OpenVINO

This is a test of the Intel OpenVINO, a toolkit around neural networks, using its built-in benchmarking support and analyzing the throughput and latency for various models.

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

Project Site

openvino.ai

Source Repository

github.com

Test Created

7 October 2020

Last Updated

6 March 2024

Test Maintainer

Michael Larabel 

Test Type

System

Average Install Time

21 Minutes, 26 Seconds

Average Run Time

4 Minutes, 26 Seconds

Test Dependencies

Python + C/C++ Compiler Toolchain + CMake

Accolades

300k+ Downloads

Supported Platforms


Public Result Uploads *Reported Installs **Reported Test Completions **Test Profile Page Views ***OpenBenchmarking.orgEventsOpenVINO Popularity Statisticspts/openvino2020.102020.112020.122021.012021.022021.032021.042021.052021.062021.072021.082021.092021.102021.112021.122022.012022.022022.032022.042022.052022.062022.072022.082022.092022.102022.112022.122023.012023.022023.032023.042023.052023.062023.072023.082023.092023.102023.112023.122024.012024.022024.032024.0411K22K33K44K55K
* 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 updated weekly as of 21 April 2024.
Face Detection FP16-INT89.2%Age Gender Recognition Retail 0013 FP16-INT88.9%Face Detection Retail FP16-INT89.4%Vehicle Detection FP16-INT88.9%Weld Porosity Detection FP16-INT88.6%Person Re-Identification Retail FP168.9%Handwritten English Recognition FP16-INT89.2%Machine Translation EN To DE FP169.4%Noise Suppression Poconet-Like FP169.4%Road Segmentation ADAS FP16-INT89.4%Person Detection FP168.9%Model Option PopularityOpenBenchmarking.org

Revision History

pts/openvino-1.5.0   [View Source]   Wed, 06 Mar 2024 14:47:11 GMT
Update against OpenVINO 2024.0 upstream.

pts/openvino-1.4.0   [View Source]   Sun, 29 Oct 2023 14:20:45 GMT
Update against OpenVINO 2023.2.dev to allow building on Ubuntu 23.10.

pts/openvino-1.3.0   [View Source]   Fri, 29 Sep 2023 10:07:22 GMT
Update against OpenVINO 2023.1 upstream.

pts/openvino-1.2.0   [View Source]   Wed, 21 Dec 2022 18:32:34 GMT
Update against OpenVINO 2022.3 upstream.

pts/openvino-1.1.0   [View Source]   Wed, 31 Aug 2022 15:43:22 GMT
Update against OpenVINO upstream to fix build on newer distros.

pts/openvino-1.0.4   [View Source]   Thu, 19 Nov 2020 13:48:10 GMT
Fix for running OpenVINO on AArch64.

pts/openvino-1.0.3   [View Source]   Thu, 08 Oct 2020 18:19:27 GMT
Expose Intel GPU option that works on Intel GPU setups having their GPU compute stack installed at OpenVINO build time.

pts/openvino-1.0.2   [View Source]   Wed, 07 Oct 2020 20:43:13 GMT
Add python2 as a system dependency - still needed for OpenVINO, yuck.

pts/openvino-1.0.1   [View Source]   Wed, 07 Oct 2020 18:36:20 GMT
Ensure OpenCV gets included as a system dependency.

pts/openvino-1.0.0   [View Source]   Wed, 07 Oct 2020 16:55:46 GMT
Initial commit of OpenVINO benchmark.

Suites Using This Test

Multi-Core

Machine Learning

HPC - High Performance Computing

Intel oneAPI

Creator Workloads


Performance Metrics

Analyze Test Configuration:

OpenVINO 2021.1

Model: Face Detection 0106 FP32 - Device: CPU

OpenBenchmarking.org metrics for this test profile configuration based on 504 public results since 7 October 2020 with the latest data as of 4 August 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
FPS (Average)
100th
4
18.44 +/- 0.03
99th
11
15.20 +/- 0.13
96th
8
9.03 +/- 0.04
95th
3
8.87 +/- 0.08
93rd
8
8.12 +/- 0.20
92nd
8
7.72 +/- 0.06
91st
5
7.67 +/- 0.39
89th
8
7.31 +/- 0.08
87th
7
6.98 +/- 0.01
86th
6
6.09 +/- 0.03
83rd
8
5.70 +/- 0.02
81st
9
5.64 +/- 0.01
79th
8
5.57 +/- 0.09
78th
3
5.40 +/- 0.02
77th
6
5.36 +/- 0.03
76th
6
5.24 +/- 0.05
Mid-Tier
75th
< 5.21
74th
7
5.03 +/- 0.01
73rd
8
4.63 +/- 0.08
69th
11
3.95 +/- 0.07
68th
3
3.87 +/- 0.01
66th
21
3.69 +/- 0.09
61st
7
3.31 +/- 0.02
60th
4
3.28 +/- 0.47
60th
10
3.27 +/- 0.05
57th
8
3.16 +/- 0.13
56th
13
3.13 +/- 0.10
56th
8
3.08 +/- 0.12
52nd
7
2.76 +/- 0.02
Median
50th
2.67
49th
6
2.52 +/- 0.01
48th
6
2.39 +/- 0.09
47th
6
2.28 +/- 0.01
43rd
6
2.10 +/- 0.01
42nd
12
2.07 +/- 0.07
41st
9
2.02 +/- 0.02
38th
3
1.94 +/- 0.02
37th
8
1.90 +/- 0.04
37th
8
1.90 +/- 0.04
34th
5
1.78 +/- 0.02
31st
6
1.70 +/- 0.01
29th
6
1.67 +/- 0.02
28th
8
1.61 +/- 0.11
26th
3
1.47 +/- 0.01
Low-Tier
25th
< 1.45
25th
6
1.34 +/- 0.02
25th
6
1.33 +/- 0.01
24th
3
1.32 +/- 0.01
22nd
3
1.27 +/- 0.01
22nd
6
1.26 +/- 0.01
20th
6
1.22 +/- 0.01
18th
3
1.16 +/- 0.05
15th
6
1.05 +/- 0.01
14th
3
1.01 +/- 0.01
10th
3
0.80 +/- 0.02
10th
3
0.69 +/- 0.01
9th
3
0.66 +/- 0.02
8th
7
0.65 +/- 0.01
OpenBenchmarking.orgDistribution Of Public Results - Model: Face Detection 0106 FP32 - Device: CPU504 Results Range From 0 To 19 FPS246810121416182022244080120160200

Based on OpenBenchmarking.org data, the selected test / test configuration (OpenVINO 2021.1 - Model: Face Detection 0106 FP32 - Device: CPU) has an average run-time of 4 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 BenchmarkModel: Face Detection 0106 FP32 - Device: CPURun-Time246810Min: 3 / Avg: 3.47 / Max: 5

Based on public OpenBenchmarking.org results, the selected test / test configuration has an average standard deviation of 0.1%.

OpenBenchmarking.orgPercent, Fewer Is BetterAverage Deviation Between RunsModel: Face Detection 0106 FP32 - Device: CPUDeviation246810Min: 0 / Avg: 0.14 / Max: 1

Does It Scale Well With Increasing Cores?

Yes, based on the automated analysis of the collected public benchmark data, this test / test settings does 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 BaseOpenVINO CPU Core ScalingModel: Face Detection 0106 FP32 - Device: CPU46812162432642.685.368.0410.7213.4

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.
 
MOVDQA PUNPCKLQDQ CVTSI2SD MULSD MOVDQU MOVAPD DIVSD MOVUPD ADDSD SUBSD COMISD MOVD ANDPD UCOMISD MOVMSKPD XORPD PADDQ PSRLDQ SHUFPD UNPCKHPD CVTTSD2SI PSHUFD PSHUFLW UNPCKLPD CVTDQ2PS CVTDQ2PD CVTSD2SS CVTPD2PS CVTTPD2DQ CMPLEPD SUBPD PSUBQ CVTSS2SD ANDNPD CMPNLESD ORPD CMPLTPD ADDPD CVTPS2PD CVTTPS2DQ MAXSD DIVPD MOVHPD PMULUDQ CMPNEQPD SQRTSD PUNPCKHQDQ MINSD
SSSE3 (SSSE3)
Used by default on supported hardware.
 
PSHUFB
Requires passing a supported compiler/build flag (verified with targets: sandybridge, skylake, tigerlake, cascadelake, sapphirerapids, alderlake, znver2, znver3).
Found on Intel processors since Sandy Bridge (2011).
Found on AMD processors since Bulldozer (2011).

 
VZEROUPPER VEXTRACTF128 VINSERTF128 VPERMILPD VBROADCASTSS VBROADCASTSD VPERM2F128 VPERMILPS
Requires passing a supported compiler/build flag (verified with targets: skylake, tigerlake, cascadelake, sapphirerapids, alderlake, znver2, znver3).
Found on Intel processors since Haswell (2013).
Found on AMD processors since Excavator (2016).

 
VINSERTI128 VPBROADCASTW VPBROADCASTD VPBROADCASTQ VPERMQ VEXTRACTI128 VPERM2I128 VPBROADCASTB VPGATHERQQ VPGATHERDQ VPMASKMOVD VPMASKMOVQ VPSRLVQ VPERMD
FMA (FMA)
Requires passing a supported compiler/build flag (verified with targets: skylake, tigerlake, cascadelake, sapphirerapids, alderlake, znver2, znver3).
Found on Intel processors since Haswell (2013).
Found on AMD processors since Bulldozer (2011).

 
VFMADD132SD VFMADD213SS VFMADD132SS VFMADD231SD VFMADD231SS VFNMADD132SD VFMADD213SD VFMSUB132SD VFMADD213PS VFMADD132PS VFNMADD132SS VFNMADD231SS VFMSUB231SS VFMSUB231SD VFMSUB132SS VFMADD132PD
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: 1 October 2023

This test profile binary relies on the shared libraries libOpenCL.so.1, libopenvino.so.2310, libm.so.6, libc.so.6, libz.so.1, libtbb.so.2, libtbb.so.12, libjpeg.so.8, libwebp.so.7, libpng16.so.16, libgdcmMSFF.so.3.0, libtiff.so.5, libopenjp2.so.7, libgdal.so.30, libgdcmDSED.so.3.0, libgdcmDICT.so.3.0, libgdcmjpeg8.so.3.0, libgdcmjpeg12.so.3.0, libgdcmjpeg16.so.3.0, libcharls.so.2, libuuid.so.1, libjson-c.so.5, libgdcmIOD.so.3.0, libgdcmCommon.so.3.0, libzstd.so.1, liblzma.so.5, libjbig.so.0, libdeflate.so.0, libheif.so.1, libarmadillo.so.10, libpoppler.so.118, libfreexl.so.1, libodbc.so.2, libodbcinst.so.2, libkmlbase.so.1, libkmldom.so.1, libkmlengine.so.1, libexpat.so.1, libxerces-c-3.2.so, libnetcdf.so.19, libmfhdfalt.so.0, libdfalt.so.0, libogdi.so.4.1, libgif.so.7, libgeotiff.so.5, libcfitsio.so.9, libpq.so.5, liblz4.so.1, libblosc.so.1, libproj.so.22, libsqlite3.so.0, libspatialite.so.7, libpcre2-8.so.0, libcurl.so.4, libfyba.so.0, libxml2.so.2, libmysqlclient.so.21, libcrypto.so.3, libaom.so.3, libde265.so.0, libx265.so.199, libdav1d.so.5, libblas.so.3, liblapack.so.3, libarpack.so.2, libsuperlu.so.5, libfreetype.so.6, libfontconfig.so.1, liblcms2.so.2, libnss3.so, libsmime3.so, libplc4.so, libnspr4.so, libgeos.so.3.10.2, libltdl.so.7, libminizip.so.1, liburiparser.so.1, libcurl-gnutls.so.4, libicuuc.so.70, libsz.so.2, libtirpc.so.3, libdl.so.2, libbz2.so.1.0, libssl.so.3, libldap-2.5.so.0, libsnappy.so.1, librttopo.so.1, libnghttp2.so.14, libidn2.so.0, librtmp.so.1, libssh.so.4, libpsl.so.5, liblber-2.5.so.0, libbrotlidec.so.1, libfyut.so.0, libfygm.so.0, libresolv.so.2, libpthread.so.0, libnuma.so.1, libopenblas.so.0, libgfortran.so.5, libnssutil3.so, libplds4.so, libnettle.so.8, libgnutls.so.30, libicudata.so.70, libaec.so.0, libkrb5.so.3, libk5crypto.so.3, libkrb5support.so.0, libsasl2.so.2, libunistring.so.2, libhogweed.so.6, libgmp.so.10, libbrotlicommon.so.1, libquadmath.so.0, libp11-kit.so.0, libtasn1.so.6, libkeyutils.so.1, libffi.so.8.

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)