vulkan tests

AMD Ryzen Threadripper 3990X 64-Core testing with a Gigabyte TRX40 AORUS PRO WIFI (F6 BIOS) and AMD Radeon RX 5700 8GB on Ubuntu 23.04 via the Phoronix Test Suite.

Compare your own system(s) to this result file with the Phoronix Test Suite by running the command: phoronix-test-suite benchmark 2308016-PTS-VULKANTE10
Jump To Table - Results

View

Do Not Show Noisy Results
Do Not Show Results With Incomplete Data
Do Not Show Results With Little Change/Spread
List Notable Results
Show Result Confidence Charts

Limit displaying results to tests within:

NVIDIA GPU Compute 4 Tests
Vulkan Compute 4 Tests

Statistics

Show Overall Harmonic Mean(s)
Show Overall Geometric Mean
Show Geometric Means Per-Suite/Category
Show Wins / Losses Counts (Pie Chart)
Normalize Results
Remove Outliers Before Calculating Averages

Graph Settings

Force Line Graphs Where Applicable
Convert To Scalar Where Applicable
Prefer Vertical Bar Graphs

Multi-Way Comparison

Condense Multi-Option Tests Into Single Result Graphs

Table

Show Detailed System Result Table

Run Management

Highlight
Result
Hide
Result
Result
Identifier
Performance Per
Dollar
Date
Run
  Test
  Duration
a
August 01 2023
  7 Hours, 16 Minutes
b
August 01 2023
  14 Hours, 50 Minutes
c
August 01 2023
  7 Hours, 18 Minutes
Invert Hiding All Results Option
  9 Hours, 48 Minutes

Only show results where is faster than
Only show results matching title/arguments (delimit multiple options with a comma):
Do not show results matching title/arguments (delimit multiple options with a comma):

a

Kernel Notes: Transparent Huge Pages: madvise
Compiler Notes: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-defaulted --enable-offload-targets=nvptx-none=/build/gcc-12-Pa930Z/gcc-12-12.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-12-Pa930Z/gcc-12-12.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -v
Processor Notes: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0x830107a
Graphics Notes: BAR1 / Visible vRAM Size: 256 MB - vBIOS Version: 113-D1820201-101
Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: Mitigation of untrained return thunk; SMT enabled with STIBP protection + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Retpolines IBPB: conditional STIBP: always-on RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

b

c

Processor: AMD Ryzen Threadripper 3990X 64-Core @ 2.90GHz (64 Cores / 128 Threads), Motherboard: Gigabyte TRX40 AORUS PRO WIFI (F6 BIOS), Chipset: AMD Starship/Matisse, Memory: 128GB, Disk: Samsung SSD 970 EVO Plus 500GB, Graphics: AMD Radeon RX 5700 8GB (1750/875MHz), Audio: AMD Navi 10 HDMI Audio, Monitor: DELL P2415Q, Network: Intel I211 + Intel Wi-Fi 6 AX200

OS: Ubuntu 23.04, Kernel: 6.2.0-26-generic (x86_64), Desktop: GNOME Shell 44.0, Display Server: X Server + Wayland, OpenGL: 4.6 Mesa 23.0.2 (LLVM 15.0.7 DRM 3.49), Compiler: GCC 12.2.0, File-System: ext4, Screen Resolution: 3840x2160

vulkan testsOpenBenchmarking.orgPhoronix Test SuiteAMD Ryzen Threadripper 3990X 64-Core @ 2.90GHz (64 Cores / 128 Threads)Gigabyte TRX40 AORUS PRO WIFI (F6 BIOS)AMD Starship/Matisse128GBSamsung SSD 970 EVO Plus 500GBAMD Radeon RX 5700 8GB (1750/875MHz)AMD Navi 10 HDMI AudioDELL P2415QIntel I211 + Intel Wi-Fi 6 AX200Ubuntu 23.046.2.0-26-generic (x86_64)GNOME Shell 44.0X Server + Wayland4.6 Mesa 23.0.2 (LLVM 15.0.7 DRM 3.49)GCC 12.2.0ext43840x2160ProcessorMotherboardChipsetMemoryDiskGraphicsAudioMonitorNetworkOSKernelDesktopDisplay ServerOpenGLCompilerFile-SystemScreen ResolutionVulkan Tests PerformanceSystem Logs- Transparent Huge Pages: madvise- --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-defaulted --enable-offload-targets=nvptx-none=/build/gcc-12-Pa930Z/gcc-12-12.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-12-Pa930Z/gcc-12-12.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -v - Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0x830107a- BAR1 / Visible vRAM Size: 256 MB - vBIOS Version: 113-D1820201-101- itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: Mitigation of untrained return thunk; SMT enabled with STIBP protection + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Retpolines IBPB: conditional STIBP: always-on RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

abcResult OverviewPhoronix Test Suite100%101%101%102%102%NCNNvkpeakVkFFTVkResample

vulkan testsncnn: Vulkan GPU - alexnetncnn: CPU - resnet50ncnn: CPU - regnety_400mncnn: CPU - vgg16ncnn: Vulkan GPU - vision_transformerncnn: Vulkan GPU - resnet50ncnn: CPU - mobilenetncnn: Vulkan GPU - yolov4-tinyncnn: Vulkan GPU - mobilenetncnn: CPU - squeezenet_ssdncnn: CPU - resnet18ncnn: CPU - googlenetncnn: Vulkan GPU - resnet18ncnn: CPU - yolov4-tinyncnn: Vulkan GPU - regnety_400mncnn: Vulkan GPU - vgg16ncnn: Vulkan GPU - squeezenet_ssdncnn: Vulkan GPU - googlenetvkpeak: fp32-scalarvkpeak: fp32-vec4vkpeak: fp16-vec4vkfft: FFT + iFFT R2C / C2Rvkpeak: int16-vec4vkfft: FFT + iFFT C2C 1D batched in half precisionvkpeak: fp16-scalarvkfft: FFT + iFFT C2C multidimensional in single precisionvkfft: FFT + iFFT C2C Bluestein in single precisionvkpeak: fp64-vec4vkpeak: fp64-scalarvkpeak: int32-vec4vkpeak: int32-scalarvkfft: FFT + iFFT C2C Bluestein benchmark in double precisionvkpeak: int16-scalarvkresample: 2x - Singlevkfft: FFT + iFFT C2C 1D batched in double precisionvkfft: FFT + iFFT C2C 1D batched in single precisionvkfft: FFT + iFFT C2C 1D batched in single precision, no reshufflingncnn: Vulkan GPU - FastestDetncnn: Vulkan GPU - blazefacencnn: Vulkan GPU - efficientnet-b0ncnn: Vulkan GPU - mnasnetncnn: Vulkan GPU - shufflenet-v2ncnn: Vulkan GPU-v3-v3 - mobilenet-v3ncnn: Vulkan GPU-v2-v2 - mobilenet-v2ncnn: CPU - FastestDetncnn: CPU - vision_transformerncnn: CPU - alexnetncnn: CPU - blazefacencnn: CPU - efficientnet-b0ncnn: CPU - mnasnetncnn: CPU - shufflenet-v2ncnn: CPU-v3-v3 - mobilenet-v3ncnn: CPU-v2-v2 - mobilenet-v2abc15.2235.2045.4345.6475.1738.9125.2136.7224.6622.3019.0730.8719.2037.7246.7347.5221.8131.437836.617781.3412516.942552512531.351009037853.66169937268493.28493.361574.821214.4130667856.2112.77018431589446414014.257.2218.4214.1814.7512.9715.7114.6673.4711.765.8918.0013.1314.0812.7313.8315.3439.0348.0347.5372.2639.8425.7837.7925.2522.2719.4031.5519.4237.3846.2248.1322.0131.247856.197808.5612550.912544712550.381010637860.29169767258492.96493.011574.601213.4830647852.0312.77118428589476413914.537.0919.4713.7214.2713.2514.7813.7074.1113.977.1219.6912.9014.3813.4514.4913.0738.1847.0246.8872.7038.4325.9637.3625.3722.8918.9131.6419.0437.1446.1047.5021.7831.467866.837808.4812559.312545112563.141008397866.33169677259493.57493.511575.911214.4030647857.0112.77518426589386413113.886.0818.0212.9313.2912.5713.6316.0375.9612.376.9519.0615.2714.1213.6016.01OpenBenchmarking.org

