CUDA does not recognize NVIDIA GeForce GTX 1080 Ti

I have a NVIDIA GeForce GTX 1080 Ti which I intend to use for deep learning. I have installed the GPU, and ran through the CUDA installation guide. However, I couldn’t verify that the CUDA toolkit were communicating correctly with the CUDA-capable hardware.

In my Graphics/Displays tab of System Information does say:

NVIDIA GeForce GTX 1080 Ti:

  Chipset Model:	NVIDIA GeForce GTX 1080 Ti
  Type:	GPU
  Bus:	PCIe
  PCIe Lane Width:	x4
  VRAM (Total):	11 GB
  Vendor:	NVIDIA (0x10de)
  Device ID:	0x1b06
  Revision ID:	0x00a1
  ROM Revision:	VBIOS
  Metal:	Supported, feature set macOS GPUFamily1 v3

However, in my NVIDIA Driver Manager, ECC tab said: NVIDIA GeForce GTX 1080 Ti (ECC Not Supported).

Here is what the bandwidthTest look like

Last login: Fri Nov  3 13:26:53 on ttys001
/Users/Valuedcustomer/NVIDIA_CUDA-9.0_Samples/bin/x86_64/darwin/release/bandwidthTest ; exit;
Minhs-MacBook-Pro:~ Valuedcustomer$ /Users/Valuedcustomer/NVIDIA_CUDA-9.0_Samples/bin/x86_64/darwin/release/bandwidthTest ; exit;
[CUDA Bandwidth Test] - Starting...
Running on...

Here is what the deviceQuery look like

Last login: Fri Nov  3 13:14:26 on ttys003
/Users/Valuedcustomer/NVIDIA_CUDA-9.0_Samples/bin/x86_64/darwin/release/deviceQuery ; exit;
Minhs-MacBook-Pro:~ Valuedcustomer$ /Users/Valuedcustomer/NVIDIA_CUDA-9.0_Samples/bin/x86_64/darwin/release/deviceQuery ; exit;
/Users/Valuedcustomer/NVIDIA_CUDA-9.0_Samples/bin/x86_64/darwin/release/deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

Any pointer or help would be really appreciated. Thanks.

I have no idea about the rest, but ECC(error correcting code) is RAM/VRAM feature, that’s commonly supported only on professional and server grade hardware (e.i xeons for intel CPU-s and QUADRO/TESLA for NVIDIA GPU-s).
As a side note, I also use GTX 1080 ti for GPGPU and had never had any issues, other than bad performance with double precision floating points, thing I generally try to avoid using anyway, but depending on the application, something like lack of some pro-only features, or somewhat reduced memory bandwidth might hinder your experience.