intel i5 nuc ... mini nvr

[icerabbit]

I'm still working on expanding things. Received two more cameras in the mean time and another box of cable, and so I will hopefully be hooking up three in a few days. Some other duties have distracted me from the surveillance project. Plus going forward I need to figure out some cable routes where no path currently exists. First ones were easy. Now I'm at the kind where a basement, sill plates, finished rooms, an attic and tight soffits are involved ... and the finished spaces are the 10 ft physical vertical barrier I need to overcome somehow. Which is ignoring the end-to-end cable pulling hurdle a bit too.

Right this minute the NUC with axxon is running 30-35% with live view on a 1920x1080 display, showing 3x 3mp Hikv cams set to 1080p and 1x 720p foscam. Which is kind of interesting that, in that it stays that high at night ven though the visual data is a lot less compared to the daytime color picture. Some cams look mostly black, so a fair amount less visual data, but I guess for the software a pixel is a pixel, when there's some image noise.

Without live view on. When I exit from the axxon environment. It drops down to 5-10%.

I'm curious as well how this will pan out with more cameras, as the intent from the get-go has been to have live view on full-time.

(something happens in recent days between the browser (firefox) and the forum software where paragraphs or blank line spaces are removed from my postings, my apologies) (trying to put them back in using chrome)

 

[dr.]

Right this minute the NUC with axxon is running 30-35% with live view on a 1920x1080 display, showing 3x 3mp Hikv cams set to 1080p and 1x 720p foscam. Which is kind of interesting that, in that it stays that high at night ven though the visual data is a lot less compared to the daytime color picture. Some cams look mostly black, so a fair amount less visual data, but I guess for the software a pixel is a pixel, when there's some image noise.

Without live view on. When I exit from the axxon environment. It drops down to 5-10%.

I'm curious as well how this will pan out with more cameras, as the intent from the get-go has been to have live view on full-time.

(something happens in recent days between the browser (firefox) and the forum software where paragraphs or blank line spaces are removed from my postings, my apologies) (trying to put them back in using chrome)

If you have no client/liveview open, there should be next to zero CPU usage for background recording. The CPU is doing nothing. Disconnect that foscam. Maybe it's trying to do software motion control or some BS like sending terrible out of spec h264 streams/errors/etc. 5-10% background recording is unacceptable.

I think hikvision is defaulted to VBR. That foscam is likely doing something bad. There will be a slight drop at night if you have VBR set on the hikvision, but it won't be earth shattering. No change means you are set in CBR or again, that foscam is screwy.

I would love to know ivms-4200 usage (with only hikvision cameras) once you force sub-stream. I think 1-5% background recording and 5-15% client is possible for up to 8-10 cams.


Make sure you have the latest intel graphics drivers / etc installed and advanced options checked. Do windows update.

 

[icerabbit]

The drivers should be ok. I just built the thing the other day and had the typical pc build thing going on after assembly with os install, drivers, updates, ...

I have a monitoring utility installed to see what's going on at a glance.

With live view off it might have been a tick high last night due to some recording happening. Currently it is running 3-6% with live view off. Apphost 5 and 6 are the top processes. Not sure which cams those relate to. There is one cam that despite motion settings seems to record without motion. I've been in communication with axxon on that, but they're still stuck on why one bullet only gives half the fps and want me to send in wireshark reports which I haven't had time to look at yet.

I'll be more than happy to drop the foscam - it is a cheap indoor PTZ I picked up for a trial and/or cat monitor Ditto on iVMS. I can give it another whirl. The only other things were that iVMS usability for daily review or something after the fact didn't seem all that practical and had some issues where it didn't match some tutorial I had. Less playback controls than advertised. Knew about where I needed to be time wise, took forever to get there ... would have been faster on analog tape with a magnifier. But it is a learning process. And bugs and lack of intuitive design get in the way.

It may be next week before I get to it though. I have a few things lined up.

 

[Razer]

Just saw this thread, I have 9 of the NUC systems going now, and three of them are running my camera rotations in my main office. I'm running 8 55" TVs off of three NUCs, all of the the basic i3 models and they are doing great. One system is running 5 TVs and is running Exacq, another is running just one with Exacq, and the last is running two, one screen Exacq and another running a GPS tracking map.

I've also started switching our office computers to the NUCs in i5 versions and they are running like champs so far. I'll have one on my desk as my primary desktop here in a couple of weeks. I built these using Samsung SSDs and 4 to 16gb of ram depending. Love the so far, and I'll be getting a lot more. I'm purchasing the newer versions that take a full 2.5 inch size hard drive.

