NVIDIA GeForce RTX 3090 AMD Ryzen 9 7950X 16-Core testing with a ASUS TUF GAMING X670E-PLUS WIFI (0613 BIOS) and Gigabyte NVIDIA GeForce RTX 4090 24GB on Ubuntu 22.04 via the Phoronix Test Suite.
HTML result view exported from: https://openbenchmarking.org/result/2210138-NE-2112069PT61&grs&sor .
NVIDIA GeForce RTX 3090 Processor Motherboard Chipset Memory Disk Graphics Audio Monitor Network OS Kernel Desktop Display Server Display Driver OpenGL OpenCL Vulkan Compiler File-System Screen Resolution RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 RTX 4090 AMD Ryzen 9 5950X 16-Core @ 3.40GHz (16 Cores / 32 Threads) ASUS ROG CROSSHAIR VIII HERO (WI-FI) (3801 BIOS) AMD Starship/Matisse 32GB 1000GB Sabrent Rocket 4.0 Plus NVIDIA GeForce RTX 3090 24GB NVIDIA GA102 HD Audio ASUS MG28U Realtek RTL8125 2.5GbE + Intel I211 + Intel Wi-Fi 6 AX200 Ubuntu 21.10 5.13.0-22-generic (x86_64) GNOME Shell 40.5 X Server 1.20.13 NVIDIA 495.44 4.6.0 OpenCL 3.0 CUDA 11.5.100 1.2.186 GCC 11.2.0 + Clang 13.0.0-2 ext4 3840x2160 AMD Ryzen 9 7950X 16-Core @ 4.50GHz (16 Cores / 32 Threads) ASUS TUF GAMING X670E-PLUS WIFI (0613 BIOS) AMD Device 14d8 2048GB XPG GAMMIX S70 BLADE + 4001GB SSD 870 QVO 4TB Gigabyte NVIDIA GeForce RTX 4090 24GB NVIDIA Device 22ba PI-KVM Video Realtek RTL8125 2.5GbE + MEDIATEK Device 0608 Ubuntu 22.04 5.15.0-25-generic (x86_64) GNOME Shell 42.4 X Server 1.21.1.3 NVIDIA 520.56.06 OpenCL 3.0 CUDA 11.8.87 1.3.205 GCC 13.0.0 20221013 + Clang 14.0.0-1ubuntu1 1920x1080 OpenBenchmarking.org Kernel Details - Transparent Huge Pages: madvise Compiler Details - RTX 3090: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-targets=nvptx-none=/build/gcc-11-ZPT0kp/gcc-11-11.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-ZPT0kp/gcc-11-11.2.0/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-build-config=bootstrap-lto-lean --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 - NVIDIA RTX 3090: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-targets=nvptx-none=/build/gcc-11-ZPT0kp/gcc-11-11.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-ZPT0kp/gcc-11-11.2.0/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-build-config=bootstrap-lto-lean --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 - NVIDIA 3090: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-targets=nvptx-none=/build/gcc-11-ZPT0kp/gcc-11-11.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-ZPT0kp/gcc-11-11.2.0/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-build-config=bootstrap-lto-lean --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 - RTX 4090: --build=x86_64-linux-gnu --disable-multilib --enable-checking=release --enable-languages=c,c++ --host=x86_64-linux-gnu --target=x86_64-linux-gnu -v Processor Details - RTX 3090: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0xa201016 - NVIDIA RTX 3090: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0xa201016 - NVIDIA 3090: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0xa201016 - RTX 4090: Scaling Governor: acpi-cpufreq performance (Boost: Enabled) - CPU Microcode: 0xa601203 Graphics Details - RTX 3090: BAR1 / Visible vRAM Size: 32768 MiB - NVIDIA RTX 3090: BAR1 / Visible vRAM Size: 32768 MiB - NVIDIA 3090: BAR1 / Visible vRAM Size: 32768 MiB - RTX 4090: BAR1 / Visible vRAM Size: 32768 MiB - vBIOS Version: 95.02.18.00.d2 OpenCL Details - RTX 3090: GPU Compute Cores: 10496 - NVIDIA RTX 3090: GPU Compute Cores: 10496 - NVIDIA 3090: GPU Compute Cores: 10496 - RTX 4090: GPU Compute Cores: 16384 Python Details - RTX 3090: Python 3.9.7 - NVIDIA RTX 3090: Python 3.9.7 - NVIDIA 3090: Python 3.9.7 - RTX 4090: Python 3.10.6 Security Details - RTX 3090: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + srbds: Not affected + tsx_async_abort: Not affected - NVIDIA RTX 3090: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + srbds: Not affected + tsx_async_abort: Not affected - NVIDIA 3090: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional IBRS_FW STIBP: always-on RSB filling + srbds: Not affected + tsx_async_abort: Not affected - RTX 4090: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Retpolines IBPB: conditional IBRS_FW STIBP: always-on RSB filling + srbds: Not affected + tsx_async_abort: Not affected
NVIDIA GeForce RTX 3090 shoc: OpenCL - Bus Speed Readback shoc: OpenCL - Bus Speed Download shoc: OpenCL - Triad shoc: OpenCL - GEMM SGEMM_N viennacl: CPU BLAS - sAXPY viennacl: CPU BLAS - sCOPY paraview: Wavelet Volume - 3840 x 2160 paraview: Wavelet Volume - 3840 x 2160 viennacl: CPU BLAS - dAXPY viennacl: CPU BLAS - dCOPY shoc: OpenCL - Reduction viennacl: CPU BLAS - sDOT clpeak: Integer Compute INT vkpeak: int16-vec4 viennacl: CPU BLAS - dDOT clpeak: Single-Precision Float viennacl: OpenCL BLAS - dGEMM-TT hashcat: TrueCrypt RIPEMD160 + XTS hashcat: 7-Zip hashcat: SHA1 shoc: OpenCL - Max SP Flops hashcat: SHA-512 vkpeak: int16-scalar vkpeak: fp64-scalar blender: Pabellon Barcelona - CUDA vkpeak: fp64-vec4 viennacl: OpenCL BLAS - dGEMM-TN vkpeak: int32-vec4 vkpeak: int32-scalar vkpeak: fp16-vec4 vkpeak: fp16-scalar vkpeak: fp32-vec4 vkresample: 2x - Double vkpeak: fp32-scalar clpeak: Double-Precision Double shoc: OpenCL - MD5 Hash viennacl: OpenCL BLAS - dGEMM-NT hashcat: MD5 financebench: Black-Scholes OpenCL ncnn: Vulkan GPU-v3-v3 - mobilenet-v3 paraview: Wavelet Volume - 2560 x 1440 paraview: Wavelet Volume - 2560 x 1440 paraview: Wavelet Contour - 3840 x 2160 paraview: Wavelet Contour - 3840 x 2160 viennacl: OpenCL BLAS - dGEMM-NN luxcorerender: Danish Mood - GPU octanebench: Total Score paraview: Many Spheres - 3840 x 2160 paraview: Many Spheres - 3840 x 2160 paraview: Many Spheres - 1920 x 1200 paraview: Many Spheres - 1920 x 1200 blender: Fishy Cat - NVIDIA OptiX rodinia: OpenCL Particle Filter paraview: Many Spheres - 2560 x 1440 paraview: Many Spheres - 2560 x 1440 paraview: Many Spheres - 1920 x 1080 paraview: Many Spheres - 1920 x 1080 blender: Barbershop - CUDA luxcorerender: DLSC - GPU blender: Fishy Cat - CUDA viennacl: CPU BLAS - dGEMV-T paraview: Wavelet Volume - 1920 x 1200 paraview: Wavelet Volume - 1920 x 1200 vkfft: blender: Classroom - CUDA blender: Barbershop - NVIDIA OptiX luxcorerender: Orange Juice - GPU paraview: Wavelet Volume - 1920 x 1080 paraview: Wavelet Volume - 1920 x 1080 luxcorerender: LuxCore Benchmark - GPU blender: Pabellon Barcelona - NVIDIA OptiX ncnn: Vulkan GPU - efficientnet-b0 blender: Classroom - NVIDIA OptiX arrayfire: Conjugate Gradient OpenCL indigobench: OpenCL GPU - Bedroom ncnn: Vulkan GPU - mnasnet ncnn: Vulkan GPU-v2-v2 - mobilenet-v2 ncnn: Vulkan GPU - shufflenet-v2 realsr-ncnn: 4x - Yes blender: BMW27 - NVIDIA OptiX ncnn: Vulkan GPU - regnety_400m ncnn: Vulkan GPU - yolov4-tiny paraview: Wavelet Contour - 2560 x 1440 paraview: Wavelet Contour - 2560 x 1440 shoc: OpenCL - S3D ncnn: Vulkan GPU - resnet18 paraview: Wavelet Contour - 1920 x 1200 paraview: Wavelet Contour - 1920 x 1200 unvanquished: 1920 x 1080 - High ncnn: Vulkan GPU - mobilenet indigobench: OpenCL GPU - Supercar realsr-ncnn: 4x - No viennacl: CPU BLAS - dGEMM-NN ncnn: Vulkan GPU - blazeface unvanquished: 1920 x 1200 - High unvanquished: 1920 x 1200 - Medium waifu2x-ncnn: 2x - 3 - Yes paraview: Wavelet Contour - 1920 x 1080 paraview: Wavelet Contour - 1920 x 1080 unvanquished: 1920 x 1200 - Ultra shoc: OpenCL - FFT SP unvanquished: 2560 x 1440 - Medium unvanquished: 2560 x 1440 - High unvanquished: 3840 x 2160 - Medium namd-cuda: ATPase Simulation - 327,506 Atoms unvanquished: 1920 x 1080 - Medium shoc: OpenCL - Texture Read Bandwidth unvanquished: 3840 x 2160 - High unvanquished: 3840 x 2160 - Ultra unvanquished: 1920 x 1080 - Ultra xonotic: 3840 x 2160 - Low viennacl: CPU BLAS - dGEMM-TN viennacl: CPU BLAS - dGEMM-NT ncnn: Vulkan GPU - resnet50 xonotic: 3840 x 2160 - High etlegacy: 1920 x 1200 viennacl: CPU BLAS - dGEMM-TT etlegacy: 2560 x 1440 xonotic: 3840 x 2160 - Ultimate mandelgpu: GPU viennacl: OpenCL BLAS - sCOPY etlegacy: 1920 x 1080 fahbench: viennacl: OpenCL BLAS - sDOT vkresample: 2x - Single xonotic: 3840 x 2160 - Ultra viennacl: OpenCL BLAS - dGEMV-T viennacl: OpenCL BLAS - sAXPY ncnn: Vulkan GPU - vgg16 cl-mem: Copy viennacl: OpenCL BLAS - dDOT cl-mem: Read luxcorerender: Rainbow Colors and Prism - GPU viennacl: OpenCL BLAS - dCOPY viennacl: OpenCL BLAS - dGEMV-N cl-mem: Write clpeak: Global Memory Bandwidth viennacl: OpenCL BLAS - dAXPY ncnn: Vulkan GPU - alexnet warsow: 3840 x 2160 warsow: 1920 x 1080 lczero: OpenCL warsow: 1920 x 1200 v-ray: NVIDIA RTX GPU warsow: 2560 x 1440 v-ray: NVIDIA CUDA GPU neatbench: GPU blender: BMW27 - CUDA ncnn: Vulkan GPU - squeezenet_ssd ncnn: Vulkan GPU - googlenet viennacl: CPU BLAS - dGEMV-N unvanquished: 2560 x 1440 - Ultra etlegacy: 3840 x 2160 waifu2x-ncnn: 2x - 3 - No RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 RTX 4090 27.1252 26.3342 25.4781 8102.38 99.6 66.1 384.84 6157.361 34.9 23.3 391.209 140 17725.14 16885.79 42.4 35227.21 602 846100 1138500 22599300000 40753.9 2887000000 13708.58 653.61 48.11 599 20672.19 20769.27 41184.89 20949.64 27806.74 117.817 20927.99 658.16 44.1932 604 71446900000 6.259 3.17 559.05 8944.746 4134.488 396.74 601 9.18 686.231258 91.46 9169.636 93.91 9414.686 11.22 3.843 93.2 9343.817 9419.399 93.95 91.13 22.86 76.1 10151.478 634.47 45831 22.76 56.19 10.45 679.67 10874.708 17.96 3.26 17.4 1.47 20.961 2.08 1.98 1.94 28.949 6.94 2.54 7.02 514.93 5366.18 429.082 1.67 561.32 5849.675 482 4.28 53.438 5.621 81.5 1.06 464.9 473.6 3.313 6274.519 602.09 463 2100.51 492.5 469 490.1 0.12792 496.1 2246.17 463.6 456.8 470.8 655.142588 93.8 88.4 3.55 556.9297001 645.3 91.9 640.2 374.4286255 481322759.6 368 652.3 353.8915 375 9.272 499.2632615 377 503 4.17 364.1 660 794.6 32.33 606 239 743.1 813.42 722 1.91 960.8 984.3 22739 984.8 2856 984.7 2119 11.04 15.96 5.85 68.1 461.7 648.4 27.1252 26.3282 25.4840 8336.71 98.2 65.5 376.77 6028.366 34.7 23.2 391.573 137 18044.25 16880.01 41.6 35136.60 601 825633 1140100 22678500000 40238.8 2884666667 13657.36 649.53 48.18 651.17 601 20594.78 20689.72 40926.17 20717.31 27457.90 118.517 20826.66 657.71 44.5453 605 71411366667 6.258 2.24 544.98 8719.750 4070.881 390.64 600 9.20 685.109656 91.94 9217.228 94.28 9451.903 11.32 3.694 93.33 9356.710 9429.536 94.05 91.29 11.57 22.87 75.5 10300.044 643.75 44252 22.69 56.35 10.41 689.81 11036.994 11.20 18.00 3.25 17.43 1.470 20.951 2.08 1.97 1.8 29.064 6.91 2.53 6.72 515.78 5374.983 430.347 1.67 563.05 5867.721 479.0 4.22 53.500 5.715 81.9 1.03 478.3 488.5 3.362 6206.035 595.52 469.8 2101.08 488.5 467.6 485.6 0.12912 487.3 2240.09 471.0 461.2 469.5 659.9059394 92.5 84.4 3.55 567.7214300 659.1 90.7 647.6 369.9242194 475794831.7 364 654.5 353.3272 371 9.292 492.4660277 377 501 4.18 364.2 657 794.3 32.60 605 240 744.7 813.45 722 1.92 980.7 975.5 22711 980.3 2856 985.7 2112 11.02 17.31 6.07 67.1 452.5 638.7 27.0904 26.3202 25.4505 8098.46 98.3 65.3 381.69 6107.065 34.3 23.2 392.03 137 18742.7 17012.23 42.2 35225.4 605 816300 1141500 22813100000 40566.2 2892900000 13709.41 658.5 48.09 658.51 602 20672.21 20924.98 41496.68 20953.72 27806.55 119.073 20960.25 658.16 44.5358 606 71436200000 6.256 2.24 565.6 9049.543 4081.298 391.63 604 9.01 681.126108 92.03 9226.854 92.13 9236.629 11.22 3.627 93.57 9380.932 9402.63 93.79 90.36 11.61 22.78 75.4 10497.894 656.12 44147 22.64 56.34 10.43 696.47 11143.486 11.22 17.96 3.42 17.98 1.485 20.943 2.08 1.97 1.8 28.891 6.92 2.53 7.43 521.09 5430.386 430.279 1.7 560.03 5836.216 440 4.23 53.6 5.673 82.6 1.05 474.6 484 3.318 6247.735 599.52 471.1 2101.82 481.2 475.7 483 0.12779 491.9 2245.23 472.9 469.7 474.5 645.8515908 92.8 86.8 3.56 566.1228693 657.8 90.8 643.7 367.1052632 472928214.8 364 659.8 354.4623 371 9.291 490.2778714 377 500 4.17 362.8 637 795.2 32.25 608 237 742.8 813.47 723 1.9 978.6 984.8 23029 984.9 2829 985.5 2115 11.09 19.73 5.75 67.5 466.6 635.5 3.3579 3.3878 3.3663 27192.4 311 206 1137.66 18202.541 87.6 58.0 953.251 332 41856.42 39651.75 96.9 81211.79 1350 1827614 2547700 49722966667 88132.7 6300233333 29775.38 1409.19 22.22 1410.44 1297 44563.49 44768.16 88463.01 44666.11 59114.75 55.371 44774.48 1408.89 94.6195 1283 151433333333 2.953 1.5 1127.10 18033.644 8296.768 796.14 1160 17.39 1300.965825 174.52 17496.043 174.29 17473.094 6.05 2.055 173.98 17441.984 17448.869 174.04 49.31 21.38 12.59 133 17621.059 1101.32 76599 13.30 33.55 17.47 1111.66 17786.529 18.31 11.06 2.11 11.13 0.9347 33.044 1.32 1.27 1.25 18.731 4.48 1.64 4.80 794.76 8282.332 646.062 1.13 800.73 8344.584 620.3 3.05 74.728 4.166 111 0.78 628.5 638.9 2.506 8320.030 798.37 615.4 2787.07 637.5 618.8 636.9 0.16788 640.1 2939.59 607.3 597.3 611.3 830.0198846 118 107 2.81 701.8573748 811.4 114 799.7 457.4973315 587587462.0 452 794.0 425.8766 447 7.760 586.6807687 441 575 3.68 411.7 720 887.9 35.74 659 221 804.3 870.62 771 1.87 985.4 971.1 973.0 980.1 4090 10.25 3.43 2.69 93.8 608.2 219.3 OpenBenchmarking.org
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Bus Speed Readback OpenBenchmarking.org GB/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Bus Speed Readback NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 RTX 4090 6 12 18 24 30 SE +/- 0.0002, N = 3 SE +/- 0.0000, N = 3 27.1252 27.1252 27.0904 3.3579 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Bus Speed Download OpenBenchmarking.org GB/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Bus Speed Download RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 RTX 4090 6 12 18 24 30 SE +/- 0.0129, N = 3 SE +/- 0.0000, N = 3 26.3342 26.3282 26.3202 3.3878 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Triad OpenBenchmarking.org GB/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Triad NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 RTX 4090 6 12 18 24 30 SE +/- 0.0039, N = 3 SE +/- 0.0005, N = 3 25.4840 25.4781 25.4505 3.3663 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: GEMM SGEMM_N OpenBenchmarking.org GFLOPS, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: GEMM SGEMM_N RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 6K 12K 18K 24K 30K SE +/- 2.47, N = 3 SE +/- 67.62, N = 3 27192.40 8336.71 8102.38 8098.46 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
ViennaCL Test: CPU BLAS - sAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - sAXPY RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 70 140 210 280 350 SE +/- 1.45, N = 3 SE +/- 0.68, N = 3 311.0 99.6 98.3 98.2 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - sCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - sCOPY RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 50 100 150 200 250 SE +/- 0.88, N = 3 SE +/- 0.09, N = 3 206.0 66.1 65.5 65.3 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ParaView Test: Wavelet Volume - Resolution: 3840 x 2160 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Wavelet Volume - Resolution: 3840 x 2160 RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 9.85, N = 3 SE +/- 4.06, N = 5 1137.66 384.84 381.69 376.77
ParaView Test: Wavelet Volume - Resolution: 3840 x 2160 OpenBenchmarking.org MiVoxels / Sec, More Is Better ParaView 5.9 Test: Wavelet Volume - Resolution: 3840 x 2160 RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 4K 8K 12K 16K 20K SE +/- 157.61, N = 3 SE +/- 65.03, N = 5 18202.54 6157.36 6107.07 6028.37
ViennaCL Test: CPU BLAS - dAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dAXPY RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 20 40 60 80 100 SE +/- 0.06, N = 3 SE +/- 0.00, N = 3 87.6 34.9 34.7 34.3 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dCOPY RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 13 26 39 52 65 SE +/- 0.06, N = 3 SE +/- 0.00, N = 3 58.0 23.3 23.2 23.2 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Reduction OpenBenchmarking.org GB/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Reduction RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 200 400 600 800 1000 SE +/- 0.25, N = 3 SE +/- 0.19, N = 3 953.25 392.03 391.57 391.21 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
ViennaCL Test: CPU BLAS - sDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - sDOT RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 70 140 210 280 350 SE +/- 2.67, N = 3 SE +/- 0.67, N = 3 332 140 137 137 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
clpeak OpenCL Test: Integer Compute INT OpenBenchmarking.org GIOPS, More Is Better clpeak OpenCL Test: Integer Compute INT RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 9K 18K 27K 36K 45K SE +/- 370.89, N = 3 SE +/- 115.05, N = 3 41856.42 18742.70 18044.25 17725.14 1. (CXX) g++ options: -O3 -rdynamic -lOpenCL
vkpeak int16-vec4 OpenBenchmarking.org GIOPS, More Is Better vkpeak 20210424 int16-vec4 RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 8K 16K 24K 32K 40K SE +/- 18.88, N = 3 SE +/- 2.98, N = 3 39651.75 17012.23 16885.79 16880.01
ViennaCL Test: CPU BLAS - dDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dDOT RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.12, N = 3 SE +/- 0.42, N = 3 96.9 42.4 42.2 41.6 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
clpeak OpenCL Test: Single-Precision Float OpenBenchmarking.org GFLOPS, More Is Better clpeak OpenCL Test: Single-Precision Float RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 20K 40K 60K 80K 100K SE +/- 435.51, N = 3 SE +/- 88.80, N = 3 81211.79 35227.21 35225.40 35136.60 1. (CXX) g++ options: -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dGEMM-TT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-TT RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 300 600 900 1200 1500 1350 605 602 601 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
Hashcat Benchmark: TrueCrypt RIPEMD160 + XTS OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: TrueCrypt RIPEMD160 + XTS RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 400K 800K 1200K 1600K 2000K SE +/- 16650.59, N = 7 SE +/- 4520.08, N = 3 1827614 846100 825633 816300
Hashcat Benchmark: 7-Zip OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: 7-Zip RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 500K 1000K 1500K 2000K 2500K SE +/- 4152.51, N = 3 SE +/- 4014.97, N = 3 2547700 1141500 1140100 1138500
Hashcat Benchmark: SHA1 OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: SHA1 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 11000M 22000M 33000M 44000M 55000M SE +/- 15117135.24, N = 3 SE +/- 32122473.96, N = 3 49722966667 22813100000 22678500000 22599300000
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Max SP Flops OpenBenchmarking.org GFLOPS, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Max SP Flops RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 20K 40K 60K 80K 100K SE +/- 781.51, N = 3 SE +/- 297.72, N = 3 88132.7 40753.9 40566.2 40238.8 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
Hashcat Benchmark: SHA-512 OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: SHA-512 RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 1300M 2600M 3900M 5200M 6500M SE +/- 1942792.95, N = 3 SE +/- 2179704.36, N = 3 6300233333 2892900000 2887000000 2884666667
vkpeak int16-scalar OpenBenchmarking.org GIOPS, More Is Better vkpeak 20210424 int16-scalar RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 6K 12K 18K 24K 30K SE +/- 36.89, N = 3 SE +/- 29.87, N = 3 29775.38 13709.41 13708.58 13657.36
vkpeak fp64-scalar OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp64-scalar RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 300 600 900 1200 1500 SE +/- 0.03, N = 3 SE +/- 0.83, N = 3 1409.19 658.50 653.61 649.53
Blender Blend File: Pabellon Barcelona - Compute: CUDA OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.0 Blend File: Pabellon Barcelona - Compute: CUDA RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 11 22 33 44 55 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 22.22 48.09 48.11 48.18
vkpeak fp64-vec4 OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp64-vec4 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 300 600 900 1200 1500 SE +/- 1.24, N = 3 SE +/- 1.42, N = 3 1410.44 658.51 651.17
ViennaCL Test: OpenCL BLAS - dGEMM-TN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-TN RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 300 600 900 1200 1500 SE +/- 3.33, N = 3 SE +/- 1.33, N = 3 1297 602 601 599 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
vkpeak int32-vec4 OpenBenchmarking.org GIOPS, More Is Better vkpeak 20210424 int32-vec4 RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 10K 20K 30K 40K 50K SE +/- 2.83, N = 3 SE +/- 44.80, N = 3 44563.49 20672.21 20672.19 20594.78
vkpeak int32-scalar OpenBenchmarking.org GIOPS, More Is Better vkpeak 20210424 int32-scalar RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 10K 20K 30K 40K 50K SE +/- 1.91, N = 3 SE +/- 44.97, N = 3 44768.16 20924.98 20769.27 20689.72
vkpeak fp16-vec4 OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp16-vec4 RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 20K 40K 60K 80K 100K SE +/- 3.14, N = 3 SE +/- 52.39, N = 3 88463.01 41496.68 41184.89 40926.17
vkpeak fp16-scalar OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp16-scalar RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 10K 20K 30K 40K 50K SE +/- 58.62, N = 3 SE +/- 45.34, N = 3 44666.11 20953.72 20949.64 20717.31
vkpeak fp32-vec4 OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp32-vec4 RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 13K 26K 39K 52K 65K SE +/- 88.56, N = 3 SE +/- 69.31, N = 3 59114.75 27806.74 27806.55 27457.90
VkResample Upscale: 2x - Precision: Double OpenBenchmarking.org ms, Fewer Is Better VkResample 1.0 Upscale: 2x - Precision: Double RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 30 60 90 120 150 SE +/- 0.01, N = 3 SE +/- 0.08, N = 3 55.37 117.82 118.52 119.07 1. (CXX) g++ options: -O3
