Picking a PC for a Blue Iris test run from owned stuff.

[Perimeter]

After reading here for a while, I decided that I will not only check about an NVR but instead give Blue Iris a free test run first. I just read through the hardware demands and wonder what I should pick. I have access to several older computers. I don't like the idea of redoing the install on another machine if I chose BI, hence I would like to test on the machine that would do the job in the end.

Currently, I plan on using only 4MP cameras. I read about preferably using substream but don't like the idea. I would like to secure the footage away from the exposed cams. My naive plan(?) was to actually record all cams at 4mp for 24/7. I plan for 4 cams at the moment. I understand that security footage is not for the family album. 10-15 fps should do the trick for me.

I can dig out an old 4th gen i3-4130 (2C/4T) with 4-16 GB Ram at 1600MHz bandwidth on a B85 mainboard. (this might be the most energy efficient machine)
Or I could fetch an 8th gen i3-8100 (4C/4T) with 4-16 GB DDR-4 at 2400 MHZ. It is actually on a good Z370 mainboard. (hence I could upgrade (buy) an i5-9500 for an extra 2C/2T or most other 9th gen CPU)


What is the verdict? Is the 4th gen machine with 8GB Ram potent enough? It can handle H264 but not H265. And what about my naive plan to record full stream full time?

Thanks in advance for your assessment.

 

[wittaj]

Since the trial is free, give it a try on what you have. A member here was running 50 cams on a 4th Gen at 30% CPU.

The hardware guide is a little dated now that substreams are available. As long as you follow the optimization wiki you are fine - use substreams, but you can still record 24/7 mainstream if you want. But use substreams for the CPU hog computing. We would say this whether you have a 4th generation or 12th gen.

Just to be clear, a 4th generation can run H265, you just cannot use hardware acceleration, but most have found the CPU% to offload the video to the internal GPU is higher than the savings so many don't run hardware acceleration anymore.

Is an i5-3470 too old for BI?

My son passed on to me his old PC. If it could run BI reliably I could move my IP cam's video off of my NAS, which would gain me needed space on that device. I've been running QNAP's QVR-Pro for the past few years. My only cost might be to add memory, I already have a purple storage HD. The PC...

ipcamtalk.com

 

[The Automation Guy]

Personally I would choose option #2 (8th gen) since it is newer, will support Windows 11 if you are forced to move to it eventually, and is likely more power efficient than option 1 (4th gen). I really don't think there is any need to upgrade the CPU however.

 

[Perimeter]

But use substreams for the CPU hog computing. We would say this whether you have a 4th generation or 12th gen.

No clue what that is. I would let the cameras do the detection.

Just to be clear, a 4th generation can run H265, you just cannot use hardware acceleration, but most have found the CPU% to offload the video to the internal GPU is higher than the savings so many don't run hardware acceleration anymore.

You somewhere pointed out that H264 is preferable anyway in most cases, IIRC. As long as the cpu can handle it, I care more for the net power consumption. 24/7 PC and injector(s) will ring up some bill.

Personally I would choose option #2 (8th gen) since it is newer, will support Windows 11 if you are forced to move to it eventually, and is likely more power efficient than option 1 (4th gen). I really don't think there is any need to upgrade the CPU however.

No doubt, it is the better machine. Question is, if the other one is good enough too. Win11 is a good point. I am not sure if it is more economic though. But they probably level out to about the same.


Do I need an Nvidia GPU? Could offer 660 and 960. But these do consume considerable amounts of power in relation to a bare PC.

 

[wittaj]

You will still want to use substreams even if camera is doing detection.

Yes most prefer H264, but some in their application and field of view it makes sense to run H265.

Most here don't let their system update so we don't care about Win11. Our systems are isolated.

Since most have issues with hardware acceleration, a GPU isn't necessary.

If you want to do lot of AI then it might make sense to have a GPU for AI only, but look at all the posts of people having issues with CodeProject AI and. GPU. There seems to be a lot less issues with the CPU version so one should spend the money there than a GPU.

 

[The Automation Guy]

Do I need an Nvidia GPU? Could offer 660 and 960. But these do consume considerable amounts of power in relation to a bare PC.

You do not need a dedicated GPU (unless you start to get into computer based AI). Simply running BI (or another software) to record and view footage can easily be done with the built in CPU graphics. BI will record directly to "disk" and take up very little processing power while recording - even multiple cameras.

I run a i7-6700 with 12 cameras and the CPU usage is between 12-16%. My 6th gen i7 chip is only slightly faster than your 8th gen i3 chip. With 4 cameras I'd expect your CPU usage to be below 10%.

