Benchmarks

From Upscale Wiki
Revision as of 17:49, 19 August 2019 by Alsa (talk | contribs) (→‎BasicSR (Training))

(diff) ← Older revision | Approved revision (diff) | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

BasicSR (Training)

GPU CPU OS user CUDA Version cudnn Version Driver Version Pytorch Version BasicSR Version Scale model architecture CPU threads (n_workers) Tile size Batch size OTF Time per 1000 iterations
GTX 1080 TI 11Gb (Asus OC version) Intel 6700K Arch Linux alsa 10.1.168-4 7.6.1.34-1 430.40-2 1.1.0-8 (python-pytorch-cuda) Victorc's Fork #754b476 4x PPON (Phase 1) 8 192 18 yes 3648s (01h, 00min, 48s)
GTX 1080 TI 11Gb (Asus OC version) Intel 6700K Arch Linux alsa 10.1.168-4 7.6.2.24-1 430.40-5 1.1.0-8 (python-pytorch-opt-cuda) Victorc's Fork #754b476 4x PPON (Phase 1) 8 192 18 yes 3534s (00h, 58min, 54s)
GTX 1080 TI 11Gb (Asus OC version) Intel 6700K Arch Linux alsa 10.1.168-4 7.6.2.24-1 430.40-5 1.1.0-8 (python-pytorch-opt-cuda) Victorc's Fork #754b476 4x PPON (Phase 2) 8 128 10 yes 260s (00h, 04min, 20s)
Tesla K80 11Gb Xeon CPU @ 2.3 GHz Google Colaboratory, GPU Runtime (Ubuntu) BoxDroppingManApe 10.0.130-10 ? 410.79 1.1.0 Victorc's Fork 1x ESRGAN 4 128 2 no 2721s (00h, 45min, 21s)

ESRGAN

GPU (or CPU) OS CUDA Version cudnn Version Driver Version Pytorch Version ESRGAN Version Scale Image dimensions Time needed for image
sample sample sample sample sample sample sample 4x sample 0.0s (00h, 00min, 00s, 000ms)
CPU (example-manufacturer Model) sample sample - - sample sample 4x sample 0.0s (00h, 00min, 00s, 000ms)