TRT4.0 at 1080TI vs TITAN V

Hello,
I use 1080TI and TITAN V to test this NVIDIA sample code(https://github.com/NVIDIA-Jetson/tf_to_trt_image_classification), TRT 4.0 coat time has same result.

is anyone can help me to explian it?
Thanks a lot.

Hello, can you clarify what you mean by “TRT 4.0 coat time has same result”?

Hello, i use same Tensorflow official model to tested TRT 4.0 have same cost time in 1080TI and TITAN V, i dont know why.

can you give the specific model you used and performance metrics you are seeing?

thank you,
NVIDIA Enterprise Support

Hello, all model download from https://github.com/NVIDIA-Jetson/tf_to_trt_image_classification/blob/master/scripts/download_models.sh

Hello,

when running benchmark python scripts/test_trt.py, the timing result will be stored at data/test_output_trt.txt. Can you share the file with us? Is this the performance metric you are saying is the same between T080Ti and TitanV?

Hello,

Yes, that i said issue.

  1. the float result in TitanV:
    data/plans/mobilenet_v1_0p25_128.plan 0.710423
    data/plans/resnet_v1_50.plan 2.52737
    data/plans/mobilenet_v1_1p0_224.plan 1.22464
    data/plans/inception_v2.plan 2.37085
    data/plans/inception_v3.plan 5.46946
    data/plans/resnet_v2_152.plan 8.41234
    data/plans/inception_v1.plan 1.44983
    data/plans/resnet_v1_152.plan 6.97273
    data/plans/inception_v4.plan 11.2623
    data/plans/resnet_v1_101.plan 4.89694
    data/plans/inception_resnet_v2.plan 9.92916
    data/plans/resnet_v2_50.plan 2.97753
    data/plans/vgg_16.plan 2.78958
    data/plans/resnet_v2_101.plan 5.90988
    data/plans/mobilenet_v1_0p5_160.plan 0.774903

  2. the half result in TitanV:
    data/plans/mobilenet_v1_0p25_128.plan 0.689513
    data/plans/resnet_v1_50.plan 2.53846
    data/plans/mobilenet_v1_1p0_224.plan 1.22564
    data/plans/inception_v2.plan 2.35924
    data/plans/inception_v3.plan 5.53435
    data/plans/resnet_v2_152.plan 8.51935
    data/plans/inception_v1.plan 1.42204
    data/plans/resnet_v1_152.plan 6.95254
    data/plans/inception_v4.plan 11.2141
    data/plans/resnet_v1_101.plan 4.94806
    data/plans/inception_resnet_v2.plan 9.8948
    data/plans/resnet_v2_50.plan 2.96348
    data/plans/vgg_16.plan 2.83347
    data/plans/resnet_v2_101.plan 5.84114
    data/plans/mobilenet_v1_0p5_160.plan 0.774131

  3. the float result in 1080TI:
    data/plans/mobilenet_v1_0p25_128.plan 0.559485
    data/plans/resnet_v1_50.plan 3.02457
    data/plans/mobilenet_v1_1p0_224.plan 1.30994
    data/plans/inception_v2.plan 2.16442
    data/plans/inception_v3.plan 5.91856
    data/plans/resnet_v2_152.plan 9.1956
    data/plans/inception_v1.plan 1.19595
    data/plans/resnet_v1_152.plan 6.38435
    data/plans/inception_v4.plan 11.187
    data/plans/resnet_v1_101.plan 4.54257
    data/plans/inception_resnet_v2.plan 11.6828
    data/plans/resnet_v2_50.plan 3.61238
    data/plans/vgg_16.plan 4.06281