TU116 still shows Tensor Cores

[Wunkolo]

This is in relation to d3aaf7cfe507acf12557db5fad01821079576a15.

My Nvidia Geforce GTX 1660 Ti still seems to show that I have 192 Tensor Cores despite being a TU116 chip. This PR can sustain the dialog of making sure it handles this particular card since I personally have one as in one of my headless linux machines.

% gpufetch -V
gpufetch v0.22
% gpufetch
[NOTE]: gpufetch is in beta. The provided information may be incomplete or wrong.
If you want to help to improve gpufetch, please compare the output of the program
with a reliable source which you know is right (e.g, techpowerup.com) and report
any inconsistencies to https://github.com/Dr-Noob/gpufetch/issues

                  MMMMMMMMMMMMMMMMMMMMMMMMMMMMMM  Name:                   NVIDIA GeForce GTX 1660 Ti
                  MMMMMMMMMMMMMMMMMMMMMMMMMMMMMM  GPU processor:          TU116
                .::   'MMMMMMMMMMMMMMMMMMMMMMMMM  Microarchitecture:      Turing (7.5)
           ccllooo;:;.       ;MMMMMMMMMMMMMMMMMM  Technology:             12nm
       cloc       :ooollcc:     :MMMMMMMMMMMMMMM  Max Frequency:          1770 MHz
    cloc      :ccl;      lolc,     ;MMMMMMMMMMMM  SMs:                    24
.cloo:    :clo    ;c:      .ool;     MMMMMMMMMMM  Cores/SM:               64
  ooo:    ooo     :ool,  .cloo.    ;lMMMMMMMMMMM  CUDA Cores:             1536
   ooo:    ooc    :ooooccooo.    :MMMM  lMMMMMMM  Tensor Cores:           192
     ooc.   ool:  :oooooo'    ,cloo.        MMMM  Memory:                 6 GiB GDDR6
      ool:.    olc:       .:cloo.          :MMMM  Memory frequency:       1500 MHz
         olc,     ;:::cccloo.          :MMMMMMMM  Bus width:              192 bit
            olcc::;              ,:ccloMMMMMMMMM  L2 Size:                2 MiB
                  :......oMMMMMMMMMMMMMMMMMMMMMM  Peak Performance:       5.44 TFLOP/s
                  :lllMMMMMMMMMMMMMMMMMMMMMMMMMM  Peak Performance (MMA): 0.00 MFLOP/s

[Wunkolo]

Ah disregard this. I think I'm assuming differently from that commit.

[Dr-Noob]

No no, you are totally right, the output is still wrong. I have committed a fix, can you give it a try?

[Wunkolo]

I assumed the commit was just to ensure that the count was zero and not necessarily removing it from the output entirely. But if was intended to remove it entirely, then I can test again. Here's my output now:

% gpufetch
[NOTE]: gpufetch is in beta. The provided information may be incomplete or wrong.
If you want to help to improve gpufetch, please compare the output of the program
with a reliable source which you know is right (e.g, techpowerup.com) and report
any inconsistencies to https://github.com/Dr-Noob/gpufetch/issues

                  MMMMMMMMMMMMMMMMMMMMMMMMMMMMMM
                  MMMMMMMMMMMMMMMMMMMMMMMMMMMMMM  Name:              NVIDIA GeForce GTX 1660 Ti
                .::   'MMMMMMMMMMMMMMMMMMMMMMMMM  GPU processor:     TU116
           ccllooo;:;.       ;MMMMMMMMMMMMMMMMMM  Microarchitecture: Turing (7.5)
       cloc       :ooollcc:     :MMMMMMMMMMMMMMM  Technology:        12nm
    cloc      :ccl;      lolc,     ;MMMMMMMMMMMM  Max Frequency:     1770 MHz
.cloo:    :clo    ;c:      .ool;     MMMMMMMMMMM  SMs:               24
  ooo:    ooo     :ool,  .cloo.    ;lMMMMMMMMMMM  Cores/SM:          64
   ooo:    ooc    :ooooccooo.    :MMMM  lMMMMMMM  CUDA Cores:        1536
     ooc.   ool:  :oooooo'    ,cloo.        MMMM  Memory:            6 GiB GDDR6
      ool:.    olc:       .:cloo.          :MMMM  Memory frequency:  1500 MHz
         olc,     ;:::cccloo.          :MMMMMMMM  Bus width:         192 bit
            olcc::;              ,:ccloMMMMMMMMM  L2 Size:           2 MiB
                  :......oMMMMMMMMMMMMMMMMMMMMMM  Peak Performance:  5.44 TFLOP/s
                  :lllMMMMMMMMMMMMMMMMMMMMMMMMMM

Seems to work now I believe.

[Dr-Noob]

Yeah, now it does not show tensor cores nor peak performance. Good to know!

Closing