RTX 4070 SUPER Intel Core i9-13900K testing with a ASUS TUF GAMING Z790-PRO WIFI (1401 BIOS) and NVIDIA GeForce RTX 3090 24GB on EndeavourOS rolling via the Phoronix Test Suite.
HTML result view exported from: https://openbenchmarking.org/result/2402116-SADD-240207012&grs .
RTX 4070 SUPER Processor Motherboard Chipset Memory Disk Graphics Audio Monitor Network OS Kernel Desktop Display Server Display Driver OpenGL OpenCL Compiler File-System Screen Resolution NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 Intel Core i9-13900K @ 5.50GHz (24 Cores / 32 Threads) ASUS TUF GAMING Z790-PRO WIFI (1401 BIOS) Intel Device 7a27 32GB 4001GB Seagate ZP4000GP304001 ASUS NVIDIA GeForce RTX 4070 SUPER 12GB Realtek ALC1220 ARZOPA Intel I226-V + Intel Device 7a70 EndeavourOS rolling 6.7.1-arch1-1 (x86_64) KDE Plasma 5.27.10 X Server 1.21.1.11 NVIDIA 550.40.07 4.6.0 OpenCL 3.0 CUDA 12.4.74 GCC 13.2.1 20230801 ext4 1920x1080 MSI NVIDIA GeForce RTX 4070 12GB GCC 13.2.1 20230801 + CUDA 12.3 NVIDIA GeForce RTX 4070 Ti 12GB NVIDIA GeForce RTX 3090 24GB PI-KVM Video 6.7.4-arch1-1 (x86_64) OpenBenchmarking.org Kernel Details - Transparent Huge Pages: always Compiler Details - NVIDIA RTX 4070 SUPER: --disable-libssp --disable-libstdcxx-pch --disable-werror --enable-__cxa_atexit --enable-bootstrap --enable-cet=auto --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-default-ssp --enable-gnu-indirect-function --enable-gnu-unique-object --enable-languages=ada,c,c++,d,fortran,go,lto,objc,obj-c++ --enable-libstdcxx-backtrace --enable-link-serialization=1 --enable-lto --enable-multilib --enable-plugin --enable-shared --enable-threads=posix --mandir=/usr/share/man --with-build-config=bootstrap-lto --with-linker-hash-style=gnu - NVIDIA RTX 4070: --disable-libssp --disable-libstdcxx-pch --disable-werror --enable-__cxa_atexit --enable-bootstrap --enable-cet=auto --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-default-ssp --enable-gnu-indirect-function --enable-gnu-unique-object --enable-languages=ada,c,c++,d,fortran,go,lto,objc,obj-c++ --enable-libstdcxx-backtrace --enable-link-serialization=1 --enable-lto --enable-multilib --enable-plugin --enable-shared --enable-threads=posix --mandir=/usr/share/man --with-build-config=bootstrap-lto --with-linker-hash-style=gnu - NVIDIA RTX 4070 TI: --disable-libssp --disable-libstdcxx-pch --disable-werror --enable-__cxa_atexit --enable-bootstrap --enable-cet=auto --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-default-ssp --enable-gnu-indirect-function --enable-gnu-unique-object --enable-languages=ada,c,c++,d,fortran,go,lto,objc,obj-c++ --enable-libstdcxx-backtrace --enable-link-serialization=1 --enable-lto --enable-multilib --enable-plugin --enable-shared --enable-threads=posix --mandir=/usr/share/man --with-build-config=bootstrap-lto --with-linker-hash-style=gnu - NVIDIA RTX 3090: --disable-libssp --disable-libstdcxx-pch --disable-werror --enable-__cxa_atexit --enable-bootstrap --enable-cet=auto --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-default-ssp --enable-gnu-indirect-function --enable-gnu-unique-object --enable-languages=ada,c,c++,d,fortran,go,lto,m2,objc,obj-c++ --enable-libstdcxx-backtrace --enable-link-serialization=1 --enable-lto --enable-multilib --enable-plugin --enable-shared --enable-threads=posix --mandir=/usr/share/man --with-build-config=bootstrap-lto --with-linker-hash-style=gnu Processor Details - Scaling Governor: intel_pstate powersave (EPP: balance_performance) - CPU Microcode: 0x11d Graphics Details - NVIDIA RTX 4070 SUPER: BAR1 / Visible vRAM Size: 16384 MiB - vBIOS Version: 95.04.69.00.c1 - NVIDIA RTX 4070: BAR1 / Visible vRAM Size: 16384 MiB - vBIOS Version: 95.04.3e.40.2a - NVIDIA RTX 4070 TI: BAR1 / Visible vRAM Size: 16384 MiB - vBIOS Version: 95.04.31.00.36 - NVIDIA RTX 3090: BAR1 / Visible vRAM Size: 256 MiB - vBIOS Version: 94.02.26.08.ba Security Details - gather_data_sampling: Not affected + itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: Not affected + spec_rstack_overflow: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced / Automatic IBRS IBPB: conditional RSB filling PBRSB-eIBRS: SW sequence + srbds: Not affected + tsx_async_abort: Not affected Environment Details - NVIDIA RTX 4070, NVIDIA RTX 4070 TI, NVIDIA RTX 3090: NVCC_PREPEND_FLAGS="-ccbin /opt/cuda/bin" Python Details - NVIDIA RTX 4070, NVIDIA RTX 4070 TI, NVIDIA RTX 3090: Python 3.11.6
RTX 4070 SUPER vkfft: FFT + iFFT C2C 1D batched in half precision opencl-benchmark: Memory Bandwidth Coalesced Write vkfft: FFT + iFFT C2C 1D batched in single precision, no reshuffling vkfft: FFT + iFFT C2C 1D batched in single precision clpeak: Global Memory Bandwidth opencl-benchmark: Memory Bandwidth Coalesced Read cl-mem: Write cl-mem: Read vkresample: 2x - Single viennacl: OpenCL BLAS - dAXPY namd-cuda: ATPase Simulation - 327,506 Atoms libplacebo: hdr_peakdetect viennacl: OpenCL BLAS - dDOT luxcorerender: Rainbow Colors and Prism - GPU viennacl: OpenCL BLAS - dCOPY realsr-ncnn: 4x - Yes opencl-benchmark: INT64 Compute vkfft: FFT + iFFT C2C 1D batched in double precision clpeak: Integer Compute INT clpeak: Single-Precision Float neatbench: GPU hashcat: MD5 hashcat: TrueCrypt RIPEMD160 + XTS opencl-benchmark: INT8 Compute hashcat: SHA-512 gpuowl: 332220523 clpeak: Double-Precision Double opencl-benchmark: FP64 Compute hashcat: SHA1 hashcat: 7-Zip viennacl: OpenCL BLAS - dGEMM-TT vkresample: 2x - Double opencl-benchmark: FP32 Compute gpuowl: 57885161 opencl-benchmark: INT32 Compute viennacl: OpenCL BLAS - dGEMM-TN viennacl: OpenCL BLAS - dGEMM-NT viennacl: OpenCL BLAS - sAXPY opencl-benchmark: INT16 Compute mandelgpu: GPU viennacl: OpenCL BLAS - dGEMM-NN gpuowl: 77936867 libplacebo: deband_heavy libplacebo: polar_nocompute rodinia: OpenCL Particle Filter blender: Classroom - NVIDIA OptiX blender: Pabellon Barcelona - NVIDIA OptiX luxcorerender: Danish Mood - GPU blender: Fishy Cat - NVIDIA OptiX pytorch: NVIDIA CUDA GPU - 256 - ResNet-152 pytorch: NVIDIA CUDA GPU - 32 - ResNet-152 pytorch: NVIDIA CUDA GPU - 16 - ResNet-50 pytorch: NVIDIA CUDA GPU - 512 - ResNet-50 luxcorerender: LuxCore Benchmark - GPU pytorch: NVIDIA CUDA GPU - 64 - ResNet-50 pytorch: NVIDIA CUDA GPU - 256 - ResNet-50 fahbench: pytorch: NVIDIA CUDA GPU - 32 - ResNet-50 pytorch: NVIDIA CUDA GPU - 64 - ResNet-152 libplacebo: hdr_lut vkfft: FFT + iFFT C2C Bluestein benchmark in double precision pytorch: NVIDIA CUDA GPU - 16 - ResNet-152 pytorch: NVIDIA CUDA GPU - 512 - ResNet-152 vkfft: FFT + iFFT R2C / C2R luxcorerender: Orange Juice - GPU blender: BMW27 - NVIDIA OptiX blender: Barbershop - NVIDIA OptiX indigobench: OpenCL GPU - Bedroom octanebench: Total Score viennacl: OpenCL BLAS - dGEMV-N waifu2x-ncnn: 2x - 3 - Yes vkfft: FFT + iFFT C2C Bluestein in single precision indigobench: OpenCL GPU - Supercar viennacl: OpenCL BLAS - sCOPY viennacl: CPU BLAS - dGEMM-TT cl-mem: Copy vkfft: FFT + iFFT C2C multidimensional in single precision viennacl: CPU BLAS - dGEMM-TN viennacl: CPU BLAS - dGEMM-NN tensorflow: GPU - 1 - AlexNet pytorch: NVIDIA CUDA GPU - 16 - Efficientnet_v2_l viennacl: OpenCL BLAS - dGEMV-T pytorch: NVIDIA CUDA GPU - 512 - Efficientnet_v2_l viennacl: CPU BLAS - dGEMM-NT viennacl: OpenCL BLAS - sDOT pytorch: NVIDIA CUDA GPU - 256 - Efficientnet_v2_l pytorch: NVIDIA CUDA GPU - 64 - Efficientnet_v2_l pytorch: NVIDIA CUDA GPU - 1 - Efficientnet_v2_l pytorch: NVIDIA CUDA GPU - 1 - ResNet-152 tensorflow: GPU - 1 - VGG-16 viennacl: CPU BLAS - sAXPY libplacebo: av1_grain_lap tensorflow: GPU - 16 - AlexNet viennacl: CPU BLAS - dDOT tensorflow: GPU - 1 - GoogLeNet viennacl: CPU BLAS - dCOPY tensorflow: GPU - 512 - AlexNet tensorflow: GPU - 16 - VGG-16 tensorflow: GPU - 256 - AlexNet tensorflow: GPU - 32 - GoogLeNet viennacl: CPU BLAS - dAXPY tensorflow: GPU - 32 - ResNet-50 viennacl: CPU BLAS - dGEMV-N tensorflow: GPU - 64 - GoogLeNet vkpeak: fp32-vec4 viennacl: CPU BLAS - sCOPY tensorflow: GPU - 64 - ResNet-50 tensorflow: GPU - 32 - AlexNet tensorflow: GPU - 1 - ResNet-50 tensorflow: GPU - 256 - VGG-16 tensorflow: GPU - 64 - VGG-16 tensorflow: GPU - 16 - ResNet-50 vkpeak: fp32-scalar tensorflow: GPU - 64 - AlexNet vkpeak: int16-scalar vkpeak: fp16-scalar vkpeak: fp16-vec4 vkpeak: int16-vec4 tensorflow: GPU - 16 - GoogLeNet vkpeak: int32-scalar vkpeak: fp64-vec4 vkpeak: int32-vec4 vkpeak: fp64-scalar tensorflow: GPU - 32 - VGG-16 ncnn: Vulkan GPU - FastestDet ncnn: Vulkan GPU - vision_transformer ncnn: Vulkan GPU - regnety_400m ncnn: Vulkan GPU - squeezenet_ssd ncnn: Vulkan GPU - yolov4-tiny ncnn: Vulkan GPU - resnet50 ncnn: Vulkan GPU - alexnet ncnn: Vulkan GPU - resnet18 ncnn: Vulkan GPU - vgg16 ncnn: Vulkan GPU - googlenet ncnn: Vulkan GPU - blazeface ncnn: Vulkan GPU - efficientnet-b0 ncnn: Vulkan GPU - mnasnet ncnn: Vulkan GPU - shufflenet-v2 ncnn: Vulkan GPU-v3-v3 - mobilenet-v3 ncnn: Vulkan GPU-v2-v2 - mobilenet-v2 ncnn: Vulkan GPU - mobilenet viennacl: CPU BLAS - dGEMV-T viennacl: CPU BLAS - sDOT financebench: Black-Scholes OpenCL luxcorerender: DLSC - GPU realsr-ncnn: 4x - No pytorch: NVIDIA CUDA GPU - 32 - Efficientnet_v2_l pytorch: NVIDIA CUDA GPU - 1 - ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 131705 455.01 75078 73929 437.65 464.86 407.5 446.2 18.489 437 0.06791 3292.37 458 27.67 423 34.885 4.214 24317 18170.54 35492.69 4070 67583033333 802967 14.307 3232733333 137.44 630.11 0.621 22132600000 1176467 613 339.593 38.594 869.07 19.889 599 584 392 17.170 587219538.2 577 646.41 2186.70 2327.55 3.480 12.60 14.29 10.56 9.45 194.58 195.39 509.45 504.27 12.82 507.45 504.67 366.0576 501.50 196.07 3905.98 4451 195.40 195.30 54794 11.72 5.57 51.30 19.801 720.973789 210 2.855 15166 52.813 334 122 331.8 50299 115 119 13.92 389 103.57 117 370 103.17 102.60 106.37 201.94 1.35 156 4171.00 31.59 96.8 12.62 70.8 35.10 1.48 34.16 15.61 87.2 5.51 102 15.52 132 5.55 33.4 4.35 5.46 33.97 15.67 1.50 2.86 844.61 11.11 6.86 63.82 46.26 16.17 8.97 117.81 11.04 0.84 5.07 3.85 2.31 2.25 3.03 8.62 109 165 5.912 13.59 6.323 102.60 557.73 137762 459.43 79057 77774 437.21 465.18 406.7 446.3 18.016 455 0.07498 3329.26 456 23.26 423 42.852 3.443 22390 14555.19 28479.39 4070 56147866667 660967 12.116 2673300000 112.61 515.17 0.510 18202466667 976967 502 415.160 31.768 714.80 16.377 494 477 389 14.284 516770131.2 473 530.32 1843.26 1968.37 4.098 14.86 16.55 8.89 11.03 187.27 187.69 458.39 459.27 10.92 458.36 459.93 317.1952 459.94 186.63 3946.90 3886 187.26 187.51 47097 10.40 6.21 58.44 18.203 647.997867 209 3.168 13714 48.517 330 118 330.3 47212 121 122 14.04 103.68 387 101.43 122 362 101.24 101.55 107.59 198.18 1.36 153 4152.41 31.45 96.7 12.78 71.0 35.21 1.50 15.63 86.8 5.55 103 15.54 131 5.55 33.32 4.34 1.50 5.49 33.93 15.66 1.5 2.67 382.82 6.21 5.18 20.74 8.72 5.78 5.11 45.52 6.06 0.84 3.59 2.24 2.08 2.15 2.48 7.20 109 166 6.906 11.74 7.092 102.90 546.76 136210 457.17 75141 73942 437.63 465.07 412.2 446.3 18.456 437 0.06788 3475.06 457 27.71 424 33.626 4.420 25431 19821.10 38691.73 4070 73312233333 858600 15.731 3462500000 145.84 667.05 0.660 23532400000 1262633 648 322.064 40.914 919.13 21.047 634 612 393 18.281 619106132.5 604 676.59 2306.56 2459.03 3.291 12.30 13.97 10.99 9.02 195.86 198.82 502.92 504.66 13.23 505.62 382.1637 505.55 197.02 3976.04 4647 194.29 194.87 55446 11.89 5.43 50.73 20.256 735.940593 211 2.854 15125 53.589 336 124 333.3 51528 125 117 14.79 103.45 391 103.50 118 365 103.24 103.20 108.59 201.19 1.38 156 4143.96 31.70 96.4 12.79 71.3 35.44 1.49 34.61 15.81 87.3 5.50 103 15.50 132 5.53 33.29 4.32 1.5 1.5 5.46 34.06 15.69 1.5 3.04 497.66 5.89 6.13 16.37 12.25 6.07 7.74 34.49 7.37 0.82 3.46 4.14 2.01 2.09 2.54 7.45 102.7 168 5.226 13.95 5.962 96.50 535.39 273221 887.31 144311 141876 816.55 864.11 753.8 825.8 10.323 724 0.10822 5055.88 659 33.29 605 30.313 3.135 30912 17923.33 34906.79 3090 67177300000 797833 13.727 3081866667 137.32 642.23 0.637 21323733333 1056000 593 333.639 39.395 866.31 20.027 594 595 498 17.001 484098913.8 592 645.99 2020.16 2116.79 3.844 15.26 17.30 10.20 10.64 161.01 163.74 419.76 416.20 13.12 419.03 416.89 343.0199 420.29 164.14 3369.88 4195 164.14 164.35 48418 12.14 6.31 54.30 20.959 674.250912 187 3.202 14205 52.014 363 113 360.8 50856 121 113 14.45 98.11 374 99.25 119 376 99.43 99.84 105.55 197.12 1.38 154 4100.36 31.98 95.2 12.82 70.2 35.58 1.49 34.46 15.67 86.2 5.57 103 15.63 26699.66 132 5.57 33.53 4.35 1.51 1.51 5.49 20353.95 33.93 13264.91 20151.44 39860.80 16329.72 15.68 20295.27 638.74 20009.73 638.84 1.5 2.65 354.57 6.73 4.90 11.29 12.70 3.60 4.12 17.88 6.14 0.87 3.34 2.16 2.04 2.21 2.34 7.27 110 132.1 5.741 12.99 5.556 99.05 525.12 OpenBenchmarking.org
VkFFT Test: FFT + iFFT C2C 1D batched in half precision OpenBenchmarking.org Benchmark Score, More Is Better VkFFT 1.2.31 Test: FFT + iFFT C2C 1D batched in half precision NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 60K 120K 180K 240K 300K SE +/- 159.17, N = 3 SE +/- 1301.92, N = 3 SE +/- 1708.38, N = 3 SE +/- 160.60, N = 3 131705 137762 136210 273221 1. (CXX) g++ options: -O3 -lrt
ProjectPhysX OpenCL-Benchmark Operation: Memory Bandwidth Coalesced Write OpenBenchmarking.org GB/s, More Is Better ProjectPhysX OpenCL-Benchmark 1.2 Operation: Memory Bandwidth Coalesced Write NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 0.14, N = 3 SE +/- 0.16, N = 3 SE +/- 0.11, N = 3 SE +/- 0.06, N = 3 455.01 459.43 457.17 887.31 1. (CXX) g++ options: -std=c++17 -pthread -lOpenCL
VkFFT Test: FFT + iFFT C2C 1D batched in single precision, no reshuffling OpenBenchmarking.org Benchmark Score, More Is Better VkFFT 1.2.31 Test: FFT + iFFT C2C 1D batched in single precision, no reshuffling NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 30K 60K 90K 120K 150K SE +/- 37.77, N = 3 SE +/- 5.84, N = 3 SE +/- 28.54, N = 3 SE +/- 37.44, N = 3 75078 79057 75141 144311 1. (CXX) g++ options: -O3 -lrt
VkFFT Test: FFT + iFFT C2C 1D batched in single precision OpenBenchmarking.org Benchmark Score, More Is Better VkFFT 1.2.31 Test: FFT + iFFT C2C 1D batched in single precision NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 30K 60K 90K 120K 150K SE +/- 7.94, N = 3 SE +/- 13.72, N = 3 SE +/- 0.88, N = 3 SE +/- 9.64, N = 3 73929 77774 73942 141876 1. (CXX) g++ options: -O3 -lrt
clpeak OpenCL Test: Global Memory Bandwidth OpenBenchmarking.org GBPS, More Is Better clpeak 1.1.2 OpenCL Test: Global Memory Bandwidth NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 0.02, N = 3 SE +/- 0.02, N = 3 SE +/- 0.02, N = 3 SE +/- 0.02, N = 3 437.65 437.21 437.63 816.55 1. (CXX) g++ options: -O3
ProjectPhysX OpenCL-Benchmark Operation: Memory Bandwidth Coalesced Read OpenBenchmarking.org GB/s, More Is Better ProjectPhysX OpenCL-Benchmark 1.2 Operation: Memory Bandwidth Coalesced Read NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 0.01, N = 3 SE +/- 0.03, N = 3 SE +/- 0.01, N = 3 SE +/- 0.07, N = 3 464.86 465.18 465.07 864.11 1. (CXX) g++ options: -std=c++17 -pthread -lOpenCL
cl-mem Benchmark: Write OpenBenchmarking.org GB/s, More Is Better cl-mem 2017-01-13 Benchmark: Write NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 160 320 480 640 800 SE +/- 1.11, N = 3 SE +/- 0.55, N = 3 SE +/- 0.12, N = 3 SE +/- 0.83, N = 3 407.5 406.7 412.2 753.8 1. (CC) gcc options: -O2 -flto -lOpenCL
