RealSR-NCNN

RealSR-NCNN is an NCNN neural network implementation of the RealSR project and accelerated using the Vulkan API. RealSR is the Real-World Super Resolution via Kernel Estimation and Noise Injection. NCNN is a high performance neural network inference framework optimized for mobile and other platforms developed by Tencent. This test profile times how long it takes to increase the resolution of a sample image by a scale of 4x with Vulkan.


RealSR-NCNN 20200818

Scale: 4x - TAA: Yes

OpenBenchmarking.org metrics for this test profile configuration based on 961 public results since 24 September 2020 with the latest data as of 9 December 2024.

Below is an overview of the generalized performance for components where there is sufficient statistically significant data based upon user-uploaded results. It is important to keep in mind particularly in the Linux/open-source space there can be vastly different OS configurations, with this overview intended to offer just general guidance as to the performance expectations.

Component
Details
Percentile Rank
# Compatible Public Results
Seconds (Average)
89th
10
31 +/- 1
80th
6
40 +/- 1
Mid-Tier
75th
> 42
Median
50th
63
47th
12
65 +/- 2
28th
5
115 +/- 5
Low-Tier
25th
> 138
24th
4
149 +/- 1
22nd
5
210 +/- 4
22nd
4
223 +/- 1
20th
8
278 +/- 4
17th
4
456 +/- 24
15th
3
496 +/- 6
12th
10
523 +/- 23
7th
4
819 +/- 4
7th
4
831 +/- 46
5th
3
997 +/- 65
5th
7
1117 +/- 18
3rd
3
1503 +/- 9
2nd
6
1683 +/- 2
1st
3
1764 +/- 2
1st
3
2530 +/- 6