Daytime vs. Nightime CPU usage quite different - anyone else notice this?

[Tuckerdude]

Hey folks....

So I've managed to get my systems all nicely tuned (with alot of help from this forum), and now have my BI installs running at a nice CPU load of between 35% and 45% depending on the machine (i9 to i7 respectively).

However, I've noticed something odd...those numbers spike up by 10-15% when night-time rolls around all the cameras (33 of them) switch to IR mode. Still within an acceptable number, but it just seems strange to me.

Why would the load increase when (in theory), there's less information to process...as in a color image vs a grey-scale one? If anything, I would have thought the reverse? Now, I get that the stream IS the stream regardless of the mode the camera is in...but if that were the case, then the load should not change at all.

Anyone have a thought or theory they could share?

Thanks!

 

[Dasstrum]

Hmm... my systems do exactly the opposite. I go from 55-60% during the day to like 35-40% at night.

Do you use profiles and have it automatically change to a different one at night?

 

[Tuckerdude]

Hey Dasstrum...

No, I've not really utilized profiles so far. I know the concept...but can you tell me what you do in your system? Odd that you are getting the reverse affect though!

 

[bp2008]

Sometimes the stream will reduce in frame rate from where you've set it, at night, but only if you had a really high frame rate to start with, and a slow exposure time at night. And that would cause the load to be reduced (not increased!).

Have you set up a schedule and multiple profiles in Blue Iris? Maybe you have a night profile and different settings in it (like continuous recording without direct to disk or something). Edit: Haha someone else already suggested this. Oops.

 

[Dasstrum]

Hey Dasstrum...

No, I've not really utilized profiles so far. I know the concept...but can you tell me what you do in your system? Odd that you are getting the reverse affect though!

I use day/night profiles a couple ways:

Day Profile:
"Cancel shadow" enabled
Lower motion sensitivity
Different Zones
Picture/Text alerts to my phone

Night Profile:
"Black and White" enabled
Higher motion sensitivity (because camera visual range is reduced)
Different zones
Disable most if not all my alerts to my phone
"Object Exceeds" setting is much lower to help cut down on bugs and stuff flying in front of the camera

 

[Tuckerdude]

Yeah...I just double checked that! I have no profiles setup at all...

It's just a head scratcher...and I wouldn't care if it was just a couple percent! But it's a big jump

 

[Dasstrum]

Maybe you have a night profile and different settings in it (like continuous recording without direct to disk or something). Edit: Haha someone else already suggested this. Oops.

Too slow! hahh... and Yeah my thoughts exactly

 

[Tuckerdude]

I can see some useful stuff in there Dasstrum. I guess I need to dive into profiles next and see what better performance I can get.

Still doesn't answer why I'm seeing these CPU bumps though

 

[nejakejnick]

The higher CPU usage is probably due to higher bandwidth, caused by a noise due to low visibility in the night.

You probably use "variable bitrate" on your cameras, which can use 50 kB/s during the day when nothing moves and 800 kB/s during the night when everything moves because of the noise.

 

[Tuckerdude]

Heya nejakejnick!

That makes so much sense. I should have thought about that! Given the number of cameras I have and the amount of noise in each during the nite..there's probably a whole lot more changing per-frame causing the spike. Not sure what I can do about that...but I guess the first thing I'll do is take a measurement of bandwidth during the day and compare it again tonight. Then I will try using CBR on my cameras and see what affect that has.

Thanks for chiming in...I DO believe you are on to the answer!

 

[Tuckerdude]

So...I had a chance to test the theory of higher bandwidth and the results are not all that revealing. I used BI Update Helper to take a baseline during the daylight hours and got a number of 2167 MP/s and then ran it again once all the cameras switched to IR mode and the number increased to 2204 MP/s. So not much of a difference!

But here's the interesting thing...as noted in my first post, my i9 setup only jumps by about 7-10 percent, but the i7 pushes up in to the 70% range and spikes into the 90's if I run the Blue Iris UI. In the case of the 90%, it seems like the on-board Intel Graphics starts to get pretty overwhelmed while trying to both encode and display the images. Whereas my i9 with more raw-horsepower and a Titan X GPU in the mix shows only a slight bump in cpu usage at night.

 

[nejakejnick]

Don't look at MP/s, look at kB/s.

 

[bp2008]

Yes, MP/s is megapixels per second which is the amount of decoded image data, not the bit rate.

 

[Tuckerdude]

So are you guys saying....bit rate at the "per camera" level? I can of course go into each camera and set them individually, but how can I measure the changes in the aggregate? Unless I'm missing something here?

 

[nejakejnick]

There is "Cameras" tab in Blue Iris's Status window, which shows you bitrate for each camera and the sum.

 

[Tuckerdude]

Thanks man! super timely as it's still dark outside....took a snapshot of all the cameras and will wait to see the difference once they have all switched over

 

[Tuckerdude]

Hey folks...

So I went ahead and did the compare between night and day, and really don't see any measurable difference. Attached is a screenshot of the BI "Cameras" tab in the Blue Iris Status Window. On the right you will see two columns showing bitrate with IR on, and the column next to it will IR off, labeled "IR Mode" and "Daytime" respectively

I'd be interested to hear any theories!

Thanks

 

[bp2008]

Did you make note of the total bit rate at the bottom of that status window?

 

[bp2008]

You're right though, I don't see a pattern of higher bit rates in IR mode. If anything it is the other way around. However as the sun may have been coming up by the time you snapped the screenshot, it may have already been getting brighter and therefore less noisy.

 

[Tuckerdude]

Hi bp...yeah here are the two numbers:

14954.50 kB/s (IR on)
15404.25 kB/s (Daytime)