vkpeak fp32-scalar OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp32-scalar RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 10K 20K 30K 40K 50K SE +/- 61.65, N = 3 SE +/- 65.72, N = 3 44774.48 20960.25 20927.99 20826.66
clpeak OpenCL Test: Double-Precision Double OpenBenchmarking.org GFLOPS, More Is Better clpeak OpenCL Test: Double-Precision Double RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 300 600 900 1200 1500 SE +/- 0.23, N = 3 SE +/- 0.03, N = 3 1408.89 658.16 658.16 657.71 1. (CXX) g++ options: -O3 -rdynamic -lOpenCL
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: MD5 Hash OpenBenchmarking.org GHash/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: MD5 Hash RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 20 40 60 80 100 SE +/- 1.03, N = 15 SE +/- 0.00, N = 3 94.62 44.55 44.54 44.19 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
ViennaCL Test: OpenCL BLAS - dGEMM-NT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-NT RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 300 600 900 1200 1500 SE +/- 3.33, N = 3 SE +/- 1.67, N = 3 1283 606 605 604 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
Hashcat Benchmark: MD5 OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: MD5 RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 30000M 60000M 90000M 120000M 150000M SE +/- 166666666.67, N = 3 SE +/- 78065236.25, N = 3 151433333333 71446900000 71436200000 71411366667
FinanceBench Benchmark: Black-Scholes OpenCL OpenBenchmarking.org ms, Fewer Is Better FinanceBench 2016-07-25 Benchmark: Black-Scholes OpenCL RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 2 4 6 8 10 SE +/- 0.024, N = 3 SE +/- 0.004, N = 3 2.953 6.256 6.258 6.259 1. (CXX) g++ options: -O3 -march=native -fopenmp
NCNN Target: Vulkan GPU-v3-v3 - Model: mobilenet-v3 OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU-v3-v3 - Model: mobilenet-v3 RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 0.7133 1.4266 2.1399 2.8532 3.5665 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 1.50 2.24 2.24 3.17 MIN: 1.47 / MAX: 2.71 MIN: 2.21 / MAX: 3.37 MIN: 2.21 / MAX: 4.61 MIN: 2.21 / MAX: 20.43 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
ParaView Test: Wavelet Volume - Resolution: 2560 x 1440 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Wavelet Volume - Resolution: 2560 x 1440 RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 6.66, N = 3 SE +/- 3.11, N = 3 1127.10 565.60 559.05 544.98
ParaView Test: Wavelet Volume - Resolution: 2560 x 1440 OpenBenchmarking.org MiVoxels / Sec, More Is Better ParaView 5.9 Test: Wavelet Volume - Resolution: 2560 x 1440 RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 4K 8K 12K 16K 20K SE +/- 106.54, N = 3 SE +/- 49.80, N = 3 18033.64 9049.54 8944.75 8719.75
ParaView Test: Wavelet Contour - Resolution: 3840 x 2160 OpenBenchmarking.org MiPolys / Sec, More Is Better ParaView 5.9 Test: Wavelet Contour - Resolution: 3840 x 2160 RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 2K 4K 6K 8K 10K SE +/- 17.54, N = 3 SE +/- 31.15, N = 3 8296.77 4134.49 4081.30 4070.88
ParaView Test: Wavelet Contour - Resolution: 3840 x 2160 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Wavelet Contour - Resolution: 3840 x 2160 RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 1.68, N = 3 SE +/- 2.99, N = 3 796.14 396.74 391.63 390.64
ViennaCL Test: OpenCL BLAS - dGEMM-NN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-NN RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 200 400 600 800 1000 1160 604 601 600 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
LuxCoreRender Scene: Danish Mood - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.5 Scene: Danish Mood - Acceleration: GPU RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 4 8 12 16 20 SE +/- 0.20, N = 3 SE +/- 0.08, N = 3 17.39 9.20 9.18 9.01 MIN: 0.07 / MAX: 21.12 MIN: 3.03 / MAX: 10.91 MIN: 3.53 / MAX: 10.86 MIN: 3.3 / MAX: 10.77
OctaneBench Total Score OpenBenchmarking.org Score, More Is Better OctaneBench 2020.1 Total Score RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 300 600 900 1200 1500 1300.97 686.23 685.11 681.13
ParaView Test: Many Spheres - Resolution: 3840 x 2160 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Many Spheres - Resolution: 3840 x 2160 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 40 80 120 160 200 SE +/- 0.24, N = 3 SE +/- 0.06, N = 3 174.52 92.03 91.94 91.46
ParaView Test: Many Spheres - Resolution: 3840 x 2160 OpenBenchmarking.org MiPolys / Sec, More Is Better ParaView 5.9 Test: Many Spheres - Resolution: 3840 x 2160 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 4K 8K 12K 16K 20K SE +/- 24.18, N = 3 SE +/- 6.01, N = 3 17496.04 9226.85 9217.23 9169.64
ParaView Test: Many Spheres - Resolution: 1920 x 1200 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Many Spheres - Resolution: 1920 x 1200 RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 40 80 120 160 200 SE +/- 0.25, N = 3 SE +/- 0.06, N = 3 174.29 94.28 93.91 92.13
ParaView Test: Many Spheres - Resolution: 1920 x 1200 OpenBenchmarking.org MiPolys / Sec, More Is Better ParaView 5.9 Test: Many Spheres - Resolution: 1920 x 1200 RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 4K 8K 12K 16K 20K SE +/- 25.11, N = 3 SE +/- 5.73, N = 3 17473.09 9451.90 9414.69 9236.63
Blender Blend File: Fishy Cat - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.0 Blend File: Fishy Cat - Compute: NVIDIA OptiX RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 3 6 9 12 15 SE +/- 0.05, N = 13 SE +/- 0.09, N = 3 6.05 11.22 11.22 11.32
Rodinia Test: OpenCL Particle Filter OpenBenchmarking.org Seconds, Fewer Is Better Rodinia 3.1 Test: OpenCL Particle Filter RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 0.8647 1.7294 2.5941 3.4588 4.3235 SE +/- 0.050, N = 3 SE +/- 0.013, N = 3 2.055 3.627 3.694 3.843 1. (CXX) g++ options: -O2 -lOpenCL
ParaView Test: Many Spheres - Resolution: 2560 x 1440 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Many Spheres - Resolution: 2560 x 1440 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 40 80 120 160 200 SE +/- 0.02, N = 3 SE +/- 0.08, N = 3 173.98 93.57 93.33 93.20
ParaView Test: Many Spheres - Resolution: 2560 x 1440 OpenBenchmarking.org MiPolys / Sec, More Is Better ParaView 5.9 Test: Many Spheres - Resolution: 2560 x 1440 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 4K 8K 12K 16K 20K SE +/- 2.22, N = 3 SE +/- 7.77, N = 3 17441.98 9380.93 9356.71 9343.82
ParaView Test: Many Spheres - Resolution: 1920 x 1080 OpenBenchmarking.org MiPolys / Sec, More Is Better ParaView 5.9 Test: Many Spheres - Resolution: 1920 x 1080 RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 4K 8K 12K 16K 20K SE +/- 9.92, N = 3 SE +/- 4.72, N = 3 17448.87 9429.54 9419.40 9402.63
ParaView Test: Many Spheres - Resolution: 1920 x 1080 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Many Spheres - Resolution: 1920 x 1080 RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 40 80 120 160 200 SE +/- 0.10, N = 3 SE +/- 0.05, N = 3 174.04 94.05 93.95 93.79