Running a full Exacq client on the i3 computers, connected to 78 Exacq servers, and displaying 6 cameras at once, usually 2 or 3mp cameras and the CPU is running 4-8% usage. Perfect. I will have these running as full DVRs here soon also, I have need of 4-5 more in the near future and will likely use these with an external USB 3 harddrive for recording.

My full size Exacq DVRs run on i3 CPUs with 2gb or 4gb of ram, and they run fine so a NUC should be easy. Here is an example of a DVR i just checked running Exacq.

i3 3220 at 3.3ghz and 4gb of ram. 18 IP cameras, 11 Hikvision 3mp, one ACTi 4mp, one ACTi 1mp, one Vivotek 5mp hemispheric, one Axis 4 channel analog encoder running 4 analog cameras a 4 IP channels. There are also 13 analog cameras running on the built in h264 encoder card in the DVR. 31 cameras total, viewing 4 at a time equals about 4-6% CPU usage, viewing all 31 at once is 48% to 52% CPU usage. Love it, all cameras are recording between 8 and 15fps with the majority at 8-10 fps.

This looks perfect for a NUC!

 

[icerabbit]

Those cpu numbers and camera loads sound very favorable for exacqvision. It certainly seems that when the recording function and the viewing function is split over two systems (server & client) that the systems both have power to spare. I may have to keep that in mind. Looks like exacq is $150 for the software with 1yr of updates included. I'm kinda fearful though that there's also a license per camera. At first glance they're not really forthcoming whether you need to pay per camera and then update those too over time. I vaguely recall I may have looked at them last year, software fee reasonable, but then got into the details, and scrapped them from the short list due to annual cost. Trying to be home budget friendly, hardware first, software second, ideally minimal recurring cost after it is all up and running. I still need to learn more about balancing / optimizing camera detection (poor) vs software detection (better ... but cpu cost), fps settings (I've got it set to 20), motion done in axxon, ...

 

[dr.]

Those cpu numbers and camera loads sound very favorable for exacqvision. It certainly seems that when the recording function and the viewing function is split over two systems (server & client) that the systems both have power to spare. I may have to keep that in mind. Looks like exacq is $150 for the software with 1yr of updates included. I'm kinda fearful though that there's also a license per camera. At first glance they're not really forthcoming whether you need to pay per camera and then update those too over time. I vaguely recall I may have looked at them last year, software fee reasonable, but then got into the details, and scrapped them from the short list due to annual cost. Trying to be home budget friendly, hardware first, software second, ideally minimal recurring cost after it is all up and running. I still need to learn more about balancing / optimizing camera detection (poor) vs software detection (better ... but cpu cost), fps settings (I've got it set to 20), motion done in axxon, ...

There's always a per camera license otherwise everyone would run those software packages. Base fee with 1-2 cam licenses + per camera license is how they are usually sold. $150 is the base fee.

Recording should use nothing especially if all advanced features are handled in camera. That is how weak ARM based hardware is in standalone NVR's. The only taxing part is the liveview/client and it should really only be limited to playback. If that is really high on the NUC and you leave it open always, you can almost negate the power savings of a NUC (haswell and broadwell should idle <20W with pico PSU's).

I had high hopes for NUC's especially the baytrail versions, but this is not reassuring. Paying a price premium over USFF just to get a slightly smaller case seems very wasteful.


NUC's may be moot given hikvision is finally making 8-channel units for cheap. ~10W too.

[h=1]DS-7108N-SN[/h]http://www.aliexpress.com/item/-/1970562350.html

 

[Razer]

Those cpu numbers and camera loads sound very favorable for exacqvision. It certainly seems that when the recording function and the viewing function is split over two systems (server & client) that the systems both have power to spare. I may have to keep that in mind. Looks like exacq is $150 for the software with 1yr of updates included. I'm kinda fearful though that there's also a license per camera. At first glance they're not really forthcoming whether you need to pay per camera and then update those too over time. I vaguely recall I may have looked at them last year, software fee reasonable, but then got into the details, and scrapped them from the short list due to annual cost. Trying to be home budget friendly, hardware first, software second, ideally minimal recurring cost after it is all up and running. I still need to learn more about balancing / optimizing camera detection (poor) vs software detection (better ... but cpu cost), fps settings (I've got it set to 20), motion done in axxon, ...