NCNN

NCNN is a high performance neural network inference framework optimized for mobile and other platforms developed by Tencent. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: alexnetbac48121620SE +/- 0.17, N = 12SE +/- 0.36, N = 3SE +/- 0.28, N = 315.3415.2213.07MIN: 12.01 / MAX: 62MIN: 11.93 / MAX: 21.72MIN: 10.71 / MAX: 16.271. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: resnet50bca918273645SE +/- 0.53, N = 3SE +/- 1.02, N = 3SE +/- 0.28, N = 339.0338.1835.20MIN: 34.39 / MAX: 46.86MIN: 32.56 / MAX: 45.61MIN: 30.71 / MAX: 97.131. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: regnety_400mbca1122334455SE +/- 0.70, N = 3SE +/- 0.33, N = 3SE +/- 1.49, N = 348.0347.0245.43MIN: 41.67 / MAX: 101.38MIN: 41.08 / MAX: 93.32MIN: 39.85 / MAX: 90.631. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: vgg16bca1122334455SE +/- 0.34, N = 3SE +/- 0.66, N = 3SE +/- 0.74, N = 347.5346.8845.64MIN: 44.76 / MAX: 58.97MIN: 43.93 / MAX: 59.11MIN: 42.07 / MAX: 55.261. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: vision_transformeracb20406080100SE +/- 2.14, N = 3SE +/- 1.92, N = 3SE +/- 0.80, N = 1275.1772.7072.26MIN: 69.97 / MAX: 147.06MIN: 69.83 / MAX: 164MIN: 69.21 / MAX: 182.161. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: resnet50bac918273645SE +/- 0.24, N = 12SE +/- 0.49, N = 3SE +/- 0.59, N = 339.8438.9138.43MIN: 32.55 / MAX: 100.04MIN: 33.52 / MAX: 45.34MIN: 33.37 / MAX: 44.581. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: mobilenetcba612182430SE +/- 0.21, N = 3SE +/- 0.28, N = 3SE +/- 0.05, N = 325.9625.7825.21MIN: 22.99 / MAX: 30.78MIN: 22.57 / MAX: 80.73MIN: 21.72 / MAX: 91.81. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: yolov4-tinybca918273645SE +/- 0.27, N = 12SE +/- 0.31, N = 3SE +/- 0.55, N = 337.7937.3636.72MIN: 33.15 / MAX: 109.38MIN: 33.45 / MAX: 45.21MIN: 32.78 / MAX: 42.581. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: mobilenetcba612182430SE +/- 0.04, N = 3SE +/- 0.39, N = 12SE +/- 0.20, N = 325.3725.2524.66MIN: 22.3 / MAX: 30.62MIN: 21.74 / MAX: 83.58MIN: 22.49 / MAX: 28.961. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: squeezenet_ssdcab510152025SE +/- 0.30, N = 3SE +/- 0.47, N = 3SE +/- 0.39, N = 322.8922.3022.27MIN: 20.98 / MAX: 29.54MIN: 20.67 / MAX: 28.29MIN: 20.98 / MAX: 27.691. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: resnet18bac510152025SE +/- 0.21, N = 3SE +/- 0.16, N = 3SE +/- 0.03, N = 319.4019.0718.91MIN: 16.25 / MAX: 94.41MIN: 16.37 / MAX: 82.83MIN: 16.19 / MAX: 66.71. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: googlenetcba714212835SE +/- 0.41, N = 3SE +/- 0.22, N = 3SE +/- 0.21, N = 331.6431.5530.87MIN: 26.58 / MAX: 37.83MIN: 29.53 / MAX: 39MIN: 27.29 / MAX: 90.041. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: resnet18bac510152025SE +/- 0.09, N = 12SE +/- 0.06, N = 3SE +/- 0.17, N = 319.4219.2019.04MIN: 16.13 / MAX: 88.96MIN: 16.68 / MAX: 25.67MIN: 16.24 / MAX: 22.761. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: yolov4-tinyabc918273645SE +/- 0.32, N = 3SE +/- 0.58, N = 3SE +/- 0.31, N = 337.7237.3837.14MIN: 34.14 / MAX: 48.53MIN: 33.26 / MAX: 45.88MIN: 33.68 / MAX: 43.481. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: regnety_400mabc1122334455SE +/- 0.37, N = 3SE +/- 0.51, N = 12SE +/- 0.73, N = 346.7346.2246.10MIN: 39.79 / MAX: 151.9MIN: 39.54 / MAX: 121.82MIN: 39.52 / MAX: 107.361. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: vgg16bac1122334455SE +/- 0.15, N = 12SE +/- 0.06, N = 3SE +/- 0.08, N = 348.1347.5247.50MIN: 44.54 / MAX: 88.13MIN: 45.43 / MAX: 59.56MIN: 45.06 / MAX: 62.771. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: squeezenet_ssdbac510152025SE +/- 0.08, N = 12SE +/- 0.21, N = 3SE +/- 0.20, N = 322.0121.8121.78MIN: 20.55 / MAX: 81.08MIN: 20.68 / MAX: 25.84MIN: 20.62 / MAX: 26.091. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: googlenetcab714212835SE +/- 0.10, N = 3SE +/- 0.18, N = 3SE +/- 0.16, N = 1231.4631.4331.24MIN: 26.64 / MAX: 38.17MIN: 26.65 / MAX: 84.75MIN: 27.01 / MAX: 86.351. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

vkpeak

Vkpeak is a Vulkan compute benchmark inspired by OpenCL's clpeak. Vkpeak provides Vulkan compute performance measurements for FP16 / FP32 / FP64 / INT16 / INT32 scalar and vec4 performance. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgGFLOPS, More Is Bettervkpeak 20230730fp32-scalarabc2K4K6K8K10KSE +/- 0.62, N = 3SE +/- 1.09, N = 3SE +/- 1.93, N = 37836.617856.197866.83

OpenBenchmarking.orgGFLOPS, More Is Bettervkpeak 20230730fp32-vec4acb2K4K6K8K10KSE +/- 7.57, N = 3SE +/- 8.31, N = 3SE +/- 7.17, N = 37781.347808.487808.56

OpenBenchmarking.orgGFLOPS, More Is Bettervkpeak 20230730fp16-vec4abc3K6K9K12K15KSE +/- 1.96, N = 3SE +/- 1.15, N = 3SE +/- 1.25, N = 312516.9412550.9112559.31

VkFFT

OpenBenchmarking.orgBenchmark Score, More Is BetterVkFFT 1.2.31Test: FFT + iFFT R2C / C2Rbca5K10K15K20K25KSE +/- 4.33, N = 3SE +/- 2.03, N = 3SE +/- 73.86, N = 32544725451255251. (CXX) g++ options: -O3

vkpeak

Vkpeak is a Vulkan compute benchmark inspired by OpenCL's clpeak. Vkpeak provides Vulkan compute performance measurements for FP16 / FP32 / FP64 / INT16 / INT32 scalar and vec4 performance. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgGIOPS, More Is Bettervkpeak 20230730int16-vec4abc3K6K9K12K15KSE +/- 0.58, N = 3SE +/- 0.77, N = 3SE +/- 1.54, N = 312531.3512550.3812563.14

VkFFT

OpenBenchmarking.orgBenchmark Score, More Is BetterVkFFT 1.2.31Test: FFT + iFFT C2C 1D batched in half precisioncab20K40K60K80K100KSE +/- 219.62, N = 3SE +/- 98.36, N = 3SE +/- 43.64, N = 31008391009031010631. (CXX) g++ options: -O3

vkpeak

Vkpeak is a Vulkan compute benchmark inspired by OpenCL's clpeak. Vkpeak provides Vulkan compute performance measurements for FP16 / FP32 / FP64 / INT16 / INT32 scalar and vec4 performance. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgGFLOPS, More Is Bettervkpeak 20230730fp16-scalarabc2K4K6K8K10KSE +/- 1.36, N = 3SE +/- 0.75, N = 3SE +/- 1.85, N = 37853.667860.297866.33