Blender Blend File: Barbershop - Compute: CUDA OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.0 Blend File: Barbershop - Compute: CUDA RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.17, N = 3 SE +/- 0.17, N = 3 49.31 90.36 91.13 91.29
LuxCoreRender Scene: DLSC - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.5 Scene: DLSC - Acceleration: GPU RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 5 10 15 20 25 SE +/- 0.02, N = 3 SE +/- 0.03, N = 3 21.38 11.61 11.57 MIN: 18.47 / MAX: 21.85 MIN: 11.41 / MAX: 11.72 MIN: 11.39 / MAX: 11.73
Blender Blend File: Fishy Cat - Compute: CUDA OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.0 Blend File: Fishy Cat - Compute: CUDA RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 5 10 15 20 25 SE +/- 0.00, N = 3 SE +/- 0.04, N = 3 12.59 22.78 22.86 22.87
ViennaCL Test: CPU BLAS - dGEMV-T OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMV-T RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 30 60 90 120 150 SE +/- 0.36, N = 3 133.0 76.1 75.5 75.4 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ParaView Test: Wavelet Volume - Resolution: 1920 x 1200 OpenBenchmarking.org MiVoxels / Sec, More Is Better ParaView 5.9 Test: Wavelet Volume - Resolution: 1920 x 1200 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 4K 8K 12K 16K 20K SE +/- 64.74, N = 3 SE +/- 74.83, N = 3 17621.06 10497.89 10300.04 10151.48
ParaView Test: Wavelet Volume - Resolution: 1920 x 1200 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Wavelet Volume - Resolution: 1920 x 1200 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 200 400 600 800 1000 SE +/- 4.05, N = 3 SE +/- 4.68, N = 3 1101.32 656.12 643.75 634.47
VkFFT OpenBenchmarking.org Benchmark Score, More Is Better VkFFT 1.1.1 RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 16K 32K 48K 64K 80K SE +/- 489.60, N = 3 SE +/- 384.18, N = 9 76599 45831 44252 44147 1. (CXX) g++ options: -O3
Blender Blend File: Classroom - Compute: CUDA OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.0 Blend File: Classroom - Compute: CUDA RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 5 10 15 20 25 SE +/- 0.01, N = 3 SE +/- 0.03, N = 3 13.30 22.64 22.69 22.76
Blender Blend File: Barbershop - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.0 Blend File: Barbershop - Compute: NVIDIA OptiX RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 13 26 39 52 65 SE +/- 0.02, N = 3 SE +/- 0.06, N = 3 33.55 56.19 56.34 56.35
LuxCoreRender Scene: Orange Juice - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.5 Scene: Orange Juice - Acceleration: GPU RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.04, N = 3 SE +/- 0.10, N = 3 17.47 10.45 10.43 10.41 MIN: 0.39 / MAX: 25.03 MIN: 8.53 / MAX: 13.8 MIN: 8.54 / MAX: 13.75 MIN: 8.47 / MAX: 13.76
ParaView Test: Wavelet Volume - Resolution: 1920 x 1080 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Wavelet Volume - Resolution: 1920 x 1080 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 200 400 600 800 1000 SE +/- 2.67, N = 3 SE +/- 5.50, N = 3 1111.66 696.47 689.81 679.67
ParaView Test: Wavelet Volume - Resolution: 1920 x 1080 OpenBenchmarking.org MiVoxels / Sec, More Is Better ParaView 5.9 Test: Wavelet Volume - Resolution: 1920 x 1080 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 4K 8K 12K 16K 20K SE +/- 42.71, N = 3 SE +/- 88.00, N = 3 17786.53 11143.49 11036.99 10874.71
LuxCoreRender Scene: LuxCore Benchmark - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.5 Scene: LuxCore Benchmark - Acceleration: GPU RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 5 10 15 20 25 SE +/- 0.09, N = 3 SE +/- 0.01, N = 3 18.31 11.22 11.20 MIN: 7.1 / MAX: 23.24 MIN: 3.57 / MAX: 13.17 MIN: 3.54 / MAX: 13.16
Blender Blend File: Pabellon Barcelona - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.0 Blend File: Pabellon Barcelona - Compute: NVIDIA OptiX RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.01, N = 3 SE +/- 0.02, N = 3 11.06 17.96 17.96 18.00
NCNN Target: Vulkan GPU - Model: efficientnet-b0 OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: efficientnet-b0 RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 0.7695 1.539 2.3085 3.078 3.8475 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 2.11 3.25 3.26 3.42 MIN: 2.09 / MAX: 2.75 MIN: 3.22 / MAX: 4.42 MIN: 3.23 / MAX: 3.42 MIN: 3.24 / MAX: 4.26 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
Blender Blend File: Classroom - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.0 Blend File: Classroom - Compute: NVIDIA OptiX RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 4 8 12 16 20 SE +/- 0.02, N = 3 SE +/- 0.02, N = 3 11.13 17.40 17.43 17.98
ArrayFire Test: Conjugate Gradient OpenCL OpenBenchmarking.org ms, Fewer Is Better ArrayFire 3.7 Test: Conjugate Gradient OpenCL RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 0.3341 0.6682 1.0023 1.3364 1.6705 SE +/- 0.0019, N = 3 SE +/- 0.0055, N = 3 0.9347 1.4700 1.4700 1.4850 1. (CXX) g++ options: -rdynamic
IndigoBench Acceleration: OpenCL GPU - Scene: Bedroom OpenBenchmarking.org M samples/s, More Is Better IndigoBench 4.4 Acceleration: OpenCL GPU - Scene: Bedroom RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 8 16 24 32 40 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 33.04 20.96 20.95 20.94
NCNN Target: Vulkan GPU - Model: mnasnet OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: mnasnet RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 0.468 0.936 1.404 1.872 2.34 SE +/- 0.00, N = 3 SE +/- 0.01, N = 3 1.32 2.08 2.08 2.08 MIN: 1.31 / MAX: 2 MIN: 2.06 / MAX: 2.29 MIN: 2.05 / MAX: 3.23 MIN: 2.06 / MAX: 2.51 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU-v2-v2 - Model: mobilenet-v2 OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU-v2-v2 - Model: mobilenet-v2 RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 0.4455 0.891 1.3365 1.782 2.2275 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 1.27 1.97 1.97 1.98 MIN: 1.25 / MAX: 2.23 MIN: 1.94 / MAX: 3.15 MIN: 1.94 / MAX: 2.26 MIN: 1.94 / MAX: 6.06 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: shufflenet-v2 OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: shufflenet-v2 RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 0.4365 0.873 1.3095 1.746 2.1825 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 1.25 1.80 1.80 1.94 MIN: 1.23 / MAX: 2.16 MIN: 1.76 / MAX: 3 MIN: 1.77 / MAX: 2.69 MIN: 1.78 / MAX: 2.28 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
RealSR-NCNN Scale: 4x - TAA: Yes OpenBenchmarking.org Seconds, Fewer Is Better RealSR-NCNN 20200818 Scale: 4x - TAA: Yes RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 7 14 21 28 35 SE +/- 0.00, N = 3 SE +/- 0.06, N = 3 18.73 28.89 28.95 29.06
Blender Blend File: BMW27 - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.0 Blend File: BMW27 - Compute: NVIDIA OptiX RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 2 4 6 8 10 SE +/- 0.04, N = 15 SE +/- 0.01, N = 3 4.48 6.91 6.92 6.94