cl-mem Benchmark: Read OpenBenchmarking.org GB/s, More Is Better cl-mem 2017-01-13 Benchmark: Read NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 0.12, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.32, N = 3 446.2 446.3 446.3 825.8 1. (CC) gcc options: -O2 -flto -lOpenCL
VkResample Upscale: 2x - Precision: Single OpenBenchmarking.org ms, Fewer Is Better VkResample 1.0 Upscale: 2x - Precision: Single NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 5 10 15 20 25 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 18.49 18.02 18.46 10.32 1. (CXX) g++ options: -O3
ViennaCL Test: OpenCL BLAS - dAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dAXPY NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 160 320 480 640 800 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.33, N = 3 SE +/- 0.58, N = 3 437 455 437 724 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
NAMD CUDA ATPase Simulation - 327,506 Atoms OpenBenchmarking.org days/ns, Fewer Is Better NAMD CUDA 2.14 ATPase Simulation - 327,506 Atoms NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.0243 0.0486 0.0729 0.0972 0.1215 SE +/- 0.00031, N = 3 SE +/- 0.00021, N = 3 SE +/- 0.00061, N = 3 SE +/- 0.00042, N = 3 0.06791 0.07498 0.06788 0.10822
Libplacebo Test: hdr_peakdetect OpenBenchmarking.org FPS, More Is Better Libplacebo 5.229.1 Test: hdr_peakdetect NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 1100 2200 3300 4400 5500 SE +/- 3.65, N = 3 SE +/- 11.75, N = 3 SE +/- 99.97, N = 3 SE +/- 43.13, N = 3 3292.37 3310.02 3544.60 4997.08 1. (CXX) g++ options: -lm -pthread -ldl -fvisibility=hidden -std=c++20 -O2 -fno-math-errno -fPIC -MD -MQ -MF
ViennaCL Test: OpenCL BLAS - dDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dDOT NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 140 280 420 560 700 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.88, N = 3 458 456 457 659 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
LuxCoreRender Scene: Rainbow Colors and Prism - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.6 Scene: Rainbow Colors and Prism - Acceleration: GPU NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 8 16 24 32 40 SE +/- 0.03, N = 3 SE +/- 0.01, N = 3 SE +/- 0.07, N = 3 SE +/- 0.36, N = 5 27.67 23.26 27.71 33.29 MIN: 24.87 / MAX: 29.03 MIN: 20.92 / MAX: 24.3 MIN: 25.01 / MAX: 29.15 MIN: 30.4 / MAX: 36.21
ViennaCL Test: OpenCL BLAS - dCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dCOPY NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 130 260 390 520 650 SE +/- 0.33, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.58, N = 3 423 423 424 605 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
RealSR-NCNN Scale: 4x - TAA: Yes OpenBenchmarking.org Seconds, Fewer Is Better RealSR-NCNN 20200818 Scale: 4x - TAA: Yes NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 10 20 30 40 50 SE +/- 0.02, N = 3 SE +/- 0.23, N = 3 SE +/- 0.03, N = 3 SE +/- 0.06, N = 3 34.89 42.85 33.63 30.31
ProjectPhysX OpenCL-Benchmark Operation: INT64 Compute OpenBenchmarking.org TIOPs/s, More Is Better ProjectPhysX OpenCL-Benchmark 1.2 Operation: INT64 Compute NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.9945 1.989 2.9835 3.978 4.9725 SE +/- 0.015, N = 3 SE +/- 0.004, N = 3 SE +/- 0.016, N = 3 SE +/- 0.003, N = 3 4.214 3.443 4.420 3.135 1. (CXX) g++ options: -std=c++17 -pthread -lOpenCL
VkFFT Test: FFT + iFFT C2C 1D batched in double precision OpenBenchmarking.org Benchmark Score, More Is Better VkFFT 1.2.31 Test: FFT + iFFT C2C 1D batched in double precision NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 7K 14K 21K 28K 35K SE +/- 146.69, N = 3 SE +/- 125.94, N = 3 SE +/- 302.46, N = 3 SE +/- 50.66, N = 3 24317 22390 25431 30912 1. (CXX) g++ options: -O3 -lrt
clpeak OpenCL Test: Integer Compute INT OpenBenchmarking.org GIOPS, More Is Better clpeak 1.1.2 OpenCL Test: Integer Compute INT NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4K 8K 12K 16K 20K SE +/- 3.14, N = 3 SE +/- 15.26, N = 3 SE +/- 2.50, N = 3 SE +/- 16.49, N = 3 18170.54 14555.19 19821.10 17923.33 1. (CXX) g++ options: -O3
clpeak OpenCL Test: Single-Precision Float OpenBenchmarking.org GFLOPS, More Is Better clpeak 1.1.2 OpenCL Test: Single-Precision Float NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 8K 16K 24K 32K 40K SE +/- 0.99, N = 3 SE +/- 5.46, N = 3 SE +/- 11.67, N = 3 SE +/- 113.39, N = 3 35492.69 28479.39 38691.73 34906.79 1. (CXX) g++ options: -O3
NeatBench Acceleration: GPU OpenBenchmarking.org FPS, More Is Better NeatBench 5 Acceleration: GPU NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 900 1800 2700 3600 4500 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 4070 4070 4070 3090
Hashcat Benchmark: MD5 OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: MD5 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 16000M 32000M 48000M 64000M 80000M SE +/- 22430807.19, N = 3 SE +/- 33772046.30, N = 3 SE +/- 11283665.68, N = 3 SE +/- 53667246.37, N = 3 67583033333 56147866667 73312233333 67177300000
Hashcat Benchmark: TrueCrypt RIPEMD160 + XTS OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: TrueCrypt RIPEMD160 + XTS NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 200K 400K 600K 800K 1000K SE +/- 633.33, N = 3 SE +/- 176.38, N = 3 SE +/- 888.82, N = 3 SE +/- 1757.21, N = 3 802967 660967 858600 797833
ProjectPhysX OpenCL-Benchmark Operation: INT8 Compute OpenBenchmarking.org TIOPs/s, More Is Better ProjectPhysX OpenCL-Benchmark 1.2 Operation: INT8 Compute NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.05, N = 3 SE +/- 0.02, N = 3 SE +/- 0.03, N = 3 SE +/- 0.07, N = 3 14.31 12.12 15.73 13.73 1. (CXX) g++ options: -std=c++17 -pthread -lOpenCL
Hashcat Benchmark: SHA-512 OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: SHA-512 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 700M 1400M 2100M 2800M 3500M SE +/- 1530068.99, N = 3 SE +/- 1059874.21, N = 3 SE +/- 721110.26, N = 3 SE +/- 3288532.26, N = 3 3232733333 2673300000 3462500000 3081866667
GpuOwl Exponent: 332220523 OpenBenchmarking.org Iterations / Second, More Is Better GpuOwl 7.2.1 Exponent: 332220523 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 30 60 90 120 150 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 137.44 112.61 145.84 137.32
clpeak OpenCL Test: Double-Precision Double OpenBenchmarking.org GFLOPS, More Is Better clpeak 1.1.2 OpenCL Test: Double-Precision Double NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 140 280 420 560 700 SE +/- 0.98, N = 3 SE +/- 0.21, N = 3 SE +/- 1.33, N = 3 SE +/- 1.63, N = 3 630.11 515.17 667.05 642.23 1. (CXX) g++ options: -O3
ProjectPhysX OpenCL-Benchmark Operation: FP64 Compute OpenBenchmarking.org TFLOPs/s, More Is Better ProjectPhysX OpenCL-Benchmark 1.2 Operation: FP64 Compute NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.1485 0.297 0.4455 0.594 0.7425 SE +/- 0.000, N = 3 SE +/- 0.001, N = 3 SE +/- 0.001, N = 3 SE +/- 0.001, N = 3 0.621 0.510 0.660 0.637 1. (CXX) g++ options: -std=c++17 -pthread -lOpenCL
Hashcat Benchmark: SHA1 OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: SHA1 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 5000M 10000M 15000M 20000M 25000M SE +/- 5140363.15, N = 3 SE +/- 6318315.53, N = 3 SE +/- 15926811.78, N = 3 SE +/- 26244639.66, N = 3 22132600000 18202466667 23532400000 21323733333
Hashcat Benchmark: 7-Zip OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: 7-Zip NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 300K 600K 900K 1200K 1500K SE +/- 1991.93, N = 3 SE +/- 2062.63, N = 3 SE +/- 2339.04, N = 3 SE +/- 1587.45, N = 3 1176467 976967 1262633 1056000
ViennaCL Test: OpenCL BLAS - dGEMM-TT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-TT NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 140 280 420 560 700 SE +/- 0.00, N = 3 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 613 502 648 593 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
