GTX 1050 ti CUDA API version

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi there, I’m trying to install CUDA 9.1 on ubuntu 16.04 and have been encountering some issues. It seems that cuddn and tensorRT fail to install due to some version mismatch. Running ./deviceQuery after installing CUDA gives me this.

Device 0: "GeForce GTX 1050 Ti"
  CUDA Driver Version / Runtime Version          9.1 / 9.1
  CUDA Capability Major/Minor version number:    2.1
  Total amount of global memory:                 4532 MBytes (4752408576 bytes)

I would assume that the CUDA Capability Major/Minor version number would be 6.1 and not 2.1. I got the card from an Ebay seller that ships cards from China. I was trying to inspect information about the generic manufacturer, but I got this:

lspci -knn | grep VGA -A1
04:00.0 VGA compatible controller [0300]: NVIDIA Corporation Device [10de:1c82] (rev a1)
	Subsystem: NVIDIA Corporation Device [10de:089c]

Can I please get any information on whether this information should be correct for a GTX 1050 TI ?
All the best,

2

It should report 6.1 for compute capability.

1050 Ti typical deviceQuery output looks like this:

[url]https://datafireball.com/2016/11/30/nvidia-geforce-gtx-1050-ti/[/url]

I’m not aware of any 1050Ti with greater than 4096 MBytes total global memory.

[url]https://www.geforce.com/hardware/desktop-gpus/geforce-gtx-1050-ti/specifications[/url]

The PCI device ID and vendor ID appear to be correct:

[url]https://pci-ids.ucw.cz/read/PC/10de/1c82[/url]

but a shady vendor may be able to spoof those.