NCNN Target: Vulkan GPU - Model: regnety_400m OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: regnety_400m RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 0.5715 1.143 1.7145 2.286 2.8575 SE +/- 0.00, N = 3 SE +/- 0.01, N = 3 1.64 2.53 2.53 2.54 MIN: 1.61 / MAX: 2.15 MIN: 2.49 / MAX: 3.64 MIN: 2.51 / MAX: 2.76 MIN: 2.5 / MAX: 3.7 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: yolov4-tiny OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: yolov4-tiny RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 2 4 6 8 10 SE +/- 0.00, N = 3 SE +/- 0.04, N = 3 4.80 6.72 7.02 7.43 MIN: 4.73 / MAX: 5.37 MIN: 6.3 / MAX: 19.53 MIN: 6.34 / MAX: 30.81 MIN: 6.36 / MAX: 36.74 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
ParaView Test: Wavelet Contour - Resolution: 2560 x 1440 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Wavelet Contour - Resolution: 2560 x 1440 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 200 400 600 800 1000 SE +/- 1.13, N = 3 SE +/- 1.60, N = 3 794.76 521.09 515.78 514.93
ParaView Test: Wavelet Contour - Resolution: 2560 x 1440 OpenBenchmarking.org MiPolys / Sec, More Is Better ParaView 5.9 Test: Wavelet Contour - Resolution: 2560 x 1440 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 2K 4K 6K 8K 10K SE +/- 11.72, N = 3 SE +/- 16.63, N = 3 8282.33 5430.39 5374.98 5366.18
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: S3D OpenBenchmarking.org GFLOPS, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: S3D RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 140 280 420 560 700 SE +/- 0.34, N = 3 SE +/- 0.21, N = 3 646.06 430.35 430.28 429.08 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
NCNN Target: Vulkan GPU - Model: resnet18 OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: resnet18 RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 0.3825 0.765 1.1475 1.53 1.9125 SE +/- 0.01, N = 3 SE +/- 0.00, N = 3 1.13 1.67 1.67 1.70 MIN: 1.11 / MAX: 1.43 MIN: 1.64 / MAX: 1.9 MIN: 1.64 / MAX: 4.75 MIN: 1.65 / MAX: 9.46 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
ParaView Test: Wavelet Contour - Resolution: 1920 x 1200 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Wavelet Contour - Resolution: 1920 x 1200 RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 200 400 600 800 1000 SE +/- 0.35, N = 3 SE +/- 5.24, N = 3 800.73 563.05 561.32 560.03
ParaView Test: Wavelet Contour - Resolution: 1920 x 1200 OpenBenchmarking.org MiPolys / Sec, More Is Better ParaView 5.9 Test: Wavelet Contour - Resolution: 1920 x 1200 RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 2K 4K 6K 8K 10K SE +/- 3.68, N = 3 SE +/- 54.66, N = 3 8344.58 5867.72 5849.68 5836.22
Unvanquished Resolution: 1920 x 1080 - Effects Quality: High OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 1920 x 1080 - Effects Quality: High RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 130 260 390 520 650 SE +/- 1.32, N = 3 SE +/- 2.19, N = 3 620.3 482.0 479.0 440.0
NCNN Target: Vulkan GPU - Model: mobilenet OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: mobilenet RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 0.963 1.926 2.889 3.852 4.815 SE +/- 0.00, N = 3 SE +/- 0.01, N = 3 3.05 4.22 4.23 4.28 MIN: 3.02 / MAX: 4.05 MIN: 4.13 / MAX: 4.48 MIN: 4.16 / MAX: 4.41 MIN: 4.15 / MAX: 15.44 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
IndigoBench Acceleration: OpenCL GPU - Scene: Supercar OpenBenchmarking.org M samples/s, More Is Better IndigoBench 4.4 Acceleration: OpenCL GPU - Scene: Supercar RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 20 40 60 80 100 SE +/- 0.02, N = 3 SE +/- 0.03, N = 3 74.73 53.60 53.50 53.44
RealSR-NCNN Scale: 4x - TAA: No OpenBenchmarking.org Seconds, Fewer Is Better RealSR-NCNN 20200818 Scale: 4x - TAA: No RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 1.2859 2.5718 3.8577 5.1436 6.4295 SE +/- 0.005, N = 3 SE +/- 0.022, N = 3 4.166 5.621 5.673 5.715
ViennaCL Test: CPU BLAS - dGEMM-NN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-NN RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 20 40 60 80 100 SE +/- 0.18, N = 3 111.0 82.6 81.9 81.5 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
NCNN Target: Vulkan GPU - Model: blazeface OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: blazeface RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 0.2385 0.477 0.7155 0.954 1.1925 SE +/- 0.01, N = 3 SE +/- 0.00, N = 3 0.78 1.03 1.05 1.06 MIN: 0.76 / MAX: 1.68 MIN: 1 / MAX: 2.07 MIN: 1.01 / MAX: 2.18 MIN: 1.02 / MAX: 2.12 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
Unvanquished Resolution: 1920 x 1200 - Effects Quality: High OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 1920 x 1200 - Effects Quality: High RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 140 280 420 560 700 SE +/- 3.34, N = 3 SE +/- 4.62, N = 3 628.5 478.3 474.6 464.9
Unvanquished Resolution: 1920 x 1200 - Effects Quality: Medium OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 1920 x 1200 - Effects Quality: Medium RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 140 280 420 560 700 SE +/- 2.89, N = 3 SE +/- 1.98, N = 3 638.9 488.5 484.0 473.6
Waifu2x-NCNN Vulkan Scale: 2x - Denoise: 3 - TAA: Yes OpenBenchmarking.org Seconds, Fewer Is Better Waifu2x-NCNN Vulkan 20200818 Scale: 2x - Denoise: 3 - TAA: Yes RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 0.7565 1.513 2.2695 3.026 3.7825 SE +/- 0.006, N = 3 SE +/- 0.004, N = 3 2.506 3.313 3.318 3.362
ParaView Test: Wavelet Contour - Resolution: 1920 x 1080 OpenBenchmarking.org MiPolys / Sec, More Is Better ParaView 5.9 Test: Wavelet Contour - Resolution: 1920 x 1080 RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 2K 4K 6K 8K 10K SE +/- 16.65, N = 3 SE +/- 54.90, N = 3 8320.03 6274.52 6247.74 6206.04
ParaView Test: Wavelet Contour - Resolution: 1920 x 1080 OpenBenchmarking.org Frames / Sec, More Is Better ParaView 5.9 Test: Wavelet Contour - Resolution: 1920 x 1080 RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 1.60, N = 3 SE +/- 5.27, N = 3 798.37 602.09 599.52 595.52
Unvanquished Resolution: 1920 x 1200 - Effects Quality: Ultra OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 1920 x 1200 - Effects Quality: Ultra RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 130 260 390 520 650 SE +/- 0.79, N = 3 SE +/- 5.55, N = 3 615.4 471.1 469.8 463.0
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: FFT SP OpenBenchmarking.org GFLOPS, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: FFT SP RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 600 1200 1800 2400 3000 SE +/- 1.73, N = 3 SE +/- 0.42, N = 3 2787.07 2101.82 2101.08 2100.51 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
Unvanquished Resolution: 2560 x 1440 - Effects Quality: Medium OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 2560 x 1440 - Effects Quality: Medium RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 140 280 420 560 700 SE +/- 0.49, N = 3 SE +/- 3.75, N = 3 637.5 492.5 488.5 481.2
Unvanquished Resolution: 2560 x 1440 - Effects Quality: High OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 2560 x 1440 - Effects Quality: High RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 130 260 390 520 650 SE +/- 1.30, N = 3 SE +/- 2.97, N = 3 618.8 475.7 469.0 467.6
Unvanquished Resolution: 3840 x 2160 - Effects Quality: Medium OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 3840 x 2160 - Effects Quality: Medium RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 140 280 420 560 700 SE +/- 6.63, N = 3 SE +/- 2.17, N = 3 636.9 490.1 485.6 483.0