VkResample Upscale: 2x - Precision: Double OpenBenchmarking.org ms, Fewer Is Better VkResample 1.0 Upscale: 2x - Precision: Double NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 90 180 270 360 450 SE +/- 0.30, N = 3 SE +/- 0.77, N = 3 SE +/- 0.35, N = 3 SE +/- 0.30, N = 3 339.59 415.16 322.06 333.64 1. (CXX) g++ options: -O3
ProjectPhysX OpenCL-Benchmark Operation: FP32 Compute OpenBenchmarking.org TFLOPs/s, More Is Better ProjectPhysX OpenCL-Benchmark 1.2 Operation: FP32 Compute NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 9 18 27 36 45 SE +/- 0.03, N = 3 SE +/- 0.03, N = 3 SE +/- 0.00, N = 3 SE +/- 0.10, N = 3 38.59 31.77 40.91 39.40 1. (CXX) g++ options: -std=c++17 -pthread -lOpenCL
GpuOwl Exponent: 57885161 OpenBenchmarking.org Iterations / Second, More Is Better GpuOwl 7.2.1 Exponent: 57885161 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 1.26, N = 3 SE +/- 0.00, N = 3 SE +/- 2.53, N = 3 SE +/- 2.01, N = 3 869.07 714.80 919.13 866.31
ProjectPhysX OpenCL-Benchmark Operation: INT32 Compute OpenBenchmarking.org TIOPs/s, More Is Better ProjectPhysX OpenCL-Benchmark 1.2 Operation: INT32 Compute NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 5 10 15 20 25 SE +/- 0.00, N = 3 SE +/- 0.02, N = 3 SE +/- 0.04, N = 3 SE +/- 0.06, N = 3 19.89 16.38 21.05 20.03 1. (CXX) g++ options: -std=c++17 -pthread -lOpenCL
ViennaCL Test: OpenCL BLAS - dGEMM-TN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-TN NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 140 280 420 560 700 SE +/- 0.00, N = 3 SE +/- 0.33, N = 3 SE +/- 0.67, N = 3 SE +/- 2.03, N = 3 599 494 634 594 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dGEMM-NT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-NT NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 130 260 390 520 650 SE +/- 0.00, N = 3 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 SE +/- 2.33, N = 3 584 477 612 595 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 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 110 220 330 440 550 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.58, N = 3 392 389 393 498 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ProjectPhysX OpenCL-Benchmark Operation: INT16 Compute OpenBenchmarking.org TIOPs/s, More Is Better ProjectPhysX OpenCL-Benchmark 1.2 Operation: INT16 Compute NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.00, N = 3 SE +/- 0.02, N = 3 SE +/- 0.02, N = 3 SE +/- 0.00, N = 3 17.17 14.28 18.28 17.00 1. (CXX) g++ options: -std=c++17 -pthread -lOpenCL
MandelGPU OpenCL Device: GPU OpenBenchmarking.org Samples/sec, More Is Better MandelGPU 1.3pts1 OpenCL Device: GPU NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 130M 260M 390M 520M 650M SE +/- 467034.80, N = 3 SE +/- 1783157.89, N = 3 SE +/- 1202791.77, N = 3 SE +/- 794770.01, N = 3 587219538.2 516770131.2 619106132.5 484098913.8 1. (CC) gcc options: -O3 -lm -ftree-vectorize -funroll-loops -lglut -lOpenCL -lGL
ViennaCL Test: OpenCL BLAS - dGEMM-NN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-NN NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 130 260 390 520 650 SE +/- 0.00, N = 3 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 SE +/- 2.31, N = 3 577 473 604 592 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
GpuOwl Exponent: 77936867 OpenBenchmarking.org Iterations / Second, More Is Better GpuOwl 7.2.1 Exponent: 77936867 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 150 300 450 600 750 SE +/- 0.00, N = 3 SE +/- 0.09, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 646.41 530.32 676.59 645.99
Libplacebo Test: deband_heavy OpenBenchmarking.org FPS, More Is Better Libplacebo 5.229.1 Test: deband_heavy NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 500 1000 1500 2000 2500 SE +/- 2.26, N = 3 SE +/- 0.08, N = 3 SE +/- 0.56, N = 3 SE +/- 4.93, N = 3 2186.70 1847.98 2306.67 2017.75 1. (CXX) g++ options: -lm -pthread -ldl -fvisibility=hidden -std=c++20 -O2 -fno-math-errno -fPIC -MD -MQ -MF
Libplacebo Test: polar_nocompute OpenBenchmarking.org FPS, More Is Better Libplacebo 5.229.1 Test: polar_nocompute NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 500 1000 1500 2000 2500 SE +/- 0.24, N = 3 SE +/- 0.16, N = 3 SE +/- 0.26, N = 3 SE +/- 7.22, N = 3 2327.55 1972.78 2461.23 2119.89 1. (CXX) g++ options: -lm -pthread -ldl -fvisibility=hidden -std=c++20 -O2 -fno-math-errno -fPIC -MD -MQ -MF
Rodinia Test: OpenCL Particle Filter OpenBenchmarking.org Seconds, Fewer Is Better Rodinia 3.1 Test: OpenCL Particle Filter NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.9221 1.8442 2.7663 3.6884 4.6105 SE +/- 0.039, N = 4 SE +/- 0.008, N = 3 SE +/- 0.002, N = 3 SE +/- 0.030, N = 15 3.480 4.098 3.291 3.844 1. (CXX) g++ options: -O2 -lOpenCL
Blender Blend File: Classroom - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 4.0 Blend File: Classroom - Compute: NVIDIA OptiX NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.00, N = 3 SE +/- 0.03, N = 3 SE +/- 0.03, N = 3 SE +/- 0.02, N = 3 12.60 14.86 12.30 15.26
Blender Blend File: Pabellon Barcelona - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 4.0 Blend File: Pabellon Barcelona - Compute: NVIDIA OptiX NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.03, N = 3 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 SE +/- 0.02, N = 3 14.29 16.55 13.97 17.30
LuxCoreRender Scene: Danish Mood - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.6 Scene: Danish Mood - Acceleration: GPU NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 3 6 9 12 15 SE +/- 0.08, N = 3 SE +/- 0.06, N = 3 SE +/- 0.11, N = 3 SE +/- 0.04, N = 3 10.56 8.89 10.99 10.20 MIN: 3.7 / MAX: 12.17 MIN: 3.32 / MAX: 10.26 MIN: 4.17 / MAX: 12.71 MIN: 4.07 / MAX: 11.93
Blender Blend File: Fishy Cat - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 4.0 Blend File: Fishy Cat - Compute: NVIDIA OptiX NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 3 6 9 12 15 SE +/- 0.06, N = 13 SE +/- 0.03, N = 3 SE +/- 0.01, N = 3 SE +/- 0.08, N = 9 9.45 11.03 9.02 10.64
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 256 - Model: ResNet-152 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 256 - Model: ResNet-152 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 40 80 120 160 200 SE +/- 1.14, N = 2 SE +/- 0.17, N = 3 SE +/- 0.19, N = 2 194.58 187.27 195.86 161.01 MIN: 183.74 / MAX: 198.52 MIN: 179.9 / MAX: 188.08 MIN: 181.64 / MAX: 199.2 MIN: 138.12 / MAX: 165.16
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 32 - Model: ResNet-152 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 32 - Model: ResNet-152 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 40 80 120 160 200 SE +/- 0.29, N = 3 195.39 187.69 198.82 163.74 MIN: 183.94 / MAX: 198.7 MIN: 182.03 / MAX: 188.31 MIN: 188.33 / MAX: 201.47 MIN: 144.93 / MAX: 165.03
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 16 - Model: ResNet-50 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 16 - Model: ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 110 220 330 440 550 SE +/- 0.26, N = 3 SE +/- 2.23, N = 3 SE +/- 0.89, N = 2 509.45 458.39 502.92 419.76 MIN: 430.1 / MAX: 516.48 MIN: 404.5 / MAX: 461.01 MIN: 415.65 / MAX: 520.39 MIN: 376.2 / MAX: 422.17
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 512 - Model: ResNet-50 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 512 - Model: ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 110 220 330 440 550 SE +/- 4.43, N = 2 SE +/- 0.43, N = 2 SE +/- 0.83, N = 2 SE +/- 0.40, N = 3 504.27 459.27 504.66 416.20 MIN: 418.22 / MAX: 512.44 MIN: 405.48 / MAX: 461.88 MIN: 424.27 / MAX: 509.08 MIN: 355.45 / MAX: 419.05
LuxCoreRender Scene: LuxCore Benchmark - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.6 Scene: LuxCore Benchmark - Acceleration: GPU NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 3 6 9 12 15 SE +/- 0.02, N = 3 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 SE +/- 0.03, N = 2 12.82 10.92 13.23 13.12 MIN: 4.84 / MAX: 14.62 MIN: 4.45 / MAX: 12.42 MIN: 5.41 / MAX: 15.13 MIN: 4.85 / MAX: 15.21
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 64 - Model: ResNet-50 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 64 - Model: ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 110 220 330 440 550 SE +/- 0.92, N = 3 SE +/- 0.27, N = 3 SE +/- 1.92, N = 3 SE +/- 0.24, N = 3 507.45 458.36 505.62 419.03 MIN: 423.41 / MAX: 512.88 MIN: 404.89 / MAX: 461.01 MIN: 426.6 / MAX: 513.25 MIN: 376 / MAX: 422
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 256 - Model: ResNet-50 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 256 - Model: ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 3090 110 220 330 440 550 SE +/- 1.39, N = 3 SE +/- 0.34, N = 3 SE +/- 0.14, N = 3 504.67 459.93 416.89 MIN: 412.34 / MAX: 514.07 MIN: 403.65 / MAX: 462.74 MIN: 329.77 / MAX: 420.82
FAHBench OpenBenchmarking.org Ns Per Day, More Is Better FAHBench 2.3.2 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 80 160 240 320 400 SE +/- 0.39, N = 3 SE +/- 0.12, N = 3 SE +/- 0.26, N = 3 SE +/- 0.26, N = 3 366.06 317.20 382.16 343.02
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 32 - Model: ResNet-50 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 32 - Model: ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 110 220 330 440 550 SE +/- 2.17, N = 2 SE +/- 0.13, N = 2 SE +/- 1.69, N = 3 501.50 459.94 505.55 420.29 MIN: 415.94 / MAX: 510.69 MIN: 403.65 / MAX: 462.59 MIN: 419.93 / MAX: 512.69 MIN: 376.81 / MAX: 421.58
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 64 - Model: ResNet-152 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 64 - Model: ResNet-152 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 40 80 120 160 200 SE +/- 0.51, N = 3 SE +/- 0.34, N = 3 SE +/- 0.78, N = 2 196.07 186.63 197.02 164.14 MIN: 171.95 / MAX: 199.96 MIN: 180.51 / MAX: 187.79 MIN: 183.92 / MAX: 200.54 MIN: 149 / MAX: 165
Libplacebo Test: hdr_lut OpenBenchmarking.org FPS, More Is Better Libplacebo 5.229.1 Test: hdr_lut NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 900 1800 2700 3600 4500 SE +/- 12.09, N = 3 SE +/- 10.06, N = 3 SE +/- 5.47, N = 3 SE +/- 13.62, N = 3 3905.98 3927.11 3971.61 3313.26 1. (CXX) g++ options: -lm -pthread -ldl -fvisibility=hidden -std=c++20 -O2 -fno-math-errno -fPIC -MD -MQ -MF
VkFFT Test: FFT + iFFT C2C Bluestein benchmark in double precision OpenBenchmarking.org Benchmark Score, More Is Better VkFFT 1.2.31 Test: FFT + iFFT C2C Bluestein benchmark in double precision NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 1000 2000 3000 4000 5000 SE +/- 12.55, N = 3 SE +/- 4.51, N = 3 SE +/- 11.35, N = 3 SE +/- 9.84, N = 3 4451 3886 4647 4195 1. (CXX) g++ options: -O3 -lrt
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 16 - Model: ResNet-152 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 16 - Model: ResNet-152 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 40 80 120 160 200 SE +/- 0.29, N = 3 195.40 187.26 194.29 164.14 MIN: 186.09 / MAX: 197.7 MIN: 179.81 / MAX: 188.21 MIN: 182.25 / MAX: 197.39 MIN: 145.67 / MAX: 165.38
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 512 - Model: ResNet-152 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 512 - Model: ResNet-152 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 40 80 120 160 200 SE +/- 1.38, N = 2 SE +/- 0.05, N = 3 SE +/- 0.33, N = 2 195.30 187.51 194.87 164.35 MIN: 182 / MAX: 199.43 MIN: 181.57 / MAX: 188.05 MIN: 180.8 / MAX: 198 MIN: 149.91 / MAX: 166.09
VkFFT Test: FFT + iFFT R2C / C2R OpenBenchmarking.org Benchmark Score, More Is Better VkFFT 1.2.31 Test: FFT + iFFT R2C / C2R NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 12K 24K 36K 48K 60K SE +/- 702.53, N = 15 SE +/- 745.02, N = 13 SE +/- 520.37, N = 3 SE +/- 320.62, N = 3 54794 47097 55446 48418 1. (CXX) g++ options: -O3 -lrt
LuxCoreRender Scene: Orange Juice - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.6 Scene: Orange Juice - Acceleration: GPU NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 3 6 9 12 15 SE +/- 0.07, N = 3 SE +/- 0.03, N = 3 SE +/- 0.00, N = 3 SE +/- 0.01, N = 3 11.72 10.40 11.89 12.14 MIN: 9.6 / MAX: 15.44 MIN: 8.31 / MAX: 13.9 MIN: 9.85 / MAX: 15.88 MIN: 10.24 / MAX: 16.71
Blender Blend File: BMW27 - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 4.0 Blend File: BMW27 - Compute: NVIDIA OptiX NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 2 4 6 8 10 SE +/- 0.06, N = 13 SE +/- 0.01, N = 3 SE +/- 0.02, N = 3 SE +/- 0.06, N = 14 5.57 6.21 5.43 6.31
Blender Blend File: Barbershop - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 4.0 Blend File: Barbershop - Compute: NVIDIA OptiX NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 13 26 39 52 65 SE +/- 0.10, N = 3 SE +/- 0.04, N = 3 SE +/- 0.05, N = 3 SE +/- 0.02, N = 2 51.30 58.44 50.73 54.30
IndigoBench Acceleration: OpenCL GPU - Scene: Bedroom OpenBenchmarking.org M samples/s, More Is Better IndigoBench 4.4 Acceleration: OpenCL GPU - Scene: Bedroom NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 5 10 15 20 25 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 SE +/- 0.03, N = 3 SE +/- 0.01, N = 3 19.80 18.20 20.26 20.96
OctaneBench Total Score OpenBenchmarking.org Score, More Is Better OctaneBench 2020.1 Total Score NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 160 320 480 640 800 720.97 648.00 735.94 674.25
ViennaCL Test: OpenCL BLAS - dGEMV-N OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMV-N NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 50 100 150 200 250 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 210 209 211 187 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
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 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.7205 1.441 2.1615 2.882 3.6025 SE +/- 0.014, N = 3 SE +/- 0.028, N = 3 SE +/- 0.009, N = 3 SE +/- 0.011, N = 3 2.855 3.168 2.854 3.202
VkFFT Test: FFT + iFFT C2C Bluestein in single precision OpenBenchmarking.org Benchmark Score, More Is Better VkFFT 1.2.31 Test: FFT + iFFT C2C Bluestein in single precision NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 3K 6K 9K 12K 15K SE +/- 102.52, N = 3 SE +/- 52.09, N = 3 SE +/- 118.41, N = 3 SE +/- 115.62, N = 3 15166 13714 15125 14205 1. (CXX) g++ options: -O3 -lrt
IndigoBench Acceleration: OpenCL GPU - Scene: Supercar OpenBenchmarking.org M samples/s, More Is Better IndigoBench 4.4 Acceleration: OpenCL GPU - Scene: Supercar NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 12 24 36 48 60 SE +/- 0.03, N = 3 SE +/- 0.03, N = 3 SE +/- 0.02, N = 3 SE +/- 0.03, N = 3 52.81 48.52 53.59 52.01
ViennaCL Test: OpenCL BLAS - sCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - sCOPY NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 80 160 240 320 400 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 SE +/- 0.00, N = 3 SE +/- 1.00, N = 3 334 330 336 363 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dGEMM-TT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-TT NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 30 60 90 120 150 SE +/- 2.08, N = 3 SE +/- 1.20, N = 3 SE +/- 2.08, N = 3 SE +/- 0.88, N = 3 122 118 124 113 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
