CPU suggestion?

[seth-feinberg]

I have had a blue iris instance running with Deepstack on a Dell Optiplex with a i7-4770K, 32gb of memory and a SSD for a couple years now. I followed this guide: . I set everything up with substreams and it MOSTLY worked, but tbh it was about 90% set up, I had a couple of small children, one right after set up and I never got it working 100 percent idiot proof and whenever I NEEDED to use it always ran into trouble. So now that my youngest is about to enter pre-school and my wife is constantly complaining about not being able to access the cameras, I'm making another attempt to button this up. I'm basically starting from scratch so I don't miss anything, and noticed that our friends at The Hook Up released a new video () which i quickly realized included a GPU. This led me down a rabbit hole and eventually I found this extremely informative post: Choosing Hardware for Blue Iris.

Now, like i said, my old install was ALMOST set up the way I like it, so I can't say for sure I hadn't done something wrong but, at least at the time, I'd never noticed any performance issues. Everything I'm reading now though seems to imply it's pretty underpowered. My PC was never more that super low double digits of utilization and the memory was a non-factor. But I have 16 4k cameras running at my house and there are a few instances in that wiki that peg that AT the high end ("These are better for large workloads, such as if you have more than about 12-16 cameras. ") I plan to again utilize sub streams when I re-do it, and my current plan is to power through with following the guide in the new video as best as I can before migrating to a new more powerful PC, but I am a little confused. I see the calculation table, but I'm not buying a PC with a benchmark of 40k just to be safe (249*100*16cams) and I will again be using substreams. So what's a good buffer to shoot for with 16 4k cameras and substreams?

Also confusing me is this part: "In fact, it can sometimes be complicated getting Quick Sync hardware acceleration to work if you have installed a dedicated graphics card." The tutorial passes right over that aspect and just assumes you are using a GPU, but I'm scared of gotchas.

Any advice on how hard I should go after the PC would be very helpful, and if I should employ a GPU to speed up the AI (and if I should go for more than a 1060) would be very appreciated, thanks in advance!

 

[wittaj]

What cameras do you have? It may have AI built in that is sufficient?

The camera AI is useful to many people, but BI has way more motion setting granularity than the cameras, and some people need that additional detail, especially if wanting AI for more than a car or person. For folks that want AI and alerts on animals or specifically a UPS truck then they need the additional AI.

There isn't really a best practice because every field of view is different and use case and needs are different.

To many here, BI motion without AI is more than adequate for what they do.

To many here, camera AI is more than adequate for what they do.

To many here, using the BI AI adds additional functionality that the above alone can not do.

It comes down to testing with each field of view and which one gives you the most consistent results.


While some of that third party stuff is cool like tagging was it a dog or a bear, I don't need all that fancy stuff. If my camera triggers BI to tag an alert for human or vehicle and BI can accomplish what I need by way of a text or email or push or whatever, that is sufficient for my needs. I just want to be alerted if a person or vehicle is on my property and the camera AI does a fine job with that.

However, I do run BI AI on one camera so that it knocks out headlight shine so that the alert image includes the vehicle. The camera AI will trigger for a car, but the alert image was always just the headlights.

The true test....I have found the AI of the cameras to work even in a freakin blizzard....imagine how much the CPU would be maxing out sending all the snow pictures for analysis to CodeProject LOL. My non-AI cams in BI were triggering all night. This picture was ran through Deepstack (without the IVS or red lines on it) and it failed to recognize a person in the picture, but the camera AI did. This pic says it all and the video had the red box over it even in complete white out on the screen:

And I am doing all of this on a 4th gen computer. If you don't go crazy with BI AI and the camera AI is sufficient for you, your existing computer is probably sufficient if you use the substreams:

My humble i5-4590 still humming along with 30+ cams:

I see people posting about their i7-xxxx system bogging down with less than 10 cameras so I wanted to update my setup. I have ~33 cams running on my BI system: HP Elitedesk 800 G1 tower i5-4590 @ 3.3GHz 16GB ram 12TB total purple drives for storage 30+ total cams currently - 22 cams visible in...

ipcamtalk.com

The wiki as it relates to hardware is a little dated. If you are looking to upgrade, an i5-8500 is more than adequate for your needs.