VkFFT

OpenBenchmarking.orgBenchmark Score, More Is BetterVkFFT 1.2.31Test: FFT + iFFT C2C multidimensional in single precisioncba4K8K12K16K20KSE +/- 9.53, N = 3SE +/- 12.01, N = 3SE +/- 51.10, N = 31696716976169931. (CXX) g++ options: -O3

OpenBenchmarking.orgBenchmark Score, More Is BetterVkFFT 1.2.31Test: FFT + iFFT C2C Bluestein in single precisionbca16003200480064008000SE +/- 2.67, N = 3SE +/- 4.70, N = 3SE +/- 15.24, N = 37258725972681. (CXX) g++ options: -O3

vkpeak

Vkpeak is a Vulkan compute benchmark inspired by OpenCL's clpeak. Vkpeak provides Vulkan compute performance measurements for FP16 / FP32 / FP64 / INT16 / INT32 scalar and vec4 performance. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgGFLOPS, More Is Bettervkpeak 20230730fp64-vec4bac110220330440550SE +/- 0.03, N = 3SE +/- 0.11, N = 3SE +/- 0.10, N = 3492.96493.28493.57

OpenBenchmarking.orgGFLOPS, More Is Bettervkpeak 20230730fp64-scalarbac110220330440550SE +/- 0.03, N = 3SE +/- 0.10, N = 3SE +/- 0.07, N = 3493.01493.36493.51

OpenBenchmarking.orgGIOPS, More Is Bettervkpeak 20230730int32-vec4bac30060090012001500SE +/- 0.05, N = 3SE +/- 0.17, N = 3SE +/- 0.21, N = 31574.601574.821575.91

OpenBenchmarking.orgGIOPS, More Is Bettervkpeak 20230730int32-scalarbca30060090012001500SE +/- 0.05, N = 3SE +/- 0.32, N = 3SE +/- 0.27, N = 31213.481214.401214.41

VkFFT

OpenBenchmarking.orgBenchmark Score, More Is BetterVkFFT 1.2.31Test: FFT + iFFT C2C Bluestein benchmark in double precisionbca7001400210028003500SE +/- 1.00, N = 3SE +/- 1.53, N = 3SE +/- 4.33, N = 33064306430661. (CXX) g++ options: -O3

vkpeak

Vkpeak is a Vulkan compute benchmark inspired by OpenCL's clpeak. Vkpeak provides Vulkan compute performance measurements for FP16 / FP32 / FP64 / INT16 / INT32 scalar and vec4 performance. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgGIOPS, More Is Bettervkpeak 20230730int16-scalarbac2K4K6K8K10KSE +/- 0.51, N = 3SE +/- 0.44, N = 3SE +/- 0.96, N = 37852.037856.217857.01

VkResample

VkResample is a Vulkan-based image upscaling library based on VkFFT. The sample input file is upscaling a 4K image to 8K using Vulkan-based GPU acceleration. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterVkResample 1.0Upscale: 2x - Precision: Singlecba3691215SE +/- 0.00, N = 3SE +/- 0.00, N = 3SE +/- 0.00, N = 312.7812.7712.771. (CXX) g++ options: -O3

VkFFT

OpenBenchmarking.orgBenchmark Score, More Is BetterVkFFT 1.2.31Test: FFT + iFFT C2C 1D batched in double precisioncba4K8K12K16K20KSE +/- 1.86, N = 3SE +/- 3.61, N = 3SE +/- 6.17, N = 31842618428184311. (CXX) g++ options: -O3

OpenBenchmarking.orgBenchmark Score, More Is BetterVkFFT 1.2.31Test: FFT + iFFT C2C 1D batched in single precisioncab13K26K39K52K65KSE +/- 1.20, N = 3SE +/- 1.33, N = 3SE +/- 3.33, N = 35893858944589471. (CXX) g++ options: -O3

OpenBenchmarking.orgBenchmark Score, More Is BetterVkFFT 1.2.31Test: FFT + iFFT C2C 1D batched in single precision, no reshufflingcba14K28K42K56K70KSE +/- 7.69, N = 3SE +/- 5.46, N = 3SE +/- 6.43, N = 36413164139641401. (CXX) g++ options: -O3

NCNN

NCNN is a high performance neural network inference framework optimized for mobile and other platforms developed by Tencent. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: FastestDetbac48121620SE +/- 0.63, N = 12SE +/- 1.49, N = 3SE +/- 0.66, N = 314.5314.2513.88MIN: 11.05 / MAX: 27.1MIN: 11.35 / MAX: 22.25MIN: 11.52 / MAX: 18.831. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: blazefaceabc246810SE +/- 0.32, N = 3SE +/- 0.38, N = 12SE +/- 0.20, N = 37.227.096.08MIN: 5.7 / MAX: 10.52MIN: 4.88 / MAX: 78.29MIN: 5.01 / MAX: 7.581. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: efficientnet-b0bac510152025SE +/- 0.83, N = 12SE +/- 0.44, N = 3SE +/- 0.52, N = 319.4718.4218.02MIN: 13.59 / MAX: 97.32MIN: 14.24 / MAX: 27.56MIN: 13.61 / MAX: 27.871. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: mnasnetabc48121620SE +/- 2.23, N = 3SE +/- 0.74, N = 12SE +/- 1.40, N = 314.1813.7212.93MIN: 9.68 / MAX: 23.38MIN: 9.31 / MAX: 23.03MIN: 9.34 / MAX: 22.141. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU - Model: shufflenet-v2abc48121620SE +/- 1.24, N = 3SE +/- 0.46, N = 12SE +/- 0.21, N = 314.7514.2713.29MIN: 11.39 / MAX: 23.44MIN: 10.94 / MAX: 22.32MIN: 11.24 / MAX: 22.871. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU-v3-v3 - Model: mobilenet-v3bac3691215SE +/- 0.62, N = 12SE +/- 1.36, N = 3SE +/- 0.32, N = 313.2512.9712.57MIN: 9.52 / MAX: 53.34MIN: 9.71 / MAX: 20.92MIN: 9.6 / MAX: 17.011. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: Vulkan GPU-v2-v2 - Model: mobilenet-v2abc48121620SE +/- 3.00, N = 3SE +/- 0.85, N = 12SE +/- 0.74, N = 315.7114.7813.63MIN: 10.21 / MAX: 26.19MIN: 10.2 / MAX: 70.67MIN: 10.24 / MAX: 27.871. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: FastestDetcab48121620SE +/- 0.49, N = 3SE +/- 1.04, N = 3SE +/- 0.53, N = 316.0314.6613.70MIN: 11.61 / MAX: 24.19MIN: 11.69 / MAX: 19.64MIN: 11.46 / MAX: 17.731. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: vision_transformercba20406080100SE +/- 2.22, N = 3SE +/- 2.67, N = 3SE +/- 2.54, N = 375.9674.1173.47MIN: 71.06 / MAX: 93.63MIN: 70.44 / MAX: 113.62MIN: 70.04 / MAX: 83.771. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: alexnetbca48121620SE +/- 0.45, N = 3SE +/- 0.24, N = 3SE +/- 0.62, N = 313.9712.3711.76MIN: 11.94 / MAX: 26.76MIN: 10.09 / MAX: 15.86MIN: 9.49 / MAX: 19.161. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: blazefacebca246810SE +/- 0.58, N = 3SE +/- 0.44, N = 3SE +/- 0.59, N = 37.126.955.89MIN: 5.21 / MAX: 13.65MIN: 5.22 / MAX: 9.49MIN: 4.78 / MAX: 9.521. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: efficientnet-b0bca510152025SE +/- 1.68, N = 3SE +/- 1.07, N = 3SE +/- 1.70, N = 319.6919.0618.00MIN: 13.76 / MAX: 27.29MIN: 13.85 / MAX: 27.95MIN: 13.54 / MAX: 69.841. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: mnasnetcab48121620SE +/- 0.65, N = 3SE +/- 1.54, N = 3SE +/- 1.05, N = 315.2713.1312.90MIN: 10.33 / MAX: 25.75MIN: 9.27 / MAX: 22.76MIN: 9.19 / MAX: 19.691. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU - Model: shufflenet-v2bca48121620SE +/- 0.93, N = 3SE +/- 1.04, N = 3SE +/- 1.17, N = 314.3814.1214.08MIN: 11.05 / MAX: 20.51MIN: 11.12 / MAX: 22.91MIN: 10.99 / MAX: 18.541. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU-v3-v3 - Model: mobilenet-v3cba3691215SE +/- 0.59, N = 3SE +/- 0.98, N = 3SE +/- 1.00, N = 313.6013.4512.73MIN: 9.51 / MAX: 51.49MIN: 9.44 / MAX: 19.16MIN: 9.51 / MAX: 20.961. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20230517Target: CPU-v2-v2 - Model: mobilenet-v2cba48121620SE +/- 1.00, N = 3SE +/- 1.52, N = 3SE +/- 1.66, N = 316.0114.4913.83MIN: 11.37 / MAX: 25.15MIN: 10.12 / MAX: 23.01MIN: 9.93 / MAX: 25.721. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread

a

Kernel Notes: Transparent Huge Pages: madvise
Compiler Notes: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-defaulted --enable-offload-targets=nvptx-none=/build/gcc-12-Pa930Z/gcc-12-12.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-12-Pa930Z/gcc-12-12.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -v
Processor Notes: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0x830107a
Graphics Notes: BAR1 / Visible vRAM Size: 256 MB - vBIOS Version: 113-D1820201-101
Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: Mitigation of untrained return thunk; SMT enabled with STIBP protection + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Retpolines IBPB: conditional STIBP: always-on RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

Testing initiated at 1 August 2023 13:36 by user pts.

b

Kernel Notes: Transparent Huge Pages: madvise
Compiler Notes: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-defaulted --enable-offload-targets=nvptx-none=/build/gcc-12-Pa930Z/gcc-12-12.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-12-Pa930Z/gcc-12-12.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -v
Processor Notes: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0x830107a
Graphics Notes: BAR1 / Visible vRAM Size: 256 MB - vBIOS Version: 113-D1820201-101
Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: Mitigation of untrained return thunk; SMT enabled with STIBP protection + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Retpolines IBPB: conditional STIBP: always-on RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

Testing initiated at 1 August 2023 14:33 by user pts.

c

Processor: AMD Ryzen Threadripper 3990X 64-Core @ 2.90GHz (64 Cores / 128 Threads), Motherboard: Gigabyte TRX40 AORUS PRO WIFI (F6 BIOS), Chipset: AMD Starship/Matisse, Memory: 128GB, Disk: Samsung SSD 970 EVO Plus 500GB, Graphics: AMD Radeon RX 5700 8GB (1750/875MHz), Audio: AMD Navi 10 HDMI Audio, Monitor: DELL P2415Q, Network: Intel I211 + Intel Wi-Fi 6 AX200

OS: Ubuntu 23.04, Kernel: 6.2.0-26-generic (x86_64), Desktop: GNOME Shell 44.0, Display Server: X Server + Wayland, OpenGL: 4.6 Mesa 23.0.2 (LLVM 15.0.7 DRM 3.49), Compiler: GCC 12.2.0, File-System: ext4, Screen Resolution: 3840x2160

Kernel Notes: Transparent Huge Pages: madvise
Compiler Notes: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-defaulted --enable-offload-targets=nvptx-none=/build/gcc-12-Pa930Z/gcc-12-12.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-12-Pa930Z/gcc-12-12.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -v
Processor Notes: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0x830107a
Graphics Notes: BAR1 / Visible vRAM Size: 256 MB - vBIOS Version: 113-D1820201-101
Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: Mitigation of untrained return thunk; SMT enabled with STIBP protection + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Retpolines IBPB: conditional STIBP: always-on RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

Testing initiated at 1 August 2023 16:40 by user pts.