To clarify, the numbers I put up above were a server, that is recording, and the client is running on the computer too, showing all the cameras at once! Usually my servers do a rotation of 4 cameras and do not display all at once, so with all 31 cameras on and recording and 4 in a rotation on the client running on the same CPU I usually average 10-15% usage of the CPU. The server is also hosting the full web client too, as we use that 95% of the time for our usage.

Exacq has the lowest CPU usage of any I've tried by far, but is not for everyone. Cost can be high, but the $150 license is MSRP, you can pay less, and they have a more basic license which is only $50 per camera, and then $10 each additional year. So a 4 camera system for 5 years total of updates would be $360 for example. The basic version will only let you see one server at a time, does not have a map integrated into the software, other little things. We would buy the basic software for our company, except for a major need to see all 78 servers at once and you have to have the higher $150 license for that. Otherwise we'd save a ton. Exacq includes a full web client, you can even search and view recorded video in it if you want, and full app support for all major phones too. I use the app every day on my way home, pull up my cameras and open my gate from the road using the app and a soft trigger. Works great and has much better range (anywhere I have internet on my phone) than my dumb opener did that's for sure! That's all in the cost built in. not saying it's the best, but it was the best, and cheapest, that met my company's needs. Avigilon was my second choice, but was more costly for everything we needed, but I did like it for sure!

Many scrap the software because of the cost, but when I can run it, and run it perfectly well, on a basic $400-$500 computer then the cost is not bad. Some spend $1500+ on just the computer components to run the cheaper software well, so I don't mind paying. It's a business, if I was a home user them maybe blue iris would work fine for me.

Exacq does have a free trial if you want to play with it some, one camera, fully functional and never expires. If you have trouble getting it let me know and I can help you out with that if you want to play.

 

[icerabbit]

Thanks Razed for clarification.

Dr. RE: small form factor premium waste

I hear you. They're not exactly cheap, but then at the same time the nucs are very competitive for their size and price. There are / may be standalone boxes that will be better value and more plug and play; but never having seen and operated one of those particular ones first hand, I didn't want to order something I wasn't going to be happy with, something that was possibly harder to operate or feature limited like some antique vcr with a dos interface. Some previews I've seen don't impress me. Having played with a few in the big box stores was meh. Plus in many cases when I looked at the details of more inexpensive nvrs, they didn't seem geared yet to dealing with 3mp cameras with a decent fps rate.

Hence looking at a computer with software solutions. The choice for the NUC was made as it is a compact dual drive system with low power needs. Vs having some type of slimline or tower or whatever that needs more space and power. The fact that the NUC can hang off the back of a monitor's vesa mount or be easily hidden somewhere and be sight unseen was an important consideration. The price of a nuc was also competitive vs some other solutions on the market. It is also more DIY friendly vs a complete part built system.

It could be that, ultimately, a nuc will be too stressed for recording and live viewing of more than 6 or 8 3mp cameras. But I guess somebody has to try it. Worst case, the NUC can get re-purposed as a regular desktop (or maybe sold, who knows). I can put a regular desktop i7 machine in gear for a while until another solution comes up.

And I fully understand that for some people live view is not important. For others it is. Some people like camera rotation view. I don't for a primary screen. Personal choice. I wanted live view to be available in an instant and idiot proof. One or two buttons. No need to undock a tablet, or phone, login, find an app, wait for it to start, ... which is why I'm redistributing the live view throughout. Hear something. TV on. Done. And yes, this requires a client to be on and thus some additional power to be consumed, but so does the rest of the network and camera kit.

 

[icerabbit]

FYI.
Axxon Next, nothing moving, nothing recording (and I believe only one detection zone on 2 of them)
40% CPU avg live view 9-up with 5 slots filled: 4x 3mp Hikv set to 2mp and 1x 720p foscam
55% CPU avg live view 9-up with 7 slots filled: 6x 3mp Hivk set to 2mp and 1x 720p foscam
63% CPU avg live view 9-up with 8 slots filled: 6x 3mp Hikv set to 2mp and 2x 720p foscam

3MB/s net incoming load

Mosaic tile resolution is indicated to be a quarter of the recording resolution (h/2 x v/2) so 960x540 vs 1920x1080

 

[icerabbit]

Regarding the request what happens with live view without the foscams

Added a couple additional live view screens in Axxon Next. Nice you can switch views on the fly after setting them up.

