Join Log in
Score:0
Louis avatar Ubuntu

Ubuntu inconsistant GPU informations

cