12 x 8MP - system requirements - discussion

Quindor said:
But I don't run Blue Iris but Nx Witness, it's "a bit" more pricey but if you need to buy a seperate 9900k system to run your NVR, well, it might just b
Click to expand...
Your software is using inferior in camera algorythms and that's explain low CPU usage. Your software is just plain recording with front interface for cameras.
Blue Iris is using more advanced motion detection and algorythms and that's why better CPU is needed.
 
SouthernYankee said:
If using VBR your bit rate depends on the frame rate, iframe value, Also the complexity if the image, motion in the image, rate of the motion in the image. Also the quality of the camera softwarefor doing the compression, it is not very standard.
Shoot a solid white wall, will have a completely different rate, then shoot race cars going around a track.

test it your self. record two images for an hour and look at the file size.
A year of so ago, i tested a 2.1MP camera shooting a white poster board, Then shooting colored streamers on a ceiling fan. The streams file was 7 time bigger than the white board.

Test do not guess.
Click to expand...

When using VBR, if you set the max bitrate too low on a stream that has lots of motion, none of this matters as it'll always be maxing the bitrate and the image will suffer.
That's what I was getting at. Set the bitrate higher, and let VBR do its thing.
 
Quindor said:
But I mainly use it to lower storage needs recording the low stream always and the high stream only when motion is detected.
Click to expand...
You mentioned that you 24x7 record 10x 8MP 15FPS (at 15% load on the system) using the camera's low stream. By low-stream, do you mean sub-stream?

Does Nx Witness show you the total bit rate (or MP/s) load that it's processing? IMO, that's an apples-to-apples way of comparing things.

Quindor said:
I don't know what advanced magic Blue Iris performs or what it's good
Click to expand...
If you have a few minutes, might be worth checking out the quick "Zone" video tutorials, and maybe the motion detection ones.
 
  • Like
Reactions: Quindor and looney2ns
Quindor said:
@aristobrat Yes, I use something called Hi/low recording in Nx Witness where it always records the low/sub stream and only when motion is detected it also records the high quality stream. Motion detection is determined by analyzing the low/sub stream but in my experience the software motion detection in Nx Witness works flawlessly. If you have sensitivity set too low and miss the motion you still always have the video of the low stream so you aren't left without video. Nx Witness does not show you how much MP/s it's doing, but as mentioned, since it's only analyzing the low/sub stream and recording the High stream to disk directly when detection goes off, it will be a LOT less. I don't know if you can setup Blue Iris in the same way because that could even the playing field in regard to server resources required a lot for the Blue Iris side, but I would image people talking more about that then?

I have looked at some videos of Blue Iris, including motion detection and such and although it supports some more advanced features such as trigger wires, other capabilities don't seem too far off. Nx Witness also works with setting zones which you can define sensitivity to, but it does only do pixel detection. For more advanced rules you can use the in camera IVS from Dahua for instance which then doesn't cost any CPU on your NVR box. But even using the software motion detection on the low stream I'm able to use recorded motion searching which works perfectly, you can basically define a region where you want to see motion events and within a second all events pop-up when something happened in that region. That way the zones become a non-issue unless you wish to use it for alerting or manually specify when you don't need high quality footage for movement in that zone.

But don't get me wrong, I'm not saying Nx Witness matches Blue Iris feature for feature, I'm quite sure it doesn't actually in regards to default motion detection that's built-in. Nx however have a complete external system which plugs-in to all kinds of fancy analytics (or you can hook up tensorflow, etc.) but I can't say I've ever used that. I'm saying that I don't understand the system specifications needed for a "generic" Blue Iris install which just needs to do basic recording with maybe some motion detection to save space on disk or quickly search back to when something happened. No expensive CPU + GPU needed for anything like that, but that's all I see in regards to Blue Iris when someone wants to setup a few 8MP cameras.
Click to expand...
NX is likely analyzing the iframes just like you can do with blue iris to drastically lower the cpu consumption. You can setup blue iris to analyse the substream by setting up a duplicate camera and pulling the substream then triggering the hi res camera with it. This will actually be superior to NX as NX cannot (at least as of a few months ago) record high res pretrigger. When the camera is triggered it would have to wait for a key frame then start recording hi res.
Nx, motion detection is weak and limited compared to BI.
You need to learn a bit more about BI and its feature set so that you can compare apples to apples. When using the limit decoding setting in blue iris you will use about the same processing power as NX as I have explained to you in other posts but you have not followed up.
 