1x Hikv 3mp @ 2mp is ~16% CPU avg, single large frame size of 1920x1080

Sticking with a grid of 9 for the multiple cam ones, to keep live view tiles smaller as if there were 7-9 in view.

3x Hikv 3mp @2mp is ~22% CPU avg

6x Hikv 3mp @2mp is ~55% CPU avg

 

[dr.]

http://www.notebookcheck.net/Intel-Core-i5-4250U-Notebook-Processor.93564.0.html

Triple check everything because those numbers are pretty terrible if the above is your CPU. Couldn't find "intel nuc D5420WYKH." The only thing that should get taxed is your GPU. 40EU of the HD5000 should be plenty.

1. Record one MJPEG substream (960x540@7fps or whatever you have yours set at) for a few minutes.
2. Close as much background tasks as possible. Now playback that recorded file in VLC or your favorite media player. Open as many instances as you have cameras. e.g. 7 VLC windows playing for 7 cameras.
3. What is the load on the machine?

This should be ridiculously low i.e. <25%.

Baytrail NUC can display 1080p video @ 10% load:
http://www.legitreviews.com/intel-nuc-dn2820fykh-bay-trail-system-review_135053/6

How about just plain 1080p to see if hardware decoding is even working?

 

[icerabbit]

I appreciate your insights and comments.

Right this past few days I've been less than enthusiastic about AN.

Before I went away for a few days, I took a couple of the advanced analytics out on the two primary cameras, hoping to save some cpu cycles. So left things running with basic motion on those and motion on the new ones I added the other day. Only to come back from the trip and seemingly having no recordings. If I put the detection screen up for an individual camera, it'll see me walk or drive through and record supposedly ... but then when I try to review any recording, even if I search for a 30 min span ... zip nada spinning color circle indefinitely.

Axxon support still wants me to do some wireshark tracing, for the one camera, seemingly ignoring a couple other points I raised. Their instructions file was corrupted, so waiting on that.

Oh and they want me to turn windows firewall OFF ENTIRELY to enable remote viewing, when I said it didn't work with the specified port as instructed in their manual. Turning Windows Firewall Off ENTIRELY? ... kind of unheard of I think.

I feel I need to take a break from this testing tweaking not working right stuff. I've gone through the AN manual a couple times. Maybe if you have multiple servers handling a few cameras with some of the advanced video analytics running one can get satisfactory results. I recall now I had to have a minimum of two motion detectors on for it to work, when I first tried it on an old spare pc. That just the basic one wasn't enough.

I may just throw ivms on it for a couple weeks, and take a break.

It should be simple. n Cameras + live view without melting + record as needed.

 

[dr.]

You have to set up each camera to point to a drive/target/dir or it won't record. The video turning red is just the trigger channel being sent and the alarm being logged. IIRC there will be another icon when it is recording. Double check your storage settings.

 

[icerabbit]

Yes, I made sure that each camera is linked to the database in the data section. That was my mistake when I first started two months ago. And when I first started it was actually recording continuously by default. This was my third install of AN and so I've gotten pretty familiar with its quirks and manual. I can get a few things running in 5 min from install completion. It worked fine with just a pair of cameras a few weeks back. Then when I piled on the rest something changed. I just double checked settings on the primary cameras, added the analytics motion area, added motion analysis, made sure to do the + and record to storage db was added. Motion will trigger fine in test mode. Go back to regular view, walk in/out drive in/out same way ... nothing.

Did a repair installation, just in case something happened under the hood that shouldn't have. No dice. It'll show the motion detection film strip. Regular view no recordings, searching for recordings taps out ... spinning indefinitely. The motion detected bar does not populate.

Tried the manual recording to playback in VLC as you requested. Thing goes red. No recording made or found.

And no activity to the external hard drive either, as the little monitoring utility I added to check on the cpu and network load also is set to show hdd i/o.

So, I just gave up and hit uninstall ... I may revisit it on a Core i7.

Giving ivms another try.

 

[icerabbit]

iVMS 4200
6x Hikv 3MP set to 2MP
Live View on with 6 on screen (1 lg, 5 sm)

~30% CPU (26-34), CPU 62C, 3MB/s data I/O

Still have to learn the details with iVMS (getting some errors when I try to apply some settings) and figure out some of the motion details to make it do its thing; but it certainly looks more promising and safer for the little guy, cpu wise.

 

[dr.]

