NCNN

NCNN is a high performance neural network inference framework optimized for mobile and other platforms developed by Tencent.

To run this test with the Phoronix Test Suite, the basic command is: phoronix-test-suite benchmark ncnn.

Project Site

github.com

Test Created

18 September 2020

Last Updated

27 July 2021

Test Maintainer

Michael Larabel 

Test Type

System

Average Install Time

1 Minute

Average Run Time

20 Minutes, 48 Seconds

Test Dependencies

CMake + C/C++ Compiler Toolchain + Vulkan

Accolades

10k+ Downloads

Supported Platforms


Public Result Uploads *Reported Installs **Reported Test Completions **Test Profile Page Views ***OpenBenchmarking.orgEventsNCNN Popularity Statisticspts/ncnn2020.092020.102020.112020.122021.012021.022021.032021.042021.052021.062021.072021.082021.092021.102021.112021.122022.012022.022022.032022.042022.054K8K12K16K20K
* Uploading of benchmark result data to OpenBenchmarking.org is always optional (opt-in) via the Phoronix Test Suite for users wishing to share their results publicly.
** Data based on those opting to upload their test results to OpenBenchmarking.org and users enabling the opt-in anonymous statistics reporting while running benchmarks from an Internet-connected platform.
*** Test profile page view reporting began March 2021.
Data current as of 24 May 2022.
CPU66.5%Vulkan GPU33.5%Target Option PopularityOpenBenchmarking.org
efficientnet-b06.6%regnety_400m6.7%mnasnet6.6%resnet186.6%mobilenet-v36.6%mobilenet6.6%vgg166.8%blazeface6.4%mobilenet-v26.6%resnet506.9%squeezenet_ssd6.8%yolov4-tiny6.7%alexnet6.8%googlenet6.6%shufflenet-v26.6%Model Option PopularityOpenBenchmarking.org

Revision History

pts/ncnn-1.3.0   [View Source]   Tue, 27 Jul 2021 16:34:42 GMT
Update against NCNN 2021-07-20 upstream, fix possible Vulkan build issue by including glslang source tree.

pts/ncnn-1.2.0   [View Source]   Fri, 18 Jun 2021 08:11:33 GMT
Update against NCNN 20210525 release.

pts/ncnn-1.1.0   [View Source]   Fri, 18 Dec 2020 08:06:41 GMT
Update against new upstream NCNN 20201218.

pts/ncnn-1.0.3   [View Source]   Fri, 25 Sep 2020 06:36:39 GMT
Drop int8 tests per https://github.com/phoronix-test-suite/test-profiles/pull/167

pts/ncnn-1.0.2   [View Source]   Thu, 24 Sep 2020 12:52:47 GMT
Expose Vulkan GPU support.

pts/ncnn-1.0.1   [View Source]   Fri, 18 Sep 2020 12:28:10 GMT
Increase the run count.

pts/ncnn-1.0.0   [View Source]   Fri, 18 Sep 2020 11:58:15 GMT
Initial commit of Tencent NCNN.

Suites Using This Test

Machine Learning

HPC - High Performance Computing

NVIDIA GPU Compute

Vulkan Compute


Performance Metrics

Analyze Test Configuration:

NCNN 20210720

Target: CPU - Model: resnet50

OpenBenchmarking.org metrics for this test profile configuration based on 458 public results since 27 July 2021 with the latest data as of 9 May 2022.

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
Percentile Rank
# Compatible Public Results
ms (Average)
Mid-Tier
75th
> 24
59th
8
29 +/- 2
Median
50th
32
48th
13
34 +/- 3
47th
4
34 +/- 1
39th
5
40 +/- 3
Low-Tier
25th
> 46
25th
5
46 +/- 5
11th
8
86 +/- 7
4th
4
138 +/- 1
OpenBenchmarking.orgDistribution Of Public Results - Target: CPU - Model: resnet50458 Results Range From 12 To 2442 ms126111015920825730635540445350255160064969874779684589494399210411090113911881237128613351384143314821531158016291678172717761825187419231972202120702119216822172266231523642413246290180270360450

Based on OpenBenchmarking.org data, the selected test / test configuration (NCNN 20210720 - Target: CPU - Model: resnet50) has an average run-time of 10 minutes. By default this test profile is set to run at least 3 times but may increase if the standard deviation exceeds pre-defined defaults or other calculations deem additional runs necessary for greater statistical accuracy of the result.

