Server Hardware Specs

Purpose

A dedicated server on-site is required to run the Nirovision software. The job of the server is to:

• Receive and store video streams.
• Perform AI inferencing (detect people, vehicles, faces, licence plates, recognise people and vehicles).
• Send the results to the Nirovision cloud.

Key components

💡 In general terms, the server specs reflect that of a low, mid or high-end gaming PC.

GPU

Most of the AI workload is performed by the GPU.

Currently, we only support NVIDIA GPUs at this stage.

• While there are many product lines in the NVIDIA product range, GeForce RTX provides the best price-to-performance ratio.
  • The most important spec to look for is the number of CUDA cores, so you want to optimise against price vs. Cuda cores.
  • The second most important spec would be GPU memory. GPU memory is used to store the Deep Neural Nets, amongst other things.
  • Geforce GTX cards have the best price/CUDA cores ratio, making them the most cost-effective GPU for AI inference applications (as opposed to AI training).

CPU

The CPU is still an important component as it performs all of the general-purpose computation (running the operating system, performing IF/ELSE logic, etc.).

We recommend Intel Core i5 or AMD Ryzen 5 CPUs as a minimum.

For larger systems, i7 or Ryzen 7 would be more suitable. Server-grade CPUs like Intel Xeon won't hurt, but might be overkill depending on the situation.

Primary Storage Drive

A small 120GB NVMe SSD is required for the Nirovision OS. 120 GB or more.

Secondary Storage Drive(s)

Nirovision records footage continuously with a rolling delete. Surveillance-grade HDDs are required for this. Use a storage calculator to determine how much is needed based on the video quality you want to save and how long you need to retain it.

If you need to store video footage for long periods, a separate VMS or NVR is recommended.

Network Interfaces

We recommend two network interfaces: one for Internet connectivity and one for the security camera LAN. (It is common practice to separate the IP camera LAN from the admin LAN).

Memory

16GB for 8 cameras

32GB for 24 cameras

64GB for 48 cameras

Sample Hardware Configs

Here are some typical deployment options.

⚠️ When CPUs are mentioned, assume current generation models. An Intel Core i7 12700k (12th Gen 2022) is very different in capability compared to an Intel Core i7 4770K (4th Gen 2013).

CPU: Intel Core i5 Processor or AMD Ryzen 5

GPU: NVIDIA GeForce RTX 3050 8GB RAM

Memory: 16GB DDR4 RAM

Primary Drive: 120GB NVMe SSD

Storage Drive: 4TB Surveillance Grade HDD

Networks Cards: 2 x NICS

16-camera server

CPU: Intel Core i7 Processor or AMD Ryzen 7

GPU: NVIDIA GeForce RTX 3070 8GB RAM

Memory: 16GB DDR4 RAM

Primary Drive: 120GB NVMe SSD

Storage Drive: 8TB Surveillance Grade HDD

Networks Cards: 2 NICS

32-camera server

CPU: Intel Core i9 Processor or AMD Ryzen 9

GPU: NVIDIA GeForce RTX 3090 24GB RAM

Memory: 32GB DDR4 RAM

Primary Drive: 120GB NVMe SSD

Storage Drive: 16TB Surveillance Grade HDD

Networks Cards: 2 NICS

Budget Options for Testing Only

• CPU-only, single camera

  CPU: Intel Core i5, AMD Ryzen 5 or better.

  GPU: None/built-in

  Memory: 16GB DDR4 RAM

  Primary Drive: 120GB Solid State Drive

  Storage Drive: 1TB Surveillance Grade HDD

• GPU, 1-2 cameras

  CPU: Intel Core i3 Processor or better

  GPU: Single NVIDIA GeForce 1650 4GB RAM

  Memory: 16GB DDR4 RAM

  Primary Drive: 120GB Solid State Drive

  Storage Drive: 2TB Surveillance Grade HDD