See this thread on how using just the camera AI may be sufficient for your needs:

Who uses Dahua AI capable cameras? Reliable AI for triggering events? Pro's/con's?


Hardware acceleration isn't needed anymore. That was only needed back before substreams were introduced. The substreams allowed other CPUs without QuickSync to be used. For kicks I disabled it and deleted the driver and my system ran fine LOL.

Around the time AI was introduced in BI, many here had their system become unstable with hardware acceleration on (even if not using DeepStack or CodeProject). Some have also been fine. I started to see that error when I was using hardware acceleration several updates into when AI was added.

This hits everyone at a different point. Some had their system go wonky immediately, some it was after a specific update, and some still don't have a problem, but the trend is showing running hardware acceleration will result in a problem at some point.

However, with substreams being introduced, the CPU% needed to offload video to a GPU (internal or external) is more than the CPU% savings seen by offloading to a GPU. Especially after about 12 cameras, the CPU goes up by using hardware acceleration.

My CPU % went down by not using hardware acceleration. Here is a recent thread where someone turned off hardware acceleration based on my post and their CPU dropped 10-15% and BI became stable.

But if you use HA, use plain intel and not the variants.

 

[seth-feinberg]

Thanks so much for all that info! First I'll answer the one thing I can answer then try to clarify the many things you said that I didn't understand:

What cameras do you have? It may have AI built in that is sufficient?

all 16 are Amcrest UltraHD 4k IP8M-T2499EW

https://www.amazon.com/dp/B0962HN4VR?ref=ppx_yo2ov_dt_b_product_details&th=1

The true test....I have found the AI of the cameras to work even in a freakin blizzard....imagine how much the CPU would be maxing out sending all the snow pictures for analysis to CodeProject LOL. My non-AI cams in BI were triggering all night. This picture was ran through Deepstack (without the IVS or red lines on it) and it failed to recognize a person in the picture, but the camera AI did. This pic says it all and the video had the red box over it even in complete white out on the screen:

This is truly amazing. After seeing this I'm not sure why you don't just run it on all the cameras? just too resource intensive? was that the conclusion?

Hardware acceleration isn't needed anymore.

When you are referring to HA in this section, do you mean Quicksync?

This hits everyone at a different point. Some had their system go wonky immediately, some it was after a specific update, and some still don't have a problem, but the trend is showing running hardware acceleration will result in a problem at some point.

So your basically saying that eschewing QS/HA AND a GPU and thus any type of BI AI will just lead to a more stable install? and that even if I don't, I'll eventually run into stability issues?

My CPU % went down by not using hardware acceleration. Here is a recent thread where someone turned off hardware acceleration based on my post and their CPU dropped 10-15% and BI became stable.

Isn't the CPU % dropping completely expected?

But if you use HA, use plain intel and not the variants.

What other variants are there to avoid? I guess I'd only known about Quicksync (assuming i understood my first question about HA=QS correctly)

 

[Swampledge]

@wittaj always gives great advice. Here’s my perspective:

Don’t feel like you need to upgrade your computer until you prove to yourself that you do. I’m running 11 cameras now on a 6th gen i7, but previously on a 2nd gen. I’m also running CodeProject AI on it, on about 4 cameras. On my older computer, I ran DeepStack on 2 of the cameras. Neither box has a GPU. You decide which cameras you want to use the AI for alert confirmation.

When @wittaj wrote about using camera AI, that means the confirmation is done in the camera, not by your BI computer. So it doesn’t consume computing power. But many of the cameras’ AI will Not alert you to an animal - they are preprogrammed to detect only people or cars.

In other words, see how much you can do with the computer hardware you already have. You might be surprised. After all, money not spent on a GPU could pay for another camera.

 

[wittaj]

OK so those are budget cams on the less than ideal MP/sensor ratio and without AI, so I guess you either update to better cameras or use BI AI LOL.

You may have mistaken my blizzard image - that was with the camera AI and not BI AI - the BI AI was pegging all night for every snow flake that triggered the camera. So almost all of my AI is being done within the camera and not thru BI as the camera AI is truly amazing. I simply have the camera AI trigger BI (instead of BI processing the triggers) and it keeps the CPU% down. I have slowly updated my cameras to all better cameras with built-in AI because that simply works better for my needs.

