I am upgrading my surveillance system from a 2 cam to a 6 cam system (maybe 8 in a year) and upgrading from an Arm based QNAP which was sufficient but heavily loaded to a new system most probably running Netcam Studio with the following specs:
i5 6600
8GB DDR4
2x 4TB WD Purple Raid 1 (Windows RAID)
120GB Entry level SSD (For OS and applications)
Windows 10 x64 Pro
The cameras will all be Brickcoms recording at 2MP at 15FPS continuously. Motion detection isn’t going to be a thing as there are constant movement in the scenes and the areas are semi public, though I may experiment with motion detection if I can figure out how to make it work for the scenes.
What I would like to know, is how do these specs (particularly CPU) stand up to this and if switching to a i7 6700 instead would be worth it?
I already purchased everything but I’m returning the motherboard due to some error on my part and while I’m at it, I noticed that the 6700 has the same TDP of 65W as the 6600, but has Hyperthreading and an 8MB cache vs the 6MB cache, so I was wondering if I made a mistake with the 6600.
Hi Adam!
8 cams - 2MP - 15 FPS - motion detection is demanding. Both processors are quad core so that is good. I don´t think the other differences are crucial. RAM is good. RAID or not is more of security. 4TB depends on how long you want to store the videos.
-Henrik
500GB per camera suits me fine based on the current 2 cameras that I’ve been running for a year now.
The reason for my concern is that video transcoding is a highly threaded application, if I’m not mistaken, so I may be able to get something like a 30% performance improvement thanks to hyperthreading, if netcam studio is in line with synthetic benchmarks.
I did some testing and it seems that Hyper Threading has a huge benefit. I tested this on my workstation, which has an i7 920 @ 2.6Ghz and with a demanding stream of 30FPS on max quality and 2MP from 1 camera. With Hyper Threading I got 30% utilization when recording but without Hyper Threading I got 55% utilization.
Excellent! Yes, I was getting a little suspicious about that. Lately we have been testing different scenarios to reduce CPU load, but it seems that NCS as it is is very good at it´s job. In version 1.3.7 it is possible to use hardware decoding with DXVA2. If you are interested to test you can download the unofficial version here https://www.dropbox.com/s/x2imt9ghqxwnwak/ncs-inst-x64.exe?dl=0
Thanks!
I installed it and tried it out with Hyperthreading, but didn’t notice a difference with 1.3.7, though I’m not sure if I configured it correctly.
Settings - Decoding/Encoding - Decoding Settings - Hardware_DXVA2. As you can see it´s also an option for QuickSync under development. I tested also DXVA2, but unfortunately with no wow-effect. The future will tell if there will be any more wow-effect ;).
Thanks. I did that and it only reduced my load by 2%, though My workstation has no integrated GPU and has a GTX 960, so maybe the hardware isn’t being used. Maybe a modern IGPU like on a Skylake may come into effect then.
When the integration with QuickSync is ready we will do a more systematic test of all the possibilities and what is needed when it comes to hardware. I hope that it will be within a month or two.
Thanks for your help and support!
-Henrik @Steve