cl-mem Benchmark: Copy OpenBenchmarking.org GB/s, More Is Better cl-mem 2017-01-13 Benchmark: Copy NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 80 160 240 320 400 SE +/- 0.03, N = 3 SE +/- 0.09, N = 3 SE +/- 0.00, N = 3 SE +/- 0.22, N = 3 331.8 330.3 333.3 360.8 1. (CC) gcc options: -O2 -flto -lOpenCL
VkFFT Test: FFT + iFFT C2C multidimensional in single precision OpenBenchmarking.org Benchmark Score, More Is Better VkFFT 1.2.31 Test: FFT + iFFT C2C multidimensional in single precision NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 11K 22K 33K 44K 55K SE +/- 407.19, N = 15 SE +/- 476.57, N = 5 SE +/- 417.77, N = 15 SE +/- 407.28, N = 15 50299 47212 51528 50856 1. (CXX) g++ options: -O3 -lrt
ViennaCL Test: CPU BLAS - dGEMM-TN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-TN NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 30 60 90 120 150 SE +/- 1.00, N = 2 SE +/- 2.31, N = 3 SE +/- 2.08, N = 3 SE +/- 2.08, N = 3 115 121 125 121 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dGEMM-NN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-NN NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 30 60 90 120 150 SE +/- 4.04, N = 3 SE +/- 1.86, N = 3 SE +/- 1.15, N = 3 SE +/- 1.86, N = 3 119 122 117 113 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
TensorFlow Device: GPU - Batch Size: 1 - Model: AlexNet OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 1 - Model: AlexNet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.22, N = 2 SE +/- 0.16, N = 3 SE +/- 0.06, N = 2 SE +/- 0.20, N = 15 13.92 14.04 14.79 14.45
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 16 - Model: Efficientnet_v2_l OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 16 - Model: Efficientnet_v2_l NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.52, N = 2 SE +/- 0.53, N = 2 103.68 103.45 98.11 MIN: 96.86 / MAX: 105.56 MIN: 95.22 / MAX: 105.88 MIN: 89.88 / MAX: 100.25
ViennaCL Test: OpenCL BLAS - dGEMV-T OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMV-T NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 80 160 240 320 400 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.33, N = 3 389 387 391 374 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 512 - Model: Efficientnet_v2_l OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 512 - Model: Efficientnet_v2_l NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.39, N = 3 SE +/- 0.36, N = 2 SE +/- 0.19, N = 3 103.57 101.43 103.50 99.25 MIN: 95.95 / MAX: 105.54 MIN: 93.27 / MAX: 103.58 MIN: 94.95 / MAX: 105.61 MIN: 91.16 / MAX: 101.18
ViennaCL Test: CPU BLAS - dGEMM-NT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-NT NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 30 60 90 120 150 SE +/- 2.08, N = 3 SE +/- 1.76, N = 3 SE +/- 1.20, N = 3 SE +/- 3.28, N = 3 117 122 118 119 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - sDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - sDOT NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 80 160 240 320 400 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.58, N = 3 370 362 365 376 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 256 - Model: Efficientnet_v2_l OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 256 - Model: Efficientnet_v2_l NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.05, N = 2 SE +/- 0.57, N = 3 103.17 101.24 103.24 99.43 MIN: 95.79 / MAX: 105.15 MIN: 93.33 / MAX: 102.92 MIN: 95.41 / MAX: 104.9 MIN: 90.49 / MAX: 101.97
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 64 - Model: Efficientnet_v2_l OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 64 - Model: Efficientnet_v2_l NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 20 40 60 80 100 SE +/- 1.49, N = 2 SE +/- 0.45, N = 3 SE +/- 0.39, N = 2 SE +/- 0.14, N = 3 102.60 101.55 103.20 99.84 MIN: 79.69 / MAX: 105.28 MIN: 93.44 / MAX: 103.08 MIN: 95.31 / MAX: 105.27 MIN: 92.73 / MAX: 101.46
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 1 - Model: Efficientnet_v2_l OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 1 - Model: Efficientnet_v2_l NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.55, N = 3 SE +/- 0.33, N = 3 106.37 107.59 108.59 105.55 MIN: 97.91 / MAX: 108.16 MIN: 98.77 / MAX: 109.43 MIN: 99.04 / MAX: 110.68 MIN: 91.76 / MAX: 107.42
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 1 - Model: ResNet-152 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 1 - Model: ResNet-152 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 40 80 120 160 200 SE +/- 0.36, N = 3 SE +/- 0.73, N = 3 SE +/- 0.09, N = 2 201.94 198.18 201.19 197.12 MIN: 183.53 / MAX: 206.5 MIN: 181.27 / MAX: 200.06 MIN: 180.79 / MAX: 203.92 MIN: 137.37 / MAX: 198.9
TensorFlow Device: GPU - Batch Size: 1 - Model: VGG-16 OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 1 - Model: VGG-16 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.3105 0.621 0.9315 1.242 1.5525 SE +/- 0.01, N = 2 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 1.35 1.36 1.38 1.38
ViennaCL Test: CPU BLAS - sAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - sAXPY NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 30 60 90 120 150 SE +/- 2.19, N = 3 SE +/- 4.81, N = 3 SE +/- 2.00, N = 3 SE +/- 0.33, N = 3 156 153 156 154 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
Libplacebo Test: av1_grain_lap OpenBenchmarking.org FPS, More Is Better Libplacebo 5.229.1 Test: av1_grain_lap NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 900 1800 2700 3600 4500 SE +/- 5.52, N = 3 SE +/- 66.69, N = 3 SE +/- 21.66, N = 3 SE +/- 12.99, N = 3 4171.00 4103.40 4140.87 4126.89 1. (CXX) g++ options: -lm -pthread -ldl -fvisibility=hidden -std=c++20 -O2 -fno-math-errno -fPIC -MD -MQ -MF
TensorFlow Device: GPU - Batch Size: 16 - Model: AlexNet OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 16 - Model: AlexNet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 7 14 21 28 35 SE +/- 0.17, N = 3 SE +/- 0.08, N = 3 SE +/- 0.07, N = 3 31.59 31.45 31.70 31.98
ViennaCL Test: CPU BLAS - dDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dDOT NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.09, N = 3 SE +/- 0.22, N = 3 SE +/- 0.58, N = 3 SE +/- 0.84, N = 3 96.8 96.7 96.4 95.2 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
TensorFlow Device: GPU - Batch Size: 1 - Model: GoogLeNet OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 1 - Model: GoogLeNet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 3 6 9 12 15 SE +/- 0.17, N = 2 SE +/- 0.10, N = 3 SE +/- 0.30, N = 2 SE +/- 0.07, N = 3 12.62 12.78 12.79 12.82
ViennaCL Test: CPU BLAS - dCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dCOPY NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 16 32 48 64 80 SE +/- 0.32, N = 3 SE +/- 0.25, N = 3 SE +/- 0.74, N = 3 SE +/- 0.72, N = 3 70.8 71.0 71.3 70.2 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
TensorFlow Device: GPU - Batch Size: 512 - Model: AlexNet OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 512 - Model: AlexNet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 8 16 24 32 40 SE +/- 0.02, N = 2 SE +/- 0.03, N = 3 SE +/- 0.09, N = 2 SE +/- 0.01, N = 3 35.10 35.21 35.44 35.58
TensorFlow Device: GPU - Batch Size: 16 - Model: VGG-16 OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 16 - Model: VGG-16 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.3375 0.675 1.0125 1.35 1.6875 SE +/- 0.00, N = 2 SE +/- 0.01, N = 2 SE +/- 0.00, N = 3 1.48 1.50 1.49 1.49
TensorFlow Device: GPU - Batch Size: 256 - Model: AlexNet OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 256 - Model: AlexNet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 TI NVIDIA RTX 3090 8 16 24 32 40 SE +/- 0.01, N = 3 SE +/- 0.07, N = 2 SE +/- 0.07, N = 3 34.16 34.61 34.46