Yes QuickSync = Hardware Acceleration, more specifically the Intel built-in GPU, or the other GPU options. There are the various options:


In the past before substreams were in thing in BI, hardware acceleration was a must in order to keep the CPU% down because everything was being processed with the mainstream video.

Now that substreams are introduced, it works better than using hardware acceleration for most people. As I said, many have seen their CPU% go DOWN after turning HA off when using substreams because it took more CPU power to offload to QuickSync HA than the savings provided by the QuickSync HA. But if you do not use substreams, then HA is darn near needed once you hit a certain point.

Plus with substreams, it allowed many older computers, and non-intel computers, to be viable options as QuickSync is proprietary to Intel.

But if you plan to run BI AI on all those cameras, you will need a powerful GPU to handle that much processing.

 

[seth-feinberg]

OK so those are budget cams on the less than ideal MP/sensor ratio and without AI, so I guess you either update to better cameras or use BI AI LOL.

You may have mistaken my blizzard image - that was with the camera AI and not BI AI - the BI AI was pegging all night for every snow flake that triggered the camera. So almost all of my AI is being done within the camera and not thru BI as the camera AI is truly amazing. I simply have the camera AI trigger BI (instead of BI processing the triggers) and it keeps the CPU% down. I have slowly updated my cameras to all better cameras with built-in AI because that simply works better for my needs.

...

But if you plan to run BI AI on all those cameras, you will need a powerful GPU to handle that much processing.

Ahhh, my mistake, well that is impressive, but I don't think i'll be upgrading my cameras in the near term so I guess BI AI. So GPU it is then? I'll go through this tutorial as best I can and see where I end up, then I guess maybe shoot for a i7-8700 with a 32gb of memory, SSD and a 1060? That all seem reasonable?

 

[wittaj]

Yeah, as @Swampledge mentioned - do the optimizations on your current machine and see how far it gets you.

Keep in mind you don't need to run BI AI on EVERY camera either. And if you do, you will probably need more powerful than a 1060 GPU!

So you have 16 cameras - does every single one of them need AI to alert you of someone? Basically, only use the AI for the cameras you want to be alerted for.

So maybe with the substreams implemented and 4 cameras using BI AI, you can get by with what you have. I ran Deepstack on 4 cameras on the 4th gen fine.

If you don't need it for every camera, then spend some time tweaking the BI motion to minimize false triggers, but at the same time, if they are cameras that are not alerting you, it isn't as critical if they over trigger a bit since it is just going to store the video for you to look at later.

Many of us got quite good at tweaking the BI motion settings to eliminate most false triggers prior to camera or BI AI. What the AI has done is allow for a little bit of loosening of the settings (let it be a little more trigger happy) and let AI then knock out the false triggers.

 

[seth-feinberg]

Many of us got quite good at tweaking the BI motion settings to eliminate most false triggers prior to camera or BI AI.

So basically, just tweaking sensitivity and masking areas? Probably oversimplied it but I assume that's basically what you mean?

As I go through this tutorial, I'm at the stage where I'm now selecting "Hardware Accelerated Decode" in the Cameras Sub Menu. Would I still select "Intel + VPP" with my 4th gen Intel? Or just "Intel"? Or is this where you're suggesting I set that to "No"?

 

[wittaj]

Yes, by taking some time to set up zones and sensitivity and min object size, you can get pretty good at knocking out the false triggers.

Yes at that location, for 16 cameras, I would choose NO for Hardware acceleration. But if you want to try HA, just use Intel and not any of the other variants.

But I would suggest start with no and see what the baseline CPU% is. Then if you want to try it with a few you can and see what happens. But I think 16 cams with HA on will cause higher CPU than not using it. From my observations here, it appears after 12 cams is when folks usually see HA being worse than not using it.

 

[seth-feinberg]

Got it, i'll give "No" a try to start with...

 

[looney2ns]

But I think 16 cams with HA on will cause higher CPU than not using it. From my observations here, it appears after 12 cams is when folks usually see HA being worse than not using it.

Not my experience at all. I'm running 17 4mp cams, all but one with HA turned on, Without HA, my cpu runs around 20% or higher.
With HA turned on for 16 cams, that drops CPU usage to 7-10% average as well as power consumption drops.
This is with an i7-7700, 16gb ram.
So its definite a YMMV.

