use all the CPU cores.

One thing to keep in mind is that a lot of us have been around BI for a long time. We have opinions set based on facts that could be 5 years old.

Back in the day, the operating models very heavily relied on decoding the full streams and if you wanted to capture the BI overlays you needed to reencode new streams and write them to disk. This was VERY image processing intensive and you really needed to have Intel CPUs with embedded GPUs. The external GPU cards were not as good.

Things have change a lot in many ways.

When you hear someone old like me give advice, it might be old news.

I think that newer versions of BI and techniques such as sub-stream processing make the dependency on GPU compute less important.

BUT... if you want to use AI, then the equation might change again.

Moving targets.