Yes, I made sure that each camera is linked to the database in the data section. That was my mistake when I first started two months ago. And when I first started it was actually recording continuously by default. This was my third install of AN and so I've gotten pretty familiar with its quirks and manual. I can get a few things running in 5 min from install completion. It worked fine with just a pair of cameras a few weeks back. Then when I piled on the rest something changed. I just double checked settings on the primary cameras, added the analytics motion area, added motion analysis, made sure to do the + and record to storage db was added. Motion will trigger fine in test mode. Go back to regular view, walk in/out drive in/out same way ... nothing.

Did a repair installation, just in case something happened under the hood that shouldn't have. No dice. It'll show the motion detection film strip. Regular view no recordings, searching for recordings taps out ... spinning indefinitely. The motion detected bar does not populate.

Tried the manual recording to playback in VLC as you requested. Thing goes red. No recording made or found.

And no activity to the external hard drive either, as the little monitoring utility I added to check on the cpu and network load also is set to show hdd i/o.

So, I just gave up and hit uninstall ... I may revisit it on a Core i7.

Giving ivms another try.

Strange. Axxonnext was the easiest to get going. Ironically my biggest issue when testing it out was I had problems stopping it after motion triggers. It would just keep recording and recording after being triggered. Something's really wrong when you can't even manually record. You can try recording the the internal SSD/hd if you have the free space. Maybe it doesn't like the external.

Another thing to check when trying out different VMS packages is that camera settings can get screwed up. Each time you are in effect giving the software "root" for each camera. I had a few times where 2-3 cameras would have crazy settings messing things up.

 

[dr.]

iVMS 4200
6x Hikv 3MP set to 2MP
Live View on with 6 on screen (1 lg, 5 sm)

~30% CPU (26-34), CPU 62C, 3MB/s data I/O

Still have to learn the details with iVMS (getting some errors when I try to apply some settings) and figure out some of the motion details to make it do its thing; but it certainly looks more promising and safer for the little guy, cpu wise.

Remember that ivms-4200 may have a bug on the live view. Being "small" doesn't mean it auto switched back to the sub-stream.

What about the VLC playback test? Record a few minutes of 2MP video. Record a few minutes of sub-stream that you have set. Use VLC to emulate the live view i.e. one vlc window of 2MP mp4 file + 5 vlc windows playing sub-stream video files. You should get the exact same CPU usage reported. I'm interested in the 6 vlc windows playing just the substream. 30% still seems really high.

 

[icerabbit]

Yes, I remember. Seems that with this install it was stepping down properly, as the CPU didn't really drop when I switched from auto to manual, and forced it into 6x sub stream.

Good news is that with fair weather and less wind (thus less motion and recording, observing an outdoor environment), a couple motion sensitivity tweaks (from 4 down to 3); with live view on for 6x substreams (set to like 1/4 HD, the step above 640x480) the CPU gauge has been dropping below 20, maybe 18% avg (considering 15-24); even with a firefox window open with a couple tabs and the small monitoring utility.

I'm pretty pleased with that.

I still have to look into the playback recording of files and 6x VLC windows. Hopefully tonight

 

[dr.]

Alright. That looks much, much better.

Recording should have zero affect on CPU usage. Using camera hardware motion controls should have zero affect on CPU usage. The live view client is basically just playing back video streams and reporting triggers/alarms.

That VLC testing should reveal if lower CPU usage is possible.

 

[icerabbit]

Dr. you'll be pleased to hear that with 6 sub streams (@ 704x480) in 9up view (so not 1 main @ 1080 viewing larger and 5 smaller subs around it) that the cpu usage has dropped to as low as 7% last night. Currently avg 12% in daytime.

The only reason I mentioned recording earlier, was that, when motion recording kicks in, I can see the recorder and streaming server process become the top loading processes briefly. Then the top process is ivms again.

Will get to the VLC thing. Only so many hours in a day I'm still running cable and installing cams and other hardware. It started easy with short easy runs, and is getting progressively harder. Finally got #5 up yesterday. Anyway. I am pleased with ivms from a cpu standpoint. Certainly solves the CPU headache I was getting from Axxon Next and video motion detection.

Still need to tweak some things with iVMS but it is getting there.

Oh, one thing I am not seeing is the vcr like playback controls in review mode. I have to right click with a popup menu? According to a tutorial there should be controls in the bottom bar? Any insights? Or preference I missed? Thanks.

PS: Forgot that the big core i7 is still looking at the cams too. 4% CPU avg (runs 2-6%) for 6 main streams in 6-up, 1lg and 5 small on screen, ivms auto select on. CPU power!