 

[wittaj]

Not my experience at all. I'm running 17 4mp cams, all but one with HA turned on, Without HA, my cpu runs around 20% or higher.
With HA turned on for 16 cams, that drops CPU usage to 7-10% average as well as power consumption drops.
This is with an i7-7700, 16gb ram.
So its definite a YMMV.

Seems like so many things are LOL. Especially considering all the different configurations out there. Maybe a newer computer doesn't use as much CPU to offload as an older CPU.

I know mine went down turning it off and so have others, I recall this thread where someone saw a big drop as well turning it off:

Blue Iris Crashing on Startup

When I open the Windows Application BI takes a very long time to load and then eventually would hang up after it completes loading. However the webserver works just fine if I leave it alone. I've restore my entire system to previous dates with no luck. Reset BI to default and without any old...

ipcamtalk.com

 

[Evroul]

Yeah, as @Swampledge mentioned - do the optimizations on your current machine and see how far it gets you.

Keep in mind you don't need to run BI AI on EVERY camera either. And if you do, you will probably need more powerful than a 1060 GPU!

So you have 16 cameras - does every single one of them need AI to alert you of someone? Basically, only use the AI for the cameras you want to be alerted for.

So maybe with the substreams implemented and 4 cameras using BI AI, you can get by with what you have. I ran Deepstack on 4 cameras on the 4th gen fine.

If you don't need it for every camera, then spend some time tweaking the BI motion to minimize false triggers, but at the same time, if they are cameras that are not alerting you, it isn't as critical if they over trigger a bit since it is just going to store the video for you to look at later.

Many of us got quite good at tweaking the BI motion settings to eliminate most false triggers prior to camera or BI AI. What the AI has done is allow for a little bit of loosening of the settings (let it be a little more trigger happy) and let AI then knock out the false triggers.

So og you use camera AI do you just use “sync with camera” in BI settings for alerts etc?

 

[wittaj]

To use the camera AI, you do it in two places. One is on the Camera page in BI you tell it to pull ONVIF triggers. The 2nd place is in the alerts you tell it to alert based on external or onvif depending on your BI version.

 

[Evroul]

To use the camera AI, you do it in two places. One is on the Camera page in BI you tell it to pull ONVIF triggers. The 2nd place is in the alerts you tell it to alert based on external or onvif depending on your BI version.

Thx - secondly as this is new for me what alerts / triggers are you guys using? IM gerning spammed with alerts due to dog walking around kids etc ( so far testing BI on my 2 new cams from Andy before I setup the 6 outside ). I mean I had the Idea the camera could detect my daughters e.g. and only alert on new faces. Or have I seen too many Mission Impossible movies

should I have Ivs on, tripwire, I’m a little lost in all the possibilities. I have BI running on a i7 9th gen so enough horsepower . Saving to WD purple. I have 2 x T54IR-TM-AS inside which go crazy with every bit of movement so before I hang up the 6 T54IR-ZE-s3 I would like to know what you guys have done . And yes we are all different with different use cases. What I want is an unfamiliar dude trespassing it records.

I guess what I am also asking is turn all off in BI ( if not wanting AI in say driveway) and the use @wittajs onvif setting and the letting Andy’s cams do the work. Just wondering if IVS is the way for what I have tried to lay out

 

[wittaj]

Use IVS and turn Motion Detection (MD) and Smart Motion Detection (SMD) off.

Then try it both ways - some fields of view a tripwire is better and some the intrusion box is better.

Yeah, we aren't quite to Mission Impossible yet LOL.

Keep in mind that most of us have found that facial identification is more gimmicky and novelty than anything else. If you have to put in 5 or 10 or 15 or 35 pictures or more of yourself in the system for it to recognize it is you...then you shouldn't expect much.... My success rate was under 5% so I moved on to other things LOL. YMMV

It can work in certain situations like a business that requires everyone to stop in front of the camera and the camera is at head height. Outside of that, the percentage of being accurate is probably not going to be super high. You will get a lot of false "confirmations" doing a search.

Someone here posted once how horrible it was inside his house identifying his neighbors and others as him. Another guy his kids and wife were being tagged as him inside the house.

Unless you spend the big bucks that casinos and airports have LOL.

