is there a "best practices" guide for camera setup in BI?

JayBart

Getting the hang of it
Sep 8, 2019
63
27
none
I just built a new setup, running 15 brand new dhaua cameras, with a somewhat dated system (which I'll upgrade at some point) -- I don't want to sacrifice quality, but are there any settings I should be mindful of, that don't really impact quality, but can significantly impact cpu usage (for example, a few revs back, selecting "direct to disk" saved lost of cpu cycles, not sure if that's the case anymore) .

Thanks!
 
Assuming you have already dialed the cameras in (took them off of default/auto settings and using the appropriate FPS and bitrate for the cameras you have), simply follow all of the optimizations in the wiki to maximize BI.

 
ok, thanks.. very helpful!
for whatever reason, I was crusing along at 40ish percent load on the cpu, then just added two today, and getting slammed -- I have all the cameras on 10fps, assuming that's not too much?
 
10FPS is fine.

40% seems high unless you aren't using substreams or it is a really bad computer for BI.

What is your "dated" system. Is it intel and if so, what i number and generation number?

 
  • Like
Reactions: dmcIPCAM
10FPS is fine.

40% seems high unless you aren't using substreams or it is a really bad computer for BI.

What is your "dated" system. Is it intel and if so, what i number and generation number?

so turns out a bunch.. almost half .. didn't have direct to disc encoding on. Turning it on for all cameras dropped me back down to 25%-30% cpu usage, 8% gpu usage.

The system is a 6950x (no quicksync), 64 gigs ram, 2080ti gpu, with a fairly large storage array (6x 22tb seagate iron wolf, and a 24tb raid 0 array of 6x 4tb sd's)
I know -- it's overkill in some areas and complete underkill in others, and not ideal for BI. it's a server I built a few years ago as an experiment to virtualize a gamming system (believe it or not, actually worked), and super reliable, runs months and months on end without a hiccup. it works, and works well now that I realized the config was screwed up, though not a model of efficiency.

That said, it's on my list to build a much newer, efficient later system using a cpu with quicksync, ditch the GPU, likely reuse the mechanical discs and one of the sd's as a boot drive.
 
  • Like
Reactions: dmcIPCAM
yup, that tracks. I'm pretty steady now at 25%. I'm sure I can optimize more -- I'm not using substreams, something I need to dive into
the system is also in a server room and I'm accessing it through a remote desktop, which I'm sure is adding to the issues. but 25% with 15 cameras at 4000x3000 resolution doesn't seem too bad.
 
  • Like
Reactions: dmcIPCAM
as a post script to this, got around to setting up substreams -- for 15 dahaua cameras recording at 4000x3000, with substreams at 720p set up, I'm running at 15-17% consistently, and about 9% on the gpu. Very nice! I may actually keep this hardware.
 
  • Like
Reactions: scoob8000
FWIW, I did what wittaj suggested about a week ago and disabled the hardware acceleration on all my cams. On average my CPU usage stayed right about the same, however memory usage went down a bit. Clock speeds and power consumption stayed about the same too. I have 14 cams, all but 3 of them have substreams, but those are low res (1080) cams anyway.
 
  • Like
Reactions: The Automation Guy
FWIW, I did what wittaj suggested about a week ago and disabled the hardware acceleration on all my cams. On average my CPU usage stayed right about the same, however memory usage went down a bit. Clock speeds and power consumption stayed about the same too. I have 14 cams, all but 3 of them have substreams, but those are low res (1080) cams anyway.
Would that be setting this to "no"?

1698695554725.png
 
  • Like
Reactions: scoob8000
Would that be setting this to "no"?
That is correct correct.

There is a global setting too under the main settings, but I found toggling it didn't stop the hw decode for me. I had to change it in each cam like your screenshot.