Distinguishing between NVidia devices and Intel graphics

chiendarret · 12-28-2022, 10:35 AM

Because - as I was told - everything that reaches the monitor comes through the Intel chip, I would like to identify separately the NVidia devices and the Intel chip

Command

Quote:

$ lspci -v | egrep -i --color 'vga|3d|2d'
02:00.0 VGA compatible controller: NVIDIA Corporation GK104 [GeForce GTX 680] (rev a1) (prog-if 00 [VGA controller])

What about the Intel chip?

My question is related to sudden inability of the NVidia devices (in the example above I have manually removed the second Nvidia device) to perform calculations. Possibly the issue reflects problems with the Intel chip.

thanks

mrmazda · 12-28-2022, 03:23 PM

Coming through the Intel IGP to reach the monitor(s) when an NVidia GPU is present usually applies to laptops, but may apply via special configuration on some desktops. On desktops, by default, installation of a discrete GPU of any brand usually disables an Intel IGP, hiding it from lspci and kernel. According to your lspci output, this latter configuration seems to apply. It may be possible to change the behavior via BIOS setup.

You may find inxi -Ga more informative than lspci if your inxi version is recent.

chiendarret · 12-31-2022, 09:28 AM

Thanks. The reason for my question was related to the error reported by the system on trying molecular dynamics (I have checked now with a different case that the problem persists):

Quote:

TCL: Minimizing for 3000 steps
FATAL ERROR: CUDA error cudaStreamSynchronize(stream) in file src/CudaTileListKernel.cu, function buildTileLists, line 1136
on Pe 8 (gig64 device 0 pci 0:2:0): an illegal memory access was encountered
FATAL ERROR: CUDA error cudaStreamSynchronize(stream) in file src/CudaTileListKernel.cu, function buildTileLists, line 1136
on Pe 8 (gig64 device 0 pci 0:2:0): an illegal memory access was encountered
[Partition 0][Node 0] End of program
FATAL ERROR: CUDA error in ComputeBondedCUDA::forceDoneCheck after polling 0 times over 0.005782 s on Pe 8 (gig64 device 0 pci 0:2:0): an illegal memory access was encountered
FATAL ERROR: CUDA error in ComputeBondedCUDA::forceDoneCheck after polling 0 times over 0.005782 s on Pe 8 (gig64 device 0 pci 0:2:0): an illegal memory access was encountered

This error was suggested, long ago on the Software forum, to arise from:

Quote:

1st suggestion: To the best of my understanding, with newer hardware Nvidia graphics devices are supplements to the integrated graphics. Everything that reaches the monitor came through the Intel chip. Cuda is an NVidia thing.
Quote:
FATAL ERROR: CUDA
Pe 2 (gig64 device 0 pci 0:2:0)
Pe 4 (gig64 device 1 pci 0:3:0)
Looks to me like there's two NVidia devices trying to send data through two Intel graphics devices...
Because those addresses would be Intel addresses.
EDIT: Do they exist? the first one should

.

Quote:

2nd suggestion: looks like video driver related issue

Quote:

3rd suggestion:illegal memory access is a software error, not a hardware (memory) problem

That problem arises with both GPUs installed or just one, so that my guess is that the GPU hardware is OK, although I had no other GPU to check that.