  • Like
Reactions: Quindor
Quindor said:
Well, no, I'm not the one comparing these features here at all and this isn't a discussion I wanted to have either. The topic started by someone wanting to know what kind of system he needed to run a software NVR with 12x8MP cameras and I happen to have experience with that running a 24x 8MP software based NVR.

I only highlighted that in my opinion it's insane that BI requires a Core i9 to do that job. I don't pretend to know much about BI either and I'm quite sure it's features in regards to software based motion detection are superior, but that isn't what this topic was about or wanted to have a discussion about. If your requirements are that you need full motion analytics of 12x 8MP cameras, yes, of course you are going to need an insane amount of CPU + GPU power to run that.

But, with that said, in my opinion, I don't understand the reason why most would would want that, I for one don't and am quite happy using motion search and limiting the amount of storage needed with the motion detection that is in Nx witness. I can just run it inside of a virtual machine on my existing infrastructure or buy a sub 200$ box that runs it comfortably, not a 1250$ machine to get started.

Let me reiterate, I don't have anything against Blue Iris, I will probably take another look at it in the future to learn more about it but it isn't the end-all be-all only NVR software out there either. I think Nx Witness and what it excels in is what some people are looking for, I certainly was. Yes it has less features in some regards but with it comes drastically simpler configuration and general ease of use which is worth a lot to some people too.

p.s. Motion triggers in Nx have a default set pre and post recording of 5 seconds each and can be adjusted per camera to your needs, Nx Witness utilizes a 1 minute memory buffer for all streams of all cameras (low + high) so it never needs to wait for anything in that regard
p.s. Limit decoding and also setting up multiple cameras to do low stream trigger high recording seems it isn't an easy setup, but ok
Click to expand...
Once again you are deliberately posting misinformation. You started by this discussion by misrepresenting facts that I now see is deliberate. Blue iris would not require an i9 for that load if you set it up and used it in the same way you setup nx.
You dont need a 1250 machine. That is what you are missing. You are clueless about blue iris and more.
NX CANNOT use pretrigger when triggered via cameras ivs. You are the one who suggested using camera ivs to get around NX's limitations.
Stop posing misinformation. That is for other forums.
Setting up limit decoding is a checkbox as is explained in the wiki, any idiot can setup multiple cameras in blue iris so I assume you can as well.
 
  • Like
Reactions: Jan Werbinski
Walrus said:
4096 bitrate on an 8MP camera would be like setting 1024 bitrate on a 2MP camera.

2MP - 4096
4MP - 8192
8MP (4K) - 16384

From there the VBR will dial it back, although I find when you set it to the highest quality, it doesn't pull down the bit rate very much.

Edit: I'm referring to H264 for the above bitrates.

For H265:
2MP - 2048
4MP - 4096
8MP (4K) - 8192
Click to expand...
================================
Sorry for the delayed response, haven't checked in for a while. Here's a snippet of (and entire) PDF of 264, 264+, 265, and 265+ "suggested" bitrates...
 

Attachments

Fenderman: I am having trouble with my system running multiple 8MP Ip cameras, and I don't want to use the limit decoding feature (doesn't catch footage I need). My CPU is Intel 8700k with SSD for OS and 8TB for data storage with 32Gb ram. What steps would I need to take to record in full 4k without bogging my system down and not limit decoding? I have 16 4k cameras at 15 fps.
 
Walrus: I know that. Fenderman insinuated there is a way around a high hardware requirement by using the subframe footage to trigger the main video. I want to know more on how this works.
 
Ok, that would have been a useful tidbit to include in your question. If you are referring to this that Fenderman said: 'NX is likely analyzing the iframes just like you can do with blue iris to drastically lower the cpu consumption' that is what the limit decoding feature does.

Regarding the next part about cloning the cam, pulling the substream, and using that to trigger the main stream, I'm guessing he means you would continuously record the substream, and once it sees motion, you use it to trigger another camera to start recording, which is the same camera but the main stream.
@fenderman
 
Last edited:
  • Like
Reactions: aristobrat
Walrus: Right now I have it setup where blue iris records the substream and a dahua nvr records all the mainstreams in h265 24/7. When an issue comes up on blue iris and I need more detail, I match the timestamp to the nvr for more clarity. Would be nice if I didn't have to use the nvr at all.
 