TensorFlow Device: GPU - Batch Size: 32 - Model: GoogLeNet OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 32 - Model: GoogLeNet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.01, N = 2 SE +/- 0.03, N = 3 SE +/- 0.03, N = 3 SE +/- 0.06, N = 3 15.61 15.63 15.81 15.67
ViennaCL Test: CPU BLAS - dAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dAXPY NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.12, N = 3 SE +/- 0.44, N = 3 SE +/- 0.57, N = 3 SE +/- 0.94, N = 3 87.2 86.8 87.3 86.2 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
TensorFlow Device: GPU - Batch Size: 32 - Model: ResNet-50 OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 32 - Model: ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 1.2533 2.5066 3.7599 5.0132 6.2665 SE +/- 0.01, N = 2 SE +/- 0.01, N = 2 SE +/- 0.02, N = 3 SE +/- 0.01, N = 3 5.51 5.55 5.50 5.57
ViennaCL Test: CPU BLAS - dGEMV-N OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMV-N NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 SE +/- 0.88, N = 3 102 103 103 103 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
TensorFlow Device: GPU - Batch Size: 64 - Model: GoogLeNet OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 64 - Model: GoogLeNet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.07, N = 3 SE +/- 0.06, N = 2 SE +/- 0.08, N = 3 15.52 15.54 15.50 15.63
vkpeak fp32-vec4 OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20230730 fp32-vec4 NVIDIA RTX 3090 6K 12K 18K 24K 30K SE +/- 1.51, N = 3 26563.72
ViennaCL Test: CPU BLAS - sCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - sCOPY NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 30 60 90 120 150 SE +/- 1.20, N = 3 SE +/- 1.20, N = 3 SE +/- 0.88, N = 3 SE +/- 1.20, N = 3 132 131 132 132 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
TensorFlow Device: GPU - Batch Size: 64 - Model: ResNet-50 OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 64 - Model: ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 1.2533 2.5066 3.7599 5.0132 6.2665 SE +/- 0.01, N = 2 SE +/- 0.00, N = 3 SE +/- 0.01, N = 2 SE +/- 0.01, N = 3 5.55 5.55 5.53 5.57
TensorFlow Device: GPU - Batch Size: 32 - Model: AlexNet OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 32 - Model: AlexNet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 8 16 24 32 40 SE +/- 0.15, N = 2 SE +/- 0.18, N = 3 SE +/- 0.04, N = 3 SE +/- 0.05, N = 3 33.40 33.32 33.29 33.53
TensorFlow Device: GPU - Batch Size: 1 - Model: ResNet-50 OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 1 - Model: ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.9788 1.9576 2.9364 3.9152 4.894 SE +/- 0.01, N = 3 SE +/- 0.02, N = 2 SE +/- 0.03, N = 3 4.35 4.34 4.32 4.35
TensorFlow Device: GPU - Batch Size: 256 - Model: VGG-16 OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 256 - Model: VGG-16 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.3398 0.6796 1.0194 1.3592 1.699 SE +/- 0.00, N = 3 1.50 1.51
TensorFlow Device: GPU - Batch Size: 64 - Model: VGG-16 OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 64 - Model: VGG-16 NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.3398 0.6796 1.0194 1.3592 1.699 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 1.50 1.50 1.51
TensorFlow Device: GPU - Batch Size: 16 - Model: ResNet-50 OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 16 - Model: ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 1.2353 2.4706 3.7059 4.9412 6.1765 SE +/- 0.00, N = 2 SE +/- 0.02, N = 3 SE +/- 0.01, N = 3 SE +/- 0.00, N = 3 5.46 5.49 5.46 5.49
vkpeak fp32-scalar OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20230730 fp32-scalar NVIDIA RTX 3090 4K 8K 12K 16K 20K SE +/- 36.15, N = 3 20263.13
TensorFlow Device: GPU - Batch Size: 64 - Model: AlexNet OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 64 - Model: AlexNet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 8 16 24 32 40 SE +/- 0.14, N = 3 SE +/- 0.06, N = 3 SE +/- 0.08, N = 3 33.97 33.93 34.06 33.93
vkpeak int16-scalar OpenBenchmarking.org GIOPS, More Is Better vkpeak 20230730 int16-scalar NVIDIA RTX 3090 3K 6K 9K 12K 15K SE +/- 0.21, N = 3 13259.97
vkpeak fp16-scalar OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20230730 fp16-scalar NVIDIA RTX 3090 4K 8K 12K 16K 20K SE +/- 34.71, N = 3 20080.47
vkpeak fp16-vec4 OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20230730 fp16-vec4 NVIDIA RTX 3090 9K 18K 27K 36K 45K SE +/- 69.71, N = 3 39771.97
vkpeak int16-vec4 OpenBenchmarking.org GIOPS, More Is Better vkpeak 20230730 int16-vec4 NVIDIA RTX 3090 4K 8K 12K 16K 20K SE +/- 1.52, N = 3 16331.16
TensorFlow Device: GPU - Batch Size: 16 - Model: GoogLeNet OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 16 - Model: GoogLeNet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.03, N = 3 SE +/- 0.03, N = 3 SE +/- 0.07, N = 3 SE +/- 0.05, N = 3 15.67 15.66 15.69 15.68
vkpeak int32-scalar OpenBenchmarking.org GIOPS, More Is Better vkpeak 20230730 int32-scalar NVIDIA RTX 3090 4K 8K 12K 16K 20K SE +/- 3.21, N = 3 20280.33
vkpeak fp64-vec4 OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20230730 fp64-vec4 NVIDIA RTX 3090 140 280 420 560 700 SE +/- 0.02, N = 3 638.72
vkpeak int32-vec4 OpenBenchmarking.org GIOPS, More Is Better vkpeak 20230730 int32-vec4 NVIDIA RTX 3090 4K 8K 12K 16K 20K SE +/- 2.34, N = 3 19996.92
vkpeak fp64-scalar OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20230730 fp64-scalar NVIDIA RTX 3090 140 280 420 560 700 SE +/- 0.03, N = 3 638.70
TensorFlow Device: GPU - Batch Size: 32 - Model: VGG-16 OpenBenchmarking.org images/sec, More Is Better TensorFlow 2.12 Device: GPU - Batch Size: 32 - Model: VGG-16 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.3375 0.675 1.0125 1.35 1.6875 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 2 SE +/- 0.00, N = 3 1.50 1.50 1.50 1.50
NCNN Target: Vulkan GPU - Model: FastestDet OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: FastestDet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 2 4 6 8 10 SE +/- 0.29, N = 9 SE +/- 0.10, N = 9 SE +/- 0.12, N = 8 SE +/- 2.14, N = 6 2.86 2.34 2.84 6.38 MIN: 2.17 / MAX: 577.17 MIN: 2 / MAX: 3.86 MIN: 2.4 / MAX: 5.07 MIN: 2.14 / MAX: 1476.09 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: vision_transformer OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: vision_transformer NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 200 400 600 800 1000 SE +/- 87.53, N = 9 SE +/- 61.31, N = 9 SE +/- 25.65, N = 9 SE +/- 76.74, N = 6 844.61 281.56 390.18 663.24 MIN: 46.34 / MAX: 1866.93 MIN: 46.48 / MAX: 1913.33 MIN: 46.49 / MAX: 1816.77 MIN: 46.42 / MAX: 1833.21 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: regnety_400m OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: regnety_400m NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 3 6 9 12 15 SE +/- 3.28, N = 9 SE +/- 0.31, N = 8 SE +/- 0.24, N = 9 SE +/- 1.17, N = 6 11.11 6.50 5.97 8.06 MIN: 5.49 / MAX: 4942.19 MIN: 5.52 / MAX: 460.02 MIN: 5.49 / MAX: 7.35 MIN: 5.43 / MAX: 1922.26 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: squeezenet_ssd OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: squeezenet_ssd NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 2 4 6 8 10 SE +/- 1.76, N = 9 SE +/- 0.17, N = 9 SE +/- 0.29, N = 9 SE +/- 1.57, N = 6 6.86 5.27 5.36 6.63 MIN: 4.34 / MAX: 1630.01 MIN: 4.53 / MAX: 7.53 MIN: 4.55 / MAX: 496.3 MIN: 4.43 / MAX: 1636.66 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: yolov4-tiny OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: yolov4-tiny NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 14 28 42 56 70 SE +/- 10.56, N = 9 SE +/- 7.50, N = 9 SE +/- 2.58, N = 9 SE +/- 5.27, N = 6 63.82 25.11 16.47 26.85 MIN: 10.28 / MAX: 858.44 MIN: 10.66 / MAX: 857.35 MIN: 10.61 / MAX: 826.68 MIN: 10.35 / MAX: 853.14 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: resnet50 OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: resnet50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 10 20 30 40 50 SE +/- 14.70, N = 9 SE +/- 0.10, N = 9 SE +/- 4.23, N = 9 SE +/- 11.48, N = 6 46.26 8.24 14.32 27.77 MIN: 7.71 / MAX: 1829.99 MIN: 7.87 / MAX: 9.87 MIN: 7.9 / MAX: 1787.49 MIN: 7.77 / MAX: 1603.33 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: alexnet OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: alexnet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 5.86, N = 9 SE +/- 3.71, N = 9 SE +/- 0.03, N = 9 SE +/- 0.02, N = 6 16.17 9.33 3.74 3.69 MIN: 3.52 / MAX: 436.52 MIN: 3.5 / MAX: 430.03 MIN: 3.61 / MAX: 3.98 MIN: 3.59 / MAX: 7.37 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: resnet18 OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: resnet18 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 3.49, N = 9 SE +/- 3.20, N = 9 SE +/- 1.33, N = 9 SE +/- 6.10, N = 6 8.97 8.58 5.47 17.41 MIN: 3.94 / MAX: 922.04 MIN: 3.98 / MAX: 912.04 MIN: 3.95 / MAX: 726.67 MIN: 4.05 / MAX: 900.27 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: vgg16 OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: vgg16 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 30 60 90 120 150 SE +/- 29.60, N = 9 SE +/- 19.29, N = 9 SE +/- 11.81, N = 9 SE +/- 22.21, N = 6 117.81 54.54 32.05 145.72 MIN: 17.16 / MAX: 647.67 MIN: 17.54 / MAX: 646.66 MIN: 17.34 / MAX: 644.35 MIN: 17.46 / MAX: 648.88 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: googlenet OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: googlenet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 3 6 9 12 15 SE +/- 1.21, N = 9 SE +/- 0.14, N = 9 SE +/- 0.14, N = 9 SE +/- 1.05, N = 6 11.04 6.06 5.87 7.49 MIN: 5.28 / MAX: 1769.19 MIN: 5.33 / MAX: 8.36 MIN: 5.2 / MAX: 6.88 MIN: 5.46 / MAX: 1242.73 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: blazeface OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: blazeface NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.1958 0.3916 0.5874 0.7832 0.979 SE +/- 0.04, N = 9 SE +/- 0.03, N = 9 SE +/- 0.03, N = 9 SE +/- 0.03, N = 6 0.84 0.84 0.81 0.86 MIN: 0.65 / MAX: 4.63 MIN: 0.64 / MAX: 0.96 MIN: 0.61 / MAX: 1.19 MIN: 0.64 / MAX: 3.3 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: efficientnet-b0 OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: efficientnet-b0 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.97, N = 9 SE +/- 0.09, N = 9 SE +/- 0.06, N = 9 SE +/- 9.20, N = 6 5.07 3.46 3.49 13.87 MIN: 3.22 / MAX: 1124.2 MIN: 2.91 / MAX: 3.79 MIN: 3.18 / MAX: 4.03 MIN: 2.86 / MAX: 2218.7 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: mnasnet OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: mnasnet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.9315 1.863 2.7945 3.726 4.6575 SE +/- 1.31, N = 9 SE +/- 0.08, N = 8 SE +/- 0.05, N = 9 SE +/- 0.06, N = 5 3.85 2.22 2.30 2.24 MIN: 1.89 / MAX: 1093.29 MIN: 1.83 / MAX: 2.54 MIN: 2.15 / MAX: 2.58 MIN: 2.07 / MAX: 6.02 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: shufflenet-v2 OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: shufflenet-v2 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 0.9383 1.8766 2.8149 3.7532 4.6915 SE +/- 0.34, N = 8 SE +/- 0.12, N = 7 SE +/- 0.10, N = 8 SE +/- 2.18, N = 6 2.31 2.11 2.03 4.17 MIN: 1.76 / MAX: 421.42 MIN: 1.77 / MAX: 2.53 MIN: 1.84 / MAX: 2.58 MIN: 1.83 / MAX: 1393.33 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU-v3-v3 - Model: mobilenet-v3 OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU-v3-v3 - Model: mobilenet-v3 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 2 4 6 8 10 SE +/- 0.16, N = 9 SE +/- 6.77, N = 8 SE +/- 0.09, N = 9 SE +/- 1.14, N = 6 2.25 8.71 2.09 3.19 MIN: 1.75 / MAX: 343.7 MIN: 1.73 / MAX: 1561.29 MIN: 1.78 / MAX: 2.85 MIN: 1.82 / MAX: 1210.31 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU-v2-v2 - Model: mobilenet-v2 OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU-v2-v2 - Model: mobilenet-v2 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 1.0553 2.1106 3.1659 4.2212 5.2765 SE +/- 0.44, N = 9 SE +/- 2.36, N = 9 SE +/- 0.07, N = 9 SE +/- 0.13, N = 6 3.03 4.69 2.43 2.65 MIN: 2.38 / MAX: 970.87 MIN: 1.91 / MAX: 1305.64 MIN: 2.09 / MAX: 5.8 MIN: 2.23 / MAX: 6.49 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: mobilenet OpenBenchmarking.org ms, Fewer Is Better NCNN 20230517 Target: Vulkan GPU - Model: mobilenet NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 3 6 9 12 15 SE +/- 0.47, N = 9 SE +/- 2.50, N = 9 SE +/- 0.98, N = 9 SE +/- 4.97, N = 6 8.62 10.14 8.43 12.07 MIN: 6.42 / MAX: 1101.3 MIN: 6.53 / MAX: 1509.26 MIN: 6.51 / MAX: 1023.8 MIN: 6.42 / MAX: 1193.34 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
ViennaCL Test: CPU BLAS - dGEMV-T OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMV-T NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 20 40 60 80 100 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 SE +/- 6.30, N = 3 SE +/- 0.33, N = 3 109.0 109.0 102.7 110.0 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - sDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - sDOT NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 40 80 120 160 200 SE +/- 2.73, N = 3 SE +/- 3.76, N = 3 SE +/- 2.40, N = 3 SE +/- 35.40, N = 3 165.0 166.0 168.0 132.1 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
FinanceBench Benchmark: Black-Scholes OpenCL OpenBenchmarking.org ms, Fewer Is Better FinanceBench 2016-07-25 Benchmark: Black-Scholes OpenCL NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 2 4 6 8 10 SE +/- 0.114, N = 15 SE +/- 0.003, N = 3 SE +/- 0.003, N = 3 SE +/- 0.006, N = 3 5.912 6.906 5.226 5.741 1. (CXX) g++ options: -O3 -march=native -fopenmp
LuxCoreRender Scene: DLSC - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.6 Scene: DLSC - Acceleration: GPU NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 4 8 12 16 20 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 SE +/- 0.00, N = 3 SE +/- 1.13, N = 12 13.59 11.74 13.95 12.99 MIN: 12.52 / MAX: 13.84 MIN: 11.35 / MAX: 11.83 MIN: 13.67 / MAX: 14.14 MIN: 0.52 / MAX: 14.69
RealSR-NCNN Scale: 4x - TAA: No OpenBenchmarking.org Seconds, Fewer Is Better RealSR-NCNN 20200818 Scale: 4x - TAA: No NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 2 4 6 8 10 SE +/- 0.150, N = 15 SE +/- 0.006, N = 3 SE +/- 0.039, N = 3 SE +/- 0.016, N = 3 6.323 7.092 5.962 5.556
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 32 - Model: Efficientnet_v2_l OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 32 - Model: Efficientnet_v2_l NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 20 40 60 80 100 SE +/- 6.65, N = 5 SE +/- 0.13, N = 3 102.60 102.90 96.50 99.05 MIN: 94.84 / MAX: 104.25 MIN: 95.98 / MAX: 104.54 MIN: 64.35 / MAX: 104.79 MIN: 91.8 / MAX: 100.69
PyTorch Device: NVIDIA CUDA GPU - Batch Size: 1 - Model: ResNet-50 OpenBenchmarking.org batches/sec, More Is Better PyTorch 2.1 Device: NVIDIA CUDA GPU - Batch Size: 1 - Model: ResNet-50 NVIDIA RTX 4070 SUPER NVIDIA RTX 4070 NVIDIA RTX 4070 TI NVIDIA RTX 3090 120 240 360 480 600 SE +/- 3.09, N = 3 SE +/- 11.16, N = 12 557.73 546.76 535.39 525.12 MIN: 513.63 / MAX: 563.37 MIN: 195.25 / MAX: 556.94 MIN: 428.43 / MAX: 572.99 MIN: 458.54 / MAX: 542.46
Phoronix Test Suite v10.8.5