NAMD CUDA ATPase Simulation - 327,506 Atoms OpenBenchmarking.org days/ns, Fewer Is Better NAMD CUDA 2.14 ATPase Simulation - 327,506 Atoms NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 RTX 4090 0.0378 0.0756 0.1134 0.1512 0.189 SE +/- 0.00047, N = 3 SE +/- 0.00009, N = 3 0.12779 0.12792 0.12912 0.16788
Unvanquished Resolution: 1920 x 1080 - Effects Quality: Medium OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 1920 x 1080 - Effects Quality: Medium RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 140 280 420 560 700 SE +/- 5.77, N = 3 SE +/- 3.91, N = 3 640.1 496.1 491.9 487.3
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Texture Read Bandwidth OpenBenchmarking.org GB/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Texture Read Bandwidth RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 600 1200 1800 2400 3000 SE +/- 0.34, N = 3 SE +/- 3.07, N = 3 2939.59 2246.17 2245.23 2240.09 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
Unvanquished Resolution: 3840 x 2160 - Effects Quality: High OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 3840 x 2160 - Effects Quality: High RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 130 260 390 520 650 SE +/- 1.30, N = 3 SE +/- 3.53, N = 3 607.3 472.9 471.0 463.6
Unvanquished Resolution: 3840 x 2160 - Effects Quality: Ultra OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 3840 x 2160 - Effects Quality: Ultra RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 130 260 390 520 650 SE +/- 1.07, N = 3 SE +/- 4.41, N = 3 597.3 469.7 461.2 456.8
Unvanquished Resolution: 1920 x 1080 - Effects Quality: Ultra OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 1920 x 1080 - Effects Quality: Ultra RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 130 260 390 520 650 SE +/- 0.91, N = 3 SE +/- 1.10, N = 3 611.3 474.5 470.8 469.5
Xonotic Resolution: 3840 x 2160 - Effects Quality: Low OpenBenchmarking.org Frames Per Second, More Is Better Xonotic 0.8.2 Resolution: 3840 x 2160 - Effects Quality: Low RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 200 400 600 800 1000 SE +/- 2.14, N = 3 SE +/- 0.55, N = 3 830.02 659.91 655.14 645.85 MIN: 217 / MAX: 1735 MIN: 109 / MAX: 1303 MIN: 103 / MAX: 1300 MIN: 117 / MAX: 1283
ViennaCL Test: CPU BLAS - dGEMM-TN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-TN RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 30 60 90 120 150 SE +/- 0.67, N = 3 SE +/- 0.32, N = 3 118.0 93.8 92.8 92.5 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dGEMM-NT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-NT RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.12, N = 3 107.0 88.4 86.8 84.4 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
NCNN Target: Vulkan GPU - Model: resnet50 OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: resnet50 RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 0.801 1.602 2.403 3.204 4.005 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 2.81 3.55 3.55 3.56 MIN: 2.79 / MAX: 3.38 MIN: 3.52 / MAX: 3.69 MIN: 3.52 / MAX: 3.75 MIN: 3.53 / MAX: 3.69 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
Xonotic Resolution: 3840 x 2160 - Effects Quality: High OpenBenchmarking.org Frames Per Second, More Is Better Xonotic 0.8.2 Resolution: 3840 x 2160 - Effects Quality: High RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 150 300 450 600 750 SE +/- 3.30, N = 3 SE +/- 1.97, N = 3 701.86 567.72 566.12 556.93 MIN: 193 / MAX: 1442 MIN: 91 / MAX: 1163 MIN: 113 / MAX: 1151 MIN: 114 / MAX: 1122
ET: Legacy Resolution: 1920 x 1200 OpenBenchmarking.org Frames Per Second, More Is Better ET: Legacy 2.78 Resolution: 1920 x 1200 RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 200 400 600 800 1000 SE +/- 0.47, N = 3 SE +/- 2.46, N = 3 811.4 659.1 657.8 645.3
ViennaCL Test: CPU BLAS - dGEMM-TT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-TT RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 30 60 90 120 150 SE +/- 0.26, N = 3 114.0 91.9 90.8 90.7 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ET: Legacy Resolution: 2560 x 1440 OpenBenchmarking.org Frames Per Second, More Is Better ET: Legacy 2.78 Resolution: 2560 x 1440 RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 200 400 600 800 1000 SE +/- 6.23, N = 3 SE +/- 7.88, N = 4 799.7 647.6 643.7 640.2
Xonotic Resolution: 3840 x 2160 - Effects Quality: Ultimate OpenBenchmarking.org Frames Per Second, More Is Better Xonotic 0.8.2 Resolution: 3840 x 2160 - Effects Quality: Ultimate RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 100 200 300 400 500 SE +/- 0.84, N = 3 SE +/- 1.53, N = 3 457.50 374.43 369.92 367.11 MIN: 72 / MAX: 1128 MIN: 65 / MAX: 751 MIN: 58 / MAX: 764 MIN: 65 / MAX: 749
MandelGPU OpenCL Device: GPU OpenBenchmarking.org Samples/sec, More Is Better MandelGPU 1.3pts1 OpenCL Device: GPU RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 130M 260M 390M 520M 650M SE +/- 430679.16, N = 3 SE +/- 1595019.30, N = 3 587587462.0 481322759.6 475794831.7 472928214.8 1. (CC) gcc options: -O3 -lm -ftree-vectorize -funroll-loops -lglut -lOpenCL -lGL
ViennaCL Test: OpenCL BLAS - sCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - sCOPY RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 100 200 300 400 500 SE +/- 0.88, N = 3 452 368 364 364 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ET: Legacy Resolution: 1920 x 1080 OpenBenchmarking.org Frames Per Second, More Is Better ET: Legacy 2.78 Resolution: 1920 x 1080 RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 200 400 600 800 1000 SE +/- 3.83, N = 3 SE +/- 3.73, N = 3 794.0 659.8 654.5 652.3
FAHBench OpenBenchmarking.org Ns Per Day, More Is Better FAHBench 2.3.2 RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 90 180 270 360 450 SE +/- 1.01, N = 3 SE +/- 0.32, N = 3 425.88 354.46 353.89 353.33
ViennaCL Test: OpenCL BLAS - sDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - sDOT RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 100 200 300 400 500 SE +/- 0.58, N = 3 447 375 371 371 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
VkResample Upscale: 2x - Precision: Single OpenBenchmarking.org ms, Fewer Is Better VkResample 1.0 Upscale: 2x - Precision: Single RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 3 6 9 12 15 SE +/- 0.002, N = 3 SE +/- 0.003, N = 3 7.760 9.272 9.291 9.292 1. (CXX) g++ options: -O3
Xonotic Resolution: 3840 x 2160 - Effects Quality: Ultra OpenBenchmarking.org Frames Per Second, More Is Better Xonotic 0.8.2 Resolution: 3840 x 2160 - Effects Quality: Ultra RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 130 260 390 520 650 SE +/- 0.80, N = 3 SE +/- 1.15, N = 3 586.68 499.26 492.47 490.28 MIN: 218 / MAX: 1144 MIN: 117 / MAX: 913 MIN: 93 / MAX: 902 MIN: 122 / MAX: 891
ViennaCL Test: OpenCL BLAS - dGEMV-T OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMV-T RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 100 200 300 400 500 SE +/- 0.33, N = 3 441 377 377 377 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - sAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - sAXPY RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 120 240 360 480 600 SE +/- 0.33, N = 3 575 503 501 500 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
