Hi all

I've been running BI fairly successfully at work for some years now, on a small scale.

I've recently been asked to dramatically increase the scale of the system - potentially up to 30 cameras, and the hardware I'm currently running on is most definitely not going to cut it.

I can throw a bit of money at this, but not a crazy amount. I was wondering if a second hand server a generation or two old would be a good starting point? It looks like any i7 (or Xeon?) from Kaby Lake onward supports high enough QuickSync?

IIRC, Blue Iris can't make use of GPU, so it's solely down to the CPU (and storage speed)?

Any advice very much appreciated!

Not sure what you heard and from where but BI CAN take advantage of both 1) Intel emedded GPU and 2) NVidia external GPU for hardware acceleration. With Intel you can't use H265 encoding but if you can live with H264 the Intel solution is more efficient.

For your system load, you'll need to calculate the pixels/sec. A 2mp camera running at 30 FPS (not recommended) would be obviously much more demanding than a 2mp camera running at 15 fps.

There is an online database with user submitted data https://biupdatehelper.hopto.org/default.html#stats

HeneryH wrote: ↑Thu Aug 29, 2019 2:35 pm Not sure what you heard and from where but BI CAN take advantage of both 1) Intel emedded GPU and 2) NVidia external GPU for hardware acceleration. With Intel you can't use H265 encoding but if you can live with H264 the Intel solution is more efficient.

For your system load, you'll need to calculate the pixels/sec. A 2mp camera running at 30 FPS (not recommended) would be obviously much more demanding than a 2mp camera running at 15 fps.

There is an online database with user submitted data https://biupdatehelper.hopto.org/default.html#stats

NVIDIA H/W Acceleration is very inefficient, sure if for extreme cases, but most people should not use this.
I also argue the embedded GPU, while it reduces CPU usage, power and heat consumption remain high and in most cases the same. Again has it's use cases, but don't dismiss raw cpu power.

Thanks for the responses folks (and sorry for slow reply, life is very very busy).

The majority of the cameras will be 2mp, and 15FPS should be fine.

However, most/all of them need to be constantly on display, not just recorded - as I understand it, that forces the machine running BI to constantly decode them all, and I would guess this would be the main processing cost. And given that, I would assume I would be much better sticking with h264 than x265 and living with the greater storage requirement.

I can buy a machine specifically for the purpose, up to maybe £1000 if I need to spend that much (less would be better, of course!) - but it needs to do the job, properly.

So, advice on where to allocate the money would be helpful. Obviously raw CPU is important, but should I simply prioritise core count/multicore performance? (While understanding that QuickSync means I should go Intel). Does Hyperthreading/SMT help? Is it worth getting the latest and greatest, or will a high end i5/i7 from gen8/9 be better?

How much does BI care about RAM?

Is it worth getting a discrete graphics card, if it can indeed be used now? (Or I have several nVidia cards from a few generations back sitting idle if they would help, e.g. GTX 650 and GTX 660 Ti).

Basically, please help me understand what to prioritise to build the best machine specifically to handle 30+ cameras in BlueIris

A very rough calc gives me 900MP/s, which the chart linked by HeneryH suggests could comfortably be handled by an i7-8700K with 32GB - does that sound about right?

Hi

I'm also very interested to hear about this, I'm considering creating a new surveillance storage/monitoring system from scratch using unraid, and could do with some hardware specifications. All cameras are hikvision and we have about 50 cameras.

They want all recording 24x7

what's needed?

Here is what I just found out.
Systems:
AMD RYZEN 2700X 16 core
NVIDIA 980 ti 8gig
32 gig DDR4 2400
1 Tb nvme m.2
ASROCK motherboard

INTEL i9 9900K 8 core 16 thread Quick Sync
32 gig DDR4 2400
1 Tb nvme m.2
ASROCK motherboard

(16) LOREX 4k 15 f/s low light cameras

Both systems perform exactly the same. Both can handle up to 12 4k cameras with direct to drive recording and motion detect. (Given worst case all 12 cameras are recording.)
After 12 cameras the frame rate drops to 3 to 5 frames per second and recording is sporadic as in stop motion and or missed events.
At 14 cameras both systems lock up. Even the mouse becomes sporadic.

One would think that installing the NVIDA card on the INTEL board would help. It does not.
I distributed 6 cameras on the INTEL and 6 on the NVIDIA. 12 cameras were still all the system would reliably handle before stuttering.

Hope this helps.
Charlie

That... seems really odd to me. Everything works fine with 12, then suddenly goes bad at 13/14?

What's CPU usage like? Are you viewing the streams/decoding?

I feel like you must be hitting a different bottleneck - maybe bandwidth on the network adapter? (Or switch, or an uplink, etc etc)

12 cameras at 4K, 15FPS would be pretty much bang on 100Mb/s

Shrike0064, your system seems to be maxing out sooner than expected. Something seems wrong. Can you grab the camera status from the status tab on the console and see what your bit rates are?

Secondly, I was under the impression that whether cameras are recording or not wouldn't make a difference. Can you try a quick experiment to change the cameras to record always rather than record on motion and see if there is a change for the worse?