Blogs

Latency Comparison Test of Software and Hardware Encoding & Decoding

2020.03.09
Video Transmission
Latency Reduction

Background

When users choose to transmit video over the network, the transmission latency is often what they care about most. During the entire transmission process, the selection of software and hardware encoders & decoders is very important, which can directly affect the latency of network transmission. This blog focuses the comparison of video transmission latency by using different software and hardware encoding & decoding devices in the same scenario.

1.Testing Equipment

Hardware Equipment:

  1. Video source: Skyworth Video Player (1080P60)
  2. Video capture device: Pro Capture HDMI 4K Plus   
  3. Hardware encoder: Pro Convert HDMI 4K Plus(Firmware V1.0.567), a similar encoder of another brand
  4. Hardware decoder: Pro Convert for NDI® to HDMI(Firmware V1.0.571), a similar decoder of another brand
  5. PC parameters: AMD Ryzen 5 1400G Quad-Core Processor  3.20GHz
  6. Monitor: three DELL E2214Hv (hereinafter referred to as Monitor 1,2,3)
  7. Switch: H3C Aoiynk S1208D

Software:

  1. Encoding software: NDI Connect Pro
  2. Decoding software: NDI Studio Monitor

2.Testing Method

2.1 Using software encoder to generate NDI stream  

Equipment Set Up:

Diagram 1: The same encoding software is used to generate NDI stream; two hardware decoders and one decoding software are used for decoding

We use Magewell capture card and NDI Connect Pro to encode the video source and transmit it to the LAN through NDI® protocol. At the same time, the loop-through interface of the capture card is connected to Monitor 1. Since the loop through isn't processed by any system, and the monitor used in this test has no operating system, the screen it displays can be regarded as the video source itself which can be used for comparison with the displayed screen by decoding .

For decoding, Magewell Pro Convert for NDI® to HDMI (hereinafter referred to as "Magewell NDI® decoder"), the decoder of another brand and NDI Studio Monitor are connected to the same LAN. And the decoded video by hardware decoders are output to Monitor 2 and Monitor 3 for display through the HDMI interface, while the decoded video by NDI Studio Monitor is displayed on PC directly.

For the testing purpose, we have specially made a video that can be used for stopwatch timing. By recording and comparing the stopwatch time of the video source(Monitor 1) and that of the decoder(Monitor 2, Monitor 3 and NDI Studio Monitor), we can obtain the time required from video encoding to the completion of decoding, which is the video transmission latency.

2.2 Using standalone hardware encoder to generate NDI stream

After setting up the above testing environment, we try to switch the encoding software to Magewell Pro Convert HDMI 4K Plus encoder(hereinafter referred to as "Magewell NDI® encoder") and the encoder of another brand for hardware encoding test. Other devices are not changed. The equipment is set up as follows:

Diagram 2: Magewell NDI encoder is used to generate NDI stream; two hardware decoders and one decoding software are used for decoding
Diagram 3: The encoder of another brand is used to generate NDI stream; two hardware decoders and one decoding software are used for decoding

3.Testing Results

3.1 Latency Comparison Test Data of Software Encoding and Software & Hardware Decoding

NO. Latency of decoder of another brand (ms)
(Monitor 1-Monitor 2)		 Latency of Magewell decoder (ms)
(Monitor 1-Monitor 3)		 Latency of decoding software (ms)
(Monitor 1-NDI Studio Monitor)
1 66 50 116
2 66 66 116
3 66 50 116
4 67 50 117
5 66 66 116
6 66 66 116
7 67 67 117
8 67 67 117
9 66 66 133
10 66 66 133
Average 66.3 61.4 119.7

3.2 Latency Comparison Test Data of Magewell Hardware Encoding and Software & Hardware Decoding

NO. Latency of decoder of another brand (ms)
(Monitor 1-Monitor 2)		 Latency of Magewell decoder (ms)
(Monitor 1-Monitor 3)		 Latency of decoding software (ms)
(Monitor 1-NDI Studio Monitor)
1 67 50 100
2 66 50 96
3 60 50 100
4 50 50 100
5 66 50 100
6 50 50 100
7 50 33 100
8 50 50 100
9 67 50 100
10 50 33 100
Average 57.6 46.6 99.6

3.3 Latency Comparison Test Data of Another Brand Hardware Encoding and Software & Hardware Decoding

NO. Latency of decoder of another brand (ms)
(Monitor 1-Monitor 2)		 Latency of Magewell decoder (ms)
(Monitor 1-Monitor 3)		 Latency of decoding software (ms)
(Monitor 1-NDI Studio Monitor)
1 50 50 116
2 50 34 100
3 84 67 134
4 83 66 133
5 66 66 133
6 50 50 117
7 67 67 133
8 50 50 117
9 50 50 117
10 50 50 117
Average 60 55 121.7

After sorting we get the below table:

4.Testing Results

With the same network environment, the network transmission latency is closely related to the selection of software and hardware encoders & decoders. Compared with software encoders and decoders, standalone hardware encoders and decoders have sufficient dedicated resources to complete the task of encoding and decoding, and are not affected by system scheduling and CPU usage by other programs running in the computer. For users with extremely high video latency requirements, it's recommended to pair a Magewell encoder with a Magewell decoder, which can greatly reduce the overall transmission latency. It should also be noted that setting the buffer duration value of Magewell NDI decoder reasonably will also optimize the latency. Users can set the buffer duration value in the Web GUI of Magewell NDI decoders. For details, it is recommended to read "Test of Buffer Duration Value’s Effect on Pro Convert Latency".