NCNN Target: Vulkan GPU - Model: vgg16 OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: vgg16 RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 0.9405 1.881 2.8215 3.762 4.7025 SE +/- 0.00, N = 3 SE +/- 0.02, N = 3 3.68 4.17 4.17 4.18 MIN: 3.66 / MAX: 4.19 MIN: 4.12 / MAX: 4.86 MIN: 4.13 / MAX: 4.34 MIN: 4.12 / MAX: 11.44 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
cl-mem Benchmark: Copy OpenBenchmarking.org GB/s, More Is Better cl-mem 2017-01-13 Benchmark: Copy RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 90 180 270 360 450 SE +/- 0.17, N = 3 SE +/- 0.19, N = 3 411.7 364.2 364.1 362.8 1. (CC) gcc options: -O2 -flto -lOpenCL
ViennaCL Test: OpenCL BLAS - dDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dDOT RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 160 320 480 640 800 SE +/- 0.58, N = 3 720 660 657 637 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
cl-mem Benchmark: Read OpenBenchmarking.org GB/s, More Is Better cl-mem 2017-01-13 Benchmark: Read RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 0.09, N = 3 SE +/- 0.84, N = 3 887.9 795.2 794.6 794.3 1. (CC) gcc options: -O2 -flto -lOpenCL
LuxCoreRender Scene: Rainbow Colors and Prism - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.5 Scene: Rainbow Colors and Prism - Acceleration: GPU RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 8 16 24 32 40 SE +/- 0.22, N = 3 SE +/- 0.21, N = 3 35.74 32.60 32.33 32.25 MIN: 34.18 / MAX: 37.28 MIN: 30.12 / MAX: 34.6 MIN: 30.03 / MAX: 34.38 MIN: 29.03 / MAX: 34.55
ViennaCL Test: OpenCL BLAS - dCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dCOPY RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 140 280 420 560 700 SE +/- 0.33, N = 3 659 608 606 605 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dGEMV-N OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMV-N NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 RTX 4090 50 100 150 200 250 SE +/- 0.33, N = 3 240 239 237 221 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
cl-mem Benchmark: Write OpenBenchmarking.org GB/s, More Is Better cl-mem 2017-01-13 Benchmark: Write RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 200 400 600 800 1000 SE +/- 0.43, N = 3 SE +/- 0.33, N = 3 804.3 744.7 743.1 742.8 1. (CC) gcc options: -O2 -flto -lOpenCL
clpeak OpenCL Test: Global Memory Bandwidth OpenBenchmarking.org GBPS, More Is Better clpeak OpenCL Test: Global Memory Bandwidth RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 200 400 600 800 1000 SE +/- 0.19, N = 3 SE +/- 0.02, N = 3 870.62 813.47 813.45 813.42 1. (CXX) g++ options: -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dAXPY RTX 4090 NVIDIA 3090 NVIDIA RTX 3090 RTX 3090 170 340 510 680 850 SE +/- 0.58, N = 3 SE +/- 0.00, N = 3 771 723 722 722 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
NCNN Target: Vulkan GPU - Model: alexnet OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: alexnet RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 0.432 0.864 1.296 1.728 2.16 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 1.87 1.90 1.91 1.92 MIN: 1.83 / MAX: 2.45 MIN: 1.87 / MAX: 2.6 MIN: 1.86 / MAX: 9.04 MIN: 1.87 / MAX: 7.42 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
Warsow Resolution: 3840 x 2160 OpenBenchmarking.org Frames Per Second, More Is Better Warsow 2.5 Beta Resolution: 3840 x 2160 RTX 4090 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 200 400 600 800 1000 SE +/- 0.07, N = 3 SE +/- 0.78, N = 3 985.4 980.7 978.6 960.8
Warsow Resolution: 1920 x 1080 OpenBenchmarking.org Frames Per Second, More Is Better Warsow 2.5 Beta Resolution: 1920 x 1080 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 RTX 4090 200 400 600 800 1000 SE +/- 5.54, N = 3 SE +/- 7.67, N = 10 984.8 984.3 975.5 971.1
LeelaChessZero Backend: OpenCL OpenBenchmarking.org Nodes Per Second, More Is Better LeelaChessZero 0.28 Backend: OpenCL NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 5K 10K 15K 20K 25K SE +/- 98.83, N = 3 23029 22739 22711 1. (CXX) g++ options: -flto -pthread
Warsow Resolution: 1920 x 1200 OpenBenchmarking.org Frames Per Second, More Is Better Warsow 2.5 Beta Resolution: 1920 x 1200 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 RTX 4090 200 400 600 800 1000 SE +/- 5.60, N = 3 SE +/- 4.70, N = 3 984.9 984.8 980.3 973.0
Chaos Group V-RAY Mode: NVIDIA RTX GPU OpenBenchmarking.org vrays, More Is Better Chaos Group V-RAY 5 Mode: NVIDIA RTX GPU NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 600 1200 1800 2400 3000 SE +/- 11.72, N = 3 2856 2856 2829
Warsow Resolution: 2560 x 1440 OpenBenchmarking.org Frames Per Second, More Is Better Warsow 2.5 Beta Resolution: 2560 x 1440 NVIDIA RTX 3090 NVIDIA 3090 RTX 3090 RTX 4090 200 400 600 800 1000 SE +/- 0.35, N = 3 SE +/- 5.45, N = 3 985.7 985.5 984.7 980.1
Chaos Group V-RAY Mode: NVIDIA CUDA GPU OpenBenchmarking.org vpaths, More Is Better Chaos Group V-RAY 5 Mode: NVIDIA CUDA GPU RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 500 1000 1500 2000 2500 SE +/- 0.33, N = 3 2119 2115 2112
NeatBench Acceleration: GPU OpenBenchmarking.org FPS, More Is Better NeatBench 5 Acceleration: GPU RTX 4090 900 1800 2700 3600 4500 4090
Blender Blend File: BMW27 - Compute: CUDA OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.0 Blend File: BMW27 - Compute: CUDA RTX 4090 NVIDIA RTX 3090 RTX 3090 NVIDIA 3090 3 6 9 12 15 SE +/- 3.94, N = 15 SE +/- 0.01, N = 3 10.25 11.02 11.04 11.09
NCNN Target: Vulkan GPU - Model: squeezenet_ssd OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: squeezenet_ssd RTX 4090 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 5 10 15 20 25 SE +/- 0.03, N = 3 SE +/- 1.24, N = 3 3.43 15.96 17.31 19.73 MIN: 3.34 / MAX: 4.06 MIN: 5.77 / MAX: 36.07 MIN: 5.63 / MAX: 36.32 MIN: 7.36 / MAX: 35.55 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: googlenet OpenBenchmarking.org ms, Fewer Is Better NCNN 20210720 Target: Vulkan GPU - Model: googlenet RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 2 4 6 8 10 SE +/- 0.03, N = 3 SE +/- 0.26, N = 3 2.69 5.75 5.85 6.07 MIN: 2.2 / MAX: 4.95 MIN: 3.72 / MAX: 31 MIN: 3.71 / MAX: 29.49 MIN: 3.71 / MAX: 31.52 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
ViennaCL Test: CPU BLAS - dGEMV-N OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMV-N RTX 4090 RTX 3090 NVIDIA 3090 NVIDIA RTX 3090 20 40 60 80 100 SE +/- 9.12, N = 3 SE +/- 0.17, N = 3 93.8 68.1 67.5 67.1 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
Unvanquished Resolution: 2560 x 1440 - Effects Quality: Ultra OpenBenchmarking.org Frames Per Second, More Is Better Unvanquished 0.52.1 Resolution: 2560 x 1440 - Effects Quality: Ultra RTX 4090 NVIDIA 3090 RTX 3090 NVIDIA RTX 3090 130 260 390 520 650 SE +/- 0.07, N = 3 SE +/- 14.16, N = 12 608.2 466.6 461.7 452.5
ET: Legacy Resolution: 3840 x 2160 OpenBenchmarking.org Frames Per Second, More Is Better ET: Legacy 2.78 Resolution: 3840 x 2160 RTX 3090 NVIDIA RTX 3090 NVIDIA 3090 RTX 4090 140 280 420 560 700 SE +/- 6.04, N = 3 SE +/- 113.10, N = 6 648.4 638.7 635.5 219.3
Phoronix Test Suite v10.8.4