Last edited:
Then try duplicating a camera, and make the main stream the input. Set it to only record only on trigger, but don't set any motion triggers for it. Make it part of a group, and then have the motion trigger from the substream camera trigger the group. That will make the main stream camera start recording.

I haven't done this, but I'm thinking that's how you'd do it.
 
  • Like
Reactions: aristobrat
As a data point, I run twelve 4K cameras plus one 4MP camera, all on H264, 10fps, one i-frame/second, on an i7-6700. I maxed out the camera bit rates to whatever the Amcrest UI would allow -- 10240Kbps VBR for the 4K cameras, and 9984Kbps VBR on the 4MP camera. I use BI's motion trigger on all cameras and do *not* have "limit decoding" turned on.

It all runs comfortably at about 45% CPU usage, with CPU temps at around 55deg C using the retail stock CPU cooling fan.

Do you really need 25-30fps? I find 10 is plenty.

I tried doing motion detection on a substream and triggering the HD camera but it didn't work that well -- I missed the triggering event sometimes, plus it's painful to get set up because BI only allows a camera to trigger another camera *group* and not just another *camera* -- so you need 2 cameras and a group for each physical camera.
 
With your vague questions & responses, you are going to get vague answers. First you said '16 4k cameras at 15 fps'. More recently you said you need 30fps.

What are the true specs of the system, and what do you need to achieve in the end?

For card games you likely want to record 30FPS (or higher) and at high quality all the time, not at low quality (substream) then triggered on motion like you were asking about.
 
Last edited:
For what it's worth:

I have 9 cameras, 6 of which are 8MP Dahua's. I run all my cameras 15 FPS H264(H). CPU is Xeon E5-2667 V2. Any system with hardware decoding will have an easier time than mine most likely.

CPU is at 30-35% running as a service, 50% running with the GUI open.

If I switch any of the cameras to H265 I run into delay issues. The camera will start to lag. ie. the timestamp in BI will increasingly get behind the timestamp on the camera, from seconds to multiple minutes and keep growing. More modern CPUs with hardware decoding and/or a gpu to offload to would fix that.

I have 5 IPC-HDW5831R-ZE. I really like them. The turret form factor is easy to install, aim and protect from vandals. After owning several fixed optics cameras I really appreciate having the zoom and focus. Allows fine tuning the angle of view to the installation and not suffering from getting a fixes lens with a bad focus.

I also just got a IPC-HFW2831T-ZS bullet. Depending on location, bullets can be nice and I can really notice the quality improvement from the bigger sensor on this camera vs the IPC-HDW5831R-ZE.

I also own a IPC-HDW5231R-Z and have owned it's replacement IPC-HDW5231R-ZE (it was stolen). These are starlight cameras and do have much better low light performance at night. However, I find the lower amount of pixels too big of a sacrifice. Yes the 8MP are nosier but the greater detail in pixel resolution more than makes up for it IMO. I keep my yard/alley fairly well lit though so YMMV.
 
  • Like
Reactions: aristobrat
Walrus: Not sure why my questions are vague to you. I have the 4k 15fps cameras now. I NEED 30fps. I know if blue iris struggles with 15fps, then going any higher will be a nightmare. I have upgraded a few of my cameras to the 30fps LOREX LNB9232S which is similar to the Dahua IPC-HFW1831E but has two way audio and black in color. These cameras have a superb picture over my previous 15fps turrets, and they also have better low light sensitivity and clarity. I know Lorex is hated around here, but I bought a few used for $150 each, and love them. Just be prepared for limited motion detection options, but for what I do they are working great, BUT...Blue Iris can only use the substream at 704x480 efficiently. I have an 8700k that can only do one camera at 4k30fps in blue iris. I would live with a software NVR solution that just recorded at 4k30fps all the time without motion detection if that was my only option. The remote monitoring of blue iris is hard to beat too.

NYY: Turrets are the best form factor, but I don't see any that do 4k30fps yet. Once you go to that resolution, its hard to go back (even if you don't really need it).
 
Last edited:
You must log in or register to reply here.
Top Bottom