 

[Perimeter]

Most here don't let their system update so we don't care about Win11. Our systems are isolated.

Am I right in assuming that if I want to get push messages to my smartphone, I can't isolate that machine?

You will still want to use substreams even if camera is doing detection.

For what reason would I want that?

 

[The Automation Guy]

For what reason would I want that?

There is no hard fast rule that you have to use substreams. However using substreams has some great benefits. First and most importantly, using substreams actually lowers the computer's CPU usage. This is because what processing that BI does to the streams (for triggering, alerts, etc) is going to require less CPU power when it is using a lower resolution image in this processing. The effectiveness of this processing is not reduced by using a substream however - just the amount of CPU power it takes to accomplish it. In other words, while we might relish our 4k resolution cameras to be able to "identify" a person in our footage, BI doesn't need 4k resolution to determine that an object has crossed a trigger line. That extra resolution just results in more pixels that BI has to process. By using substreams, BI is processing fewer pixels and therefore uses less CPU power to accomplish the same task.

Second, many of us now record continuously by recording the substream by default and the full resolution stream anytime there is a trigger/alert. This saves a ton of HD space vs recording the full resolution steam continuously. In fact, many of us did not record continuously prior to the sub stream option. This is because the amount of HD space required for that wasn't worth it, given that 99% of the time that continuous footage is worthless and doesn't show anything important. Now by recording continuously, the 1% of the time that the system doesn't trigger/alert properly, we still have some footage available even if it is at the substream quality. That's till better than not having any footage available because we were only recording on the trigger/alert and not continuously.

Hopefully that all makes sense.....

 

[Perimeter]

Hopefully that all makes sense.....

I think it does. After playing with my first 5442, I already thought that it might not make sense to record full time, as the thing does detect motion well. If I record all motion and flag for people in addition, I should generate way less footage without losing much of value. If I also capture the time without motion as substream, then that would probably be good enough.

In my specific situation, I don't really need to check the video for anything continuously, cause I will know when the event I look for has happened. So my layman thought was to fully capture 24/7 and when the need arises just look at the time of interest. Minimum maintenance and time invested. And if the consumer grade cameras would work well at night, it would all be done already.

 

[Virga]

....
Hopefully that all makes sense.....

Helpful explanation, thanks.

 

[Perimeter]

I probably go the route of the i3-8100. In any case:
I will probably want at least 2 drives in there, one for system & BI, and one for just footage? The storage drive will be a rotating disk type, likely an old 1 TB at 5200 rpm for testing. I now assume that BI is mostly residing in memory and does not do constant disk calls for program resources. I also assume that I don't have to start up this machine often. I would therefore use an older smaller SSD for system and BI. Something like 60 or 120 GB? And 8GB of ram?
The machine currently only has one 1000 Mbit NIC. I could add an old 100 Mbit NIC for times I want to hook it up to the rest of the house/world. I assume in that case I could have viewing access from my home network and smartphones perhaps?

 

[The Automation Guy]

I think you are on the right track with your thoughts.

As far as the network scheme goes, there are lots of ways to handle it. The biggest "goal" should be to have the CCTV cameras isolated from the internet because there is no reason to have them be able to access the internet. Giving them access to the internet (even if by mistake) at best allows them to "phone home" and share information about your network and at worst allows them to be hacked. Once that happens, hackers would be able to view your camera footage and use the cameras as bots for other larger hacking attempts.

There are two common ways that people tend to handle this camera isolation. One is to have the cameras on their own completely separate network. (This means using a network switch that only has CCTV cameras and the BI computer hooked up to it). If you want your BI computer to have access to the internet or other devices on the network (so you can view your camera feeds using other devices), then you'll have to add a second network adapter. This might be a second NIC, or a wireless adapter if that's built into the motherboard, etc.

The second common way to handle this network isolation is to use VLANs in your overall network structure. This is a more advanced network scheme that requires hardware that can handle VLANs (ie you'll need a router and network switch that can handle VLANs). This isn't hard to implement, but due to the hardware requirements, the first option might be easier to implement. VLANs create "Virtual Local Area Networks" which are network segments that can be (or not be) isolated from each other depending on the rules used to handle traffic. In other words, it is easy to create a camera VLAN that is restricted from communicating with any other VLANs/devices themselves (ie they can initiate any traffic), but are available when other devices communicate with them and therefore other devices can be used to view your BI feeds, manage your individual cameras, etc as desired.