NVIDIA GPU Compute NVIDIA GPU Compute

A collection of test profiles that run well on NVIDIA GPU systems with CUDA / proprietary driver stack. Other deprecated / less interesting / older tests not included but this test suite is intended to serve as guidance for current interesting NVIDIA GPU compute benchmarking albeit not exhaustive of what is available via Phoronix Test Suite / OpenBenchmarking.org.

See how your system performs with this suite using the Phoronix Test Suite. It's as easy as running the phoronix-test-suite benchmark nvidia-gpu-compute command..

Tests In This Suite

  • ArrayFire

  •         Test: Conjugate Gradient OpenCL
  • Betsy GPU Compressor

  •         Codec: ETC2 RGB - Quality: Highest
  •         Codec: ETC1 - Quality: Highest
  • Blender

  •         Blend File: Fishy Cat - Compute: NVIDIA OptiX
  •         Blend File: Barbershop - Compute: CUDA
  •         Blend File: BMW27 - Compute: NVIDIA OptiX
  •         Blend File: Classroom - Compute: CUDA
  •         Blend File: Classroom - Compute: NVIDIA OptiX
  •         Blend File: Fishy Cat - Compute: CUDA
  •         Blend File: Pabellon Barcelona - Compute: CUDA
  •         Blend File: Pabellon Barcelona - Compute: NVIDIA OptiX
  •         Blend File: Barbershop - Compute: NVIDIA OptiX
  •         Blend File: BMW27 - Compute: CUDA
  • Caffe

  •         Model: GoogleNet - Acceleration: NVIDIA CUDA - Iterations: 200
  •         Model: AlexNet - Acceleration: NVIDIA CUDA - Iterations: 200
  •         Model: GoogleNet - Acceleration: NVIDIA CUDA - Iterations: 1000
  •         Model: AlexNet - Acceleration: NVIDIA CUDA - Iterations: 1000
  •         Model: GoogleNet - Acceleration: NVIDIA CUDA - Iterations: 100
  •         Model: AlexNet - Acceleration: NVIDIA CUDA - Iterations: 100
  • Chaos Group V-RAY

  •         Mode: NVIDIA RTX GPU
  •         Mode: NVIDIA CUDA GPU
  • cl-mem

  •         Benchmark: Read
  •         Benchmark: Write
  •         Benchmark: Copy
  • clpeak

  •         OpenCL Test: Global Memory Bandwidth
  •         OpenCL Test: Single-Precision Float
  •         OpenCL Test: Double-Precision Double
  •         OpenCL Test: Integer Compute INT
  • FAHBench

  • FinanceBench

  •         Benchmark: Black-Scholes OpenCL
  • GROMACS

  • Hashcat

  •         Benchmark: MD5
  •         Benchmark: SHA1
  •         Benchmark: TrueCrypt RIPEMD160 + XTS
  •         Benchmark: 7-Zip
  •         Benchmark: SHA-512
  • IndigoBench

  •         Acceleration: OpenCL GPU - Scene: Bedroom
  •         Acceleration: OpenCL GPU - Scene: Supercar
  • LeelaChessZero

  •         Backend: OpenCL
  • Libplacebo

  • LuxCoreRender

  •         Scene: LuxCore Benchmark - Acceleration: GPU
  •         Scene: DLSC - Acceleration: GPU
  •         Scene: Rainbow Colors and Prism - Acceleration: GPU
  •         Scene: Orange Juice - Acceleration: GPU
  •         Scene: Danish Mood - Acceleration: GPU
  • MandelGPU

  •         OpenCL Device: GPU
  • Mixbench

  •         Backend: OpenCL - Benchmark: Single Precision
  •         Backend: OpenCL - Benchmark: Double Precision
  •         Backend: OpenCL - Benchmark: Integer
  •         Backend: NVIDIA CUDA - Benchmark: Single Precision
  •         Backend: NVIDIA CUDA - Benchmark: Double Precision
  •         Backend: NVIDIA CUDA - Benchmark: Half Precision
  •         Backend: NVIDIA CUDA - Benchmark: Integer
  • NAMD CUDA

  • NCNN

  •         Target: Vulkan GPU
  • NeatBench

  •         Acceleration: GPU
  • OctaneBench

  •         Total Score
  • PlaidML

  •         FP16: Yes - Mode: Inference - Network: Mobilenet - Device: OpenCL
  •         FP16: No - Mode: Training - Network: Mobilenet - Device: OpenCL
  •         FP16: No - Mode: Inference - Network: Mobilenet - Device: OpenCL
  •         FP16: No - Mode: Inference - Network: DenseNet 201 - Device: OpenCL
  •         FP16: No - Mode: Inference - Network: IMDB LSTM - Device: OpenCL
  • RealSR-NCNN

  •         Scale: 4x - TAA: Yes
  •         Scale: 4x - TAA: No
  • RedShift Demo

  • Rodinia

  •         Test: OpenCL Particle Filter
  • SHOC Scalable HeterOgeneous Computing

  •         Target: OpenCL - Benchmark: Bus Speed Readback
  •         Target: OpenCL - Benchmark: Bus Speed Download
  •         Target: OpenCL - Benchmark: FFT SP
  •         Target: OpenCL - Benchmark: Triad
  •         Target: OpenCL - Benchmark: Reduction
  •         Target: OpenCL - Benchmark: MD5 Hash
  •         Target: OpenCL - Benchmark: Max SP Flops
  •         Target: OpenCL - Benchmark: Texture Read Bandwidth
  •         Target: OpenCL - Benchmark: S3D
  •         Target: OpenCL - Benchmark: GEMM SGEMM_N
  • ViennaCL

  •         Test: OpenCL BLAS
  •         Test: CPU BLAS
  • VkFFT

  • vkpeak

  • VkResample

  •         Upscale: 2x - Precision: Double
  •         Upscale: 2x - Precision: Single
  • Waifu2x-NCNN Vulkan

  •         Scale: 2x - Denoise: 3 - TAA: Yes
  •         Scale: 2x - Denoise: 3 - TAA: No

Revision History Revision History

pts/nvidia-gpu-compute-1.1.1     Thu, 15 Apr 2021 17:07:05 GMT
Switch luxcorerender-cl to luxcorerender now that LuxCoreRender unified has CPU+GPU support in test profile.

pts/nvidia-gpu-compute-1.1.0     Sun, 28 Mar 2021 16:04:12 GMT
Add ViennaCL, indigobench, SHOC, other updates

pts/nvidia-gpu-compute-1.0.7     Wed, 16 Dec 2020 17:17:54 GMT
Add vulkan-compute to the suite rather than including the Vulkan benchmarks individually.

pts/nvidia-gpu-compute-1.0.6     Tue, 20 Oct 2020 14:12:29 GMT
Fix for NCNN argument.

pts/nvidia-gpu-compute-1.0.5     Tue, 20 Oct 2020 14:08:59 GMT
Add waifu2x-ncnn test profile.

pts/nvidia-gpu-compute-1.0.4     Wed, 07 Oct 2020 07:05:46 GMT
Add NCNN and RealSR-NCNN to suite.

pts/nvidia-gpu-compute-1.0.3     Sun, 04 Oct 2020 13:03:09 GMT
Add latest tests to suite.

pts/nvidia-gpu-compute-1.0.2     Thu, 17 Sep 2020 18:25:32 GMT
Add Hashcat

pts/nvidia-gpu-compute-1.0.1     Sun, 13 Sep 2020 18:48:54 GMT
Add Blender tests.

pts/nvidia-gpu-compute-1.0.0     Sun, 13 Sep 2020 18:43:48 GMT
Initial commit for showing tests known to be in good shape for NVIDIA GPU usage.