As always YMMV and some here use it and say it is acceptable for their use case. Based on my experience I wouldn't use it for sending off audible alarms or disarming an alarm or unlocking door or opening a garage door.

Heck even in ideal situations like a business with the camera at ideal height and optimal lighting it fails....

What do you think of Dahua's "Uyghur Warnings"?

Good luck. Wish you no ill will. I'm not here to argue with people simply to contribute when and where I can. This is also followed by lending my views on current technology and how it impacts all of us. It's undisputed that splitting of the atom changed the world but once again man decided to...

ipcamtalk.com

 

[Evroul]

Use IVS and turn Motion Detection (MD) and Smart Motion Detection (SMD) off.

Then try it both ways - some fields of view a tripwire is better and some the intrusion box is better.

Yeah, we aren't quite to Mission Impossible yet LOL.

Keep in mind that most of us have found that facial identification is more gimmicky and novelty than anything else. If you have to put in 5 or 10 or 15 or 35 pictures or more of yourself in the system for it to recognize it is you...then you shouldn't expect much.... My success rate was under 5% so I moved on to other things LOL. YMMV

It can work in certain situations like a business that requires everyone to stop in front of the camera and the camera is at head height. Outside of that, the percentage of being accurate is probably not going to be super high. You will get a lot of false "confirmations" doing a search.

Someone here posted once how horrible it was inside his house identifying his neighbors and others as him. Another guy his kids and wife were being tagged as him inside the house.

Unless you spend the big bucks that casinos and airports have LOL.

As always YMMV and some here use it and say it is acceptable for their use case. Based on my experience I wouldn't use it for sending off audible alarms or disarming an alarm or unlocking door or opening a garage door.

Heck even in ideal situations like a business with the camera at ideal height and optimal lighting it fails....

What do you think of Dahua's "Uyghur Warnings"?

Good luck. Wish you no ill will. I'm not here to argue with people simply to contribute when and where I can. This is also followed by lending my views on current technology and how it impacts all of us. It's undisputed that splitting of the atom changed the world but once again man decided to...

ipcamtalk.com

Ok thx for great explanation as always. The Bigfoot picture ( the snow one ) LOL it shows lines etc. Would you mind showing me one example of your IVS setting on a cam . As for BI I guess the AI thingy has to be unchecked and all else to default but your “use onvif”. Correct or are there any other settings one should have? Lots of questions for a new guy . I also read a lot here have issues with push notifications, mine are working via WireGuard vpn but again I’m getting hammered if I use the Blue Iris app. I tried the DMSS app. Fine for just seeing the cam live, but never ever will I turn on P2P so goes push notifications of an intruder is kinda gone? Might get a separate alarm system for this!

 

[wittaj]

Yeah the snow one was a special case as that is my spotter cam for the PTZ, so I only cared about foot traffic from one direction and not the other.

Most of mine are just a simple intrusion box.

You want to leave a little bit of room around the edges (don't have a tripwire or intrusion box right on the field of view boundary) to get the camera a split second to see the object before it has to decide if it meets the criteria to trigger.

Your push notifications will still work the same, you are simply changing it from BI motion to external/onvif.

 

[Evroul]

Do you use the BI app for IOS / Android or how do you get intrusion alerts when say at the supermarket? Have read a lot around her for my setup and some use pushover also. I’m a little disappointed If I can’t get a notification when not at home. When leaving home WiFi my WireGuard will automatically kick in so I have access to local LAN . Guess if u say using the BI app with vpn is best option I have to make sure I don’t so many alerts / triggers activated I get murderd with notifications

 

[wittaj]

I went the Pushover route as I didn't want to have a VPN open all the time as the BI app would only send a push without the picture when away from home unless you VPN or port forward.

Since you have your WireGuard setup to auto kick in, you won't have an issue.

Getting "murdered with notifications" has nothing to do with how you get the alerts whether it is the BI app or Pushover or email or text message or any other method.

You get slammed with alerts based on setting up to get said alerts.

Do you need an alert for every car that goes past or every person that walks by on the public sidewalk?

So it takes some time to figure out which alerts are important to you and then set it up to get those alerts only. BI records triggers for all the other stuff that I can review later, but if someone gets within my "personal space" I will get an alert.