Titan V slower than 1080ti tensorflow:18.08-py3 and 396.54 drivers

Hi Tom,

I used GPUZ 2.17.0 to monitor the GPU clock and memory clock of Titan V.
After applying “nvidia-smi --cuda-clocks=OVERRIDE”, the GPU clock speed could reach 1792MHz and the memory clock was 850MHz。But the memory clock was only 425MHz in TCC mode.

However, the results of bandwidthTest of two modes were very close.

# TCC mode
bandwidthTest --dtod

[CUDA Bandwidth Test] - Starting...
Running on...

 Device 0: TITAN V
 Quick Mode

 Device to Device Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     548561.6

Result = PASS

NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.
# WDDM
bandwidthTest --dtod
[CUDA Bandwidth Test] - Starting...
Running on...

 Device 0: TITAN V
 Quick Mode

 Device to Device Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     548914.5

Result = PASS

NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.

OS: Windows 10
Driver: 419.35