OpenBenchmarking.orgMinutesTime Required To Complete BenchmarkTarget: CPU - Model: resnet50Run-Time1326395265Min: 2 / Avg: 9.77 / Max: 69

Based on public OpenBenchmarking.org results, the selected test / test configuration has an average standard deviation of 0.6%.

OpenBenchmarking.orgPercent, Fewer Is BetterAverage Deviation Between RunsTarget: CPU - Model: resnet50Deviation3691215Min: 0 / Avg: 0.61 / Max: 11

Does It Scale Well With Increasing Cores?

No, based on the automated analysis of the collected public benchmark data, this test / test settings does not generally scale well with increasing CPU core counts. Data based on publicly available results for this test / test settings, separated by vendor, result divided by the reference CPU clock speed, grouped by matching physical CPU core count, and normalized against the smallest core count tested from each vendor for each CPU having a sufficient number of test samples and statistically significant data.

IntelAMDOpenBenchmarking.orgRelative Core Scaling To BaseNCNN CPU Core ScalingTarget: CPU - Model: resnet502468162432641.00212.00423.00634.00845.0105

Notable Instruction Set Usage

Notable instruction set extensions supported by this test, based on an automatic analysis by the Phoronix Test Suite / OpenBenchmarking.org analytics engine.

Instruction Set
Support
Instructions Detected
SSE2 (SSE2)
Used by default on supported hardware.
 
PUNPCKLQDQ MOVD PSHUFD PSHUFLW MOVDQU MOVDQA SUBSD MOVAPD MINSD MAXSD ADDSD CVTSI2SD DIVSD CVTSS2SD PADDQ PSUBQ SHUFPD MULSD CVTTSD2SI MULPD PSRLDQ CVTSD2SS SQRTSD PUNPCKHQDQ CVTDQ2PS CVTTPS2DQ COMISD PMULUDQ ANDPD UCOMISD MOVUPD MOVHPD UNPCKHPD ANDNPD CMPNLESD ORPD XORPD SUBPD
Used by default on supported hardware.
Found on Intel processors since Sandy Bridge (2011).
Found on AMD processors since Bulldozer (2011).

 
VZEROUPPER VBROADCASTSS VINSERTF128 VPERM2F128 VEXTRACTF128 VMASKMOVPS VBROADCASTSD VPERMILPS
Used by default on supported hardware.
Found on Intel processors since Haswell (2013).
Found on AMD processors since Excavator (2016).

 
VPERM2I128 VEXTRACTI128 VPERMD VPERMQ VPBROADCASTW VINSERTI128 VPBROADCASTD VPBROADCASTB VPBROADCASTQ VGATHERDPS VPGATHERQQ VPSRLVQ VPBLENDD VPGATHERQD VPERMPD
FMA (FMA)
Used by default on supported hardware.
Found on Intel processors since Haswell (2013).
Found on AMD processors since Bulldozer (2011).

 
VFMADD132PS VFMADD132SS VFMADD231SS VFMADD231PS VFNMADD231PS VFNMADD132PS VFMSUB231PS VFMSUB132PS VFMADD213PS VFMADD213SS VFMSUB213PS VFMSUB132SS VFMADD132SD VFNMADD213SS VFNMADD132SS VFNMADD231SS VFMADD132PD VFMADD231SD VFNMADD132SD VFMSUB231SD VFNMADD213SD VFNMADD213PS
Advanced Vector Extensions 512 (AVX512)
Requires passing a supported compiler/build flag (verified with targets: cascadelake, sapphirerapids).
 
