NVIDIA vs. Radeon OpenCL Linux GPU Compute

NVIDIA Linux GPU computing benchmarks by Michael Larabel for a future article.

HTML result view exported from: https://openbenchmarking.org/result/1901304-SP-1901305SP07&rdt&grs.

clpeak

OpenCL Test: Kernel Latency

LeelaChessZero

System Power Consumption Monitor

Darktable

Test: Boat - Acceleration: OpenCL

LeelaChessZero

Backend: OpenCL

PlaidML

System Power Consumption Monitor

LuxMark

System Power Consumption Monitor

PlaidML

System Power Consumption Monitor

cl-mem

System Power Consumption Monitor

PlaidML

System Power Consumption Monitor

PlaidML

System Power Consumption Monitor

PlaidML

System Power Consumption Monitor

PlaidML

System Power Consumption Monitor

clpeak

System Power Consumption Monitor

SHOC Scalable HeterOgeneous Computing

System Power Consumption Monitor

clpeak

System Power Consumption Monitor

PlaidML

System Power Consumption Monitor

PlaidML

System Power Consumption Monitor

PlaidML

System Power Consumption Monitor

cl-mem

System Power Consumption Monitor

SHOC Scalable HeterOgeneous Computing

System Power Consumption Monitor

cl-mem

System Power Consumption Monitor

Darktable

Test: Server Room - Acceleration: OpenCL

PlaidML

System Power Consumption Monitor

SHOC Scalable HeterOgeneous Computing

Target: OpenCL - Benchmark: Texture Read Bandwidth

clpeak

System Power Consumption Monitor

clpeak

OpenCL Test: Double-Precision Double

SHOC Scalable HeterOgeneous Computing

Target: OpenCL - Benchmark: FFT SP

JuliaGPU

System Power Consumption Monitor

LeelaChessZero

Backend: OpenCL

SHOC Scalable HeterOgeneous Computing

Target: OpenCL - Benchmark: MD5 Hash

cl-mem

Benchmark: Read

PlaidML

System Power Consumption Monitor

SHOC Scalable HeterOgeneous Computing

Target: OpenCL - Benchmark: Texture Read Bandwidth

PlaidML

FP16: No - Mode: Inference - Network: VGG16 - Device: OpenCL

PlaidML

FP16: Yes - Mode: Inference - Network: VGG16 - Device: OpenCL

PlaidML

FP16: No - Mode: Inference - Network: ResNet 50 - Device: OpenCL

clpeak

System Power Consumption Monitor

Darktable

System Power Consumption Monitor

Darktable

System Power Consumption Monitor

PlaidML

FP16: Yes - Mode: Inference - Network: VGG16 - Device: OpenCL

PlaidML

FP16: No - Mode: Training - Network: Mobilenet - Device: OpenCL

Darktable

System Power Consumption Monitor

LuxMark

OpenCL Device: GPU - Scene: Luxball HDR

cl-mem

Benchmark: Read

PlaidML

FP16: Yes - Mode: Inference - Network: ResNet 50 - Device: OpenCL

clpeak

OpenCL Test: Global Memory Bandwidth

clpeak

OpenCL Test: Transfer Bandwidth enqueueWriteBuffer

PlaidML

FP16: No - Mode: Inference - Network: ResNet 50 - Device: OpenCL

cl-mem

Benchmark: Write

cl-mem

Benchmark: Copy

PlaidML

FP16: No - Mode: Inference - Network: VGG16 - Device: OpenCL

PlaidML

FP16: No - Mode: Inference - Network: IMDB LSTM - Device: OpenCL

PlaidML

FP16: No - Mode: Training - Network: Mobilenet - Device: OpenCL

PlaidML

FP16: No - Mode: Inference - Network: Inception V3 - Device: OpenCL

PlaidML

FP16: Yes - Mode: Inference - Network: ResNet 50 - Device: OpenCL

PlaidML

FP16: No - Mode: Inference - Network: Mobilenet - Device: OpenCL

PlaidML

FP16: No - Mode: Inference - Network: Mobilenet - Device: OpenCL

JuliaGPU

OpenCL Device: GPU

LuxMark

GPU Temperature Monitor

PlaidML

FP16: No - Mode: Inference - Network: Inception V3 - Device: OpenCL

SHOC Scalable HeterOgeneous Computing

System Power Consumption Monitor

PlaidML

FP16: No - Mode: Inference - Network: IMDB LSTM - Device: OpenCL

PlaidML

FP16: Yes - Mode: Inference - Network: Inception V3 - Device: OpenCL

LeelaChessZero

GPU Temperature Monitor

PlaidML

FP16: Yes - Mode: Inference - Network: Inception V3 - Device: OpenCL

clpeak

OpenCL Test: Single-Precision Float

cl-mem

Benchmark: Write

cl-mem

Benchmark: Copy

clpeak

System Power Consumption Monitor

clpeak

System Power Consumption Monitor

PlaidML

GPU Temperature Monitor