BlueIris on a Core i3

[MSmithIPCT]

HI folks,

I currently have 1 RTSP camera. I installed BlueIris on a laptop yesterday which I can keep dedicated to it. It's a 6 year old Core i3-4000M 2.4 ghz with 16 GB ram. It seems to be working totally fine, except 1 issue (see below). I now want to order 5 more of the same cameras and wondering if I should expect similar performance to what I'm getting. On Blue Iris Update Helper the processor seems to be totally fine with 6 cameras, which I was surprised to see given that it's below the recommended specs. Is that reliable? Also, I saw on another thread in this forum, when someone asked about i3's, someone responded that the number of cameras do not affect performance. Is this true?

The only issue I've had so far is that I tell it to send an mqtt message to my home assistant mosquitto mqtt broker. It's only been a day, but often the message might take 3-5 seconds (never less than 2) before it's reflected in the binary sensor, and at least 3 times (out of maybe 20) it was never reflected at all. Compare this to my wired sensors (not mqtt) which are reflected instantaneously always. I haven't root caused it yet, so don't know who/what's at fault, however could this be an indication that it perhaps BlueIris is actually struggling?

Thanks!

 

[bp2008]

I have no idea about MQTT.

It is incorrect to say the number of cameras you have doesn't affect performance. The problem is, the number of cameras is only one small part. Just as important is the resolution of the cameras and the frame rates you are going to run them at, and other configuration options.

For example in the record you found on biupdatehelper using an i3-4000M CPU, all of the cameras were 1.2 megapixels, most at 15 FPS, one at 12 FPS. Two with Limit Decode enabled. None with direct to disc enabled. Hardware acceleration not turned on. And some of the cameras may have been clones of other cameras, making them take nearly no resources. Biupdatehelper is unable to differentiate clones from separate cameras when it gathers performance data, so that can make a system look more powerful than it really is.

You can probably get that laptop to go a long way if you tune everything according to this guide, especially the part about sub streams. But do remember that is a relatively weak CPU and most laptops aren't really meant to be run heavily 24/7. The cooling isn't very good, and it can get loud. So you will want to be as efficient as possible in your configuration.

 

[MSmithIPCT]

I have no idea about MQTT.

It is incorrect to say the number of cameras you have doesn't affect performance. The problem is, the number of cameras is only one small part. Just as important is the resolution of the cameras and the frame rates you are going to run them at, and other configuration options.

For example in the record you found on biupdatehelper using an i3-4000M CPU, all of the cameras were 1.2 megapixels, most at 15 FPS, one at 12 FPS. Two with Limit Decode enabled. None with direct to disc enabled. Hardware acceleration not turned on. And some of the cameras may have been clones of other cameras, making them take nearly no resources. Biupdatehelper is unable to differentiate clones from separate cameras when it gathers performance data, so that can make a system look more powerful than it really is.

You can probably get that laptop to go a long way if you tune everything according to this guide, especially the part about sub streams. But do remember that is a relatively weak CPU and most laptops aren't really meant to be run heavily 24/7. The cooling isn't very good, and it can get loud. So you will want to be as efficient as possible in your configuration.

I see. Makes sense. I'm using Wyze Cam V2 with RTSP firmware, which I think is 2.1 Megapixels since it's 1080p, however I have it set to their SD mode, rather than HD, so perhaps that's lower. I saw that guide yesterday and was using it. I didn't see a way to tune the cam's FPS from the Wyze app though, so I think it's at 30 FPS, but I'll continue to investigate.

So, sounds like perhaps I might not be so bad with 6 cams being that I'm currently doing OK as is. Maybe I'll just give it a shot and cross my fingers.

Thanks!

 

[BORIStheBLADE]

I say give it a try. If it doesn't work out then get different hardware.

 

[MSmithIPCT]

Placed the order! Will report back. Do need to solve my mqtt problem though. May do a separate thread for that if I determine it's BI related since as of now it doesn't appear to be performance related.

 

[samplenhold]

So with your one cam, what is the CPU % utilization?

What disk is it writing to, an internal or USB attached? Is it writing direct-to-disk? Are you recording 24/7 or just on motion?

What is your memory situation? In use vs available?

How hot does your laptop get?

All of these parameters will help you to determine what to expect by adding 5 more cams.

Heat will probably be a problem. It may help to buy a small dorm-type refrigerator and keep the laptop inside

 

[MSmithIPCT]

I was down to as low as 2% CPU after adding the intel quicksync option, but seen as high as 20% prior. I have to monitor more to get a better sense of where it currently stands.

24x7 continuous recording is my desire. I'm not certain I have that configured correctly though. It's on my todo list for tonight.

I had 12GB free of ram with 4 in use

Laptop was of normal temp to the touch. I plan to keep it in my basement which is pretty cool. I'll monitor that though.

New cams come tomorrow so will update after getting them all installed and setup.

Thanks

 

[IAmATeaf]

What do you for storage, normal drives in laptops are notoriously slow?

 

[MSmithIPCT]

Well, I'll have to see what works, but my intention is to do 24x7plus motion detection based clip recording with regular automatic backups to S3. But if it turns out that won't work, might have to adjust the plans or look into an external SSD or network based recording. Not sure if Blue Iris offers some options which might help out as well, like write to memory and flush to a network storage.

 

[Holbs]

do explain what you do with MQTT & Home Assistant. I have both but have not really dived into things yet, even with Node-Red (summer project #34). A couple other members here also dabble with HA. There is a dedicated Home Automation subforum here
I would guess....and i3 4th gen cpu might struggle with 5 cameras recording continuously depending on 2/4/8MP camera bitrates. But then again, my i7 4th gen had no problem with 14 camera's. Who knows!
Yes, you can offload clips to NAS or external SSD. However, do search the forums here about folks that have. Problems do arise. I recommend tossing all blue iris related hardware in a rack mount server case like I did, and be done with it

 

[MSmithIPCT]

Thanks! I think I've solved my mqtt delays by adjusting how long it takes to trigger a motion detection and also retrigger. Now it's seeming more instant as it should be

For how you can use it, motion detection occurs in BI, it sends MQTT message to HA. HA grabs the image. (there's some Node Red in here as well). You now have a choice here. You can send that image to a local DeepStack server or you can send it to Amazon Rekognition. Regardless, it will now do person and/or face detection on the image(s). If they come back positive then you take whatever action you like. Send yourself a notification or do more. The idea is to reduce false positives and have the best person/face detection possible.

For me, prior to this setup (using my Arlo cameras), if HA is armed and motion is detected, turn all Google Homes to full volume and say "Motion is detected in the kitchen" and it'll follow you through every room. With this new setup I can say if motion is detected and if it's a person and if the face is not one of my family members, make the announcement and if it occurs again 2 minutes later and alarm has not been disarmed by then, then flash all my hue bulbs red and turn on my siren.

Here's a tutorial on how to set it up (there's links in the comments for a tutorial on replacing Rekognition with DeepStack):

Integrating Blue Iris, Pushover, Node-RED, and Amazon Rekognition

Preface I’ve got a few guides that look fairly similar to this: Blue Iris with Pushover and Blue Iris with iOS. Why? I’m hesitant to nuke old content and there is no easy way to update them since I’ve changed the setup a few times, along with underlying components changing on me. This is the...

community.home-assistant.io