(ZMM REGISTER USE)
The test / benchmark does honor compiler flag changes.
Last automated analysis: 18 January 2022

This test profile binary relies on the shared libraries libgomp.so.1, libvulkan.so.1, libm.so.6, libmvec.so.1, libc.so.6.

Tested CPU Architectures

This benchmark has been successfully tested on the below mentioned architectures. The CPU architectures listed is where successful OpenBenchmarking.org result uploads occurred, namely for helping to determine if a given test is compatible with various alternative CPU architectures.

CPU Architecture
Kernel Identifier
Verified On
Intel / AMD x86 64-bit
x86_64
(Many Processors)
Loongson LoongArch 64-bit
loongarch64
Loongson-3A5000LL
ARMv7 32-bit
armv7l
ARMv7 Cortex-A72 4-Core
ARMv8 64-bit
arm64
Apple M1 Max, Apple M1 Pro
ARMv8 64-bit
aarch64
AArch64 rev 12, ARMv8 Cortex-A72 16-Core, ARMv8 Cortex-A72 4-Core, Ampere eMAG ARMv8 32-Core, Apple M1

Recent Test Results

OpenBenchmarking.org Results Compare

1 System - 31 Benchmark Results

AMD Ryzen 7 3800X 8-Core - MSI MEG X570 ACE - AMD Starship

Garuda Soaring - 5.17.9-zen1-1-zen - KDE Plasma 5.24.5

4 Systems - 541 Benchmark Results

AMD Ryzen 7 5800X3D 8-Core - ASRock X570 Pro4 - AMD Starship

Ubuntu 22.04 - 5.17.4-051704-generic - GNOME Shell 42.0

3 Systems - 541 Benchmark Results

AMD Ryzen 7 5800X 8-Core - ASRock X570 Pro4 - AMD Starship

Ubuntu 22.04 - 5.17.4-051704-generic - GNOME Shell 42.0

3 Systems - 541 Benchmark Results

AMD Ryzen 7 5800X3D 8-Core - ASRock X570 Pro4 - AMD Starship

Ubuntu 22.04 - 5.17.4-051704-generic - GNOME Shell 42.0

3 Systems - 531 Benchmark Results

AMD Ryzen 9 5950X 16-Core - ASUS ROG CROSSHAIR VIII HERO - AMD Starship

Ubuntu 22.04 - 5.17.4-051704-generic - GNOME Shell 42.0

1 System - 84 Benchmark Results

AMD Ryzen 7 5800X3D 8-Core - ASRock X570 Pro4 - AMD Starship

Ubuntu 22.04 - 5.17.4-051704-generic - GNOME Shell 42.0

1 System - 16 Benchmark Results

Intel Core i9-10900F - Alienware 0N43JM - Intel Comet Lake PCH

Fedora 34 - 5.13.19-200.fc34.x86_64 - GNOME Shell 40.9

4 Systems - 116 Benchmark Results

ARMv8 Cortex-A72 - BCM2835 Raspberry Pi 400 Rev 1.0 - Broadcom BCM2711

Debian 11 - 5.10.92-v8+ - LXDE

1 System - 35 Benchmark Results

AMD Ryzen 5 3600 6-Core - ASUS ROG STRIX B550-E GAMING - AMD Starship

Ubuntu 21.10 - 5.13.0-39-generic - GNOME Shell 40.5

1 System - 35 Benchmark Results

2 x Intel Xeon E5-2673 v3 - Dell 0NK5PH - Intel Xeon E7 v3

Ubuntu 20.04 - 5.13.0-35-generic - GNOME Shell 3.36.7

5 Systems - 116 Benchmark Results

ARMv8 Cortex-A72 - SolidRun CEX7 - 32GB

Ubuntu 21.10 - 5.17.0-051700-generic - KDE Plasma 5.24.3

5 Systems - 116 Benchmark Results

ARMv7 Cortex-A72 - BCM2711 Raspberry Pi 400 Rev 1.0 - Broadcom BCM2711

Raspbian 11 - 5.10.92-v7l+ - LXDE

3 Systems - 268 Benchmark Results

Apple M1 - Apple Mac mini - 8GB

Arch Linux ARM - 5.17.0-rc7-asahi-next-20220310-5-1-ARCH - KDE Plasma 5.24.3

1 System - 120 Benchmark Results

AMD Ryzen 7 3700X 8-Core - ASUS ROG STRIX X570-E GAMING - AMD Starship

Ubuntu 20.04 - 5.4.0-104-generic - GNOME Shell 3.36.9

1 System - 192 Benchmark Results

Apple M1 - Apple Mac mini - 8GB

macOS 12.3 - 21.4.0 - OpenCL 1.2

Most Popular Test Results

OpenBenchmarking.org Results Compare

2 Systems - 115 Benchmark Results

AMD Ryzen 9 5950X 16-Core - ASUS ROG CROSSHAIR VIII HERO - AMD Starship

Ubuntu 21.10 - 5.13.0-14-generic - GNOME Shell 40.2

4 Systems - 196 Benchmark Results

Intel Core i7-8700K - ASUS TUF Z370-PLUS GAMING - Intel 8th Gen Core

Ubuntu 20.04 - 5.9.0-050900rc6daily20200923-generic - GNOME Shell 3.36.4

3 Systems - 141 Benchmark Results

2 x AMD EPYC 75F3 32-Core - ASRockRack ROME2D16-2T - AMD Starship

Ubuntu 21.10 - 5.13.0-20-generic - GNOME Shell 40.5

4 Systems - 75 Benchmark Results

AMD FX-8150 Eight-Core - ASUS M5A78L-M LX PLUS - AMD RS780 + SB7x0

Ubuntu 20.10 - 5.8.0-53-generic - GNOME Shell 3.38.1

2 Systems - 82 Benchmark Results

Intel Core i9-11900K - ASUS ROG MAXIMUS XIII HERO - Intel Device 43ef

Ubuntu 21.04 - 5.11.0-25-generic - GNOME Shell 3.38.4

4 Systems - 146 Benchmark Results

Intel Core i5-7600K - Gigabyte Z270M-D3H-CF - Intel Xeon E3-1200 v6

Ubuntu 20.04 - 5.9.0-050900rc7daily20200928-generic - GNOME Shell 3.36.4

2 Systems - 140 Benchmark Results

AMD Ryzen 9 5900X 12-Core - ASUS ROG CROSSHAIR VIII HERO - AMD Starship

Ubuntu 21.10 - 5.13.0-16-generic - GNOME Shell 40.2

3 Systems - 89 Benchmark Results

AMD Ryzen 9 3950X 16-Core - ASUS ROG CROSSHAIR VII HERO - AMD Starship

Ubuntu 20.04 - 5.9.0-050900rc6daily20200922-generic - GNOME Shell 3.36.4

2 Systems - 140 Benchmark Results

2 x AMD EPYC 75F3 32-Core - ASRockRack ROME2D16-2T - AMD Starship

Clear Linux OS 35150 - 5.14.13-1083.native - GNOME Shell 41.0

4 Systems - 63 Benchmark Results

AMD Ryzen 9 3900XT 12-Core - MSI MEG X570 GODLIKE - AMD Starship

Ubuntu 20.10 - 5.11.0-rc1-phx - GNOME Shell 3.38.1

4 Systems - 93 Benchmark Results

AMD Ryzen Threadripper 3960X 24-Core - MSI Creator TRX40 - AMD Starship

Ubuntu 20.04 - 5.12.0-051200rc2daily20210307-generic - GNOME Shell 3.36.4

4 Systems - 141 Benchmark Results

2 x AMD EPYC 75F3 32-Core - ASRockRack ROME2D16-2T - AMD Starship

Clear Linux OS 35150 - 5.14.13-1083.native - GNOME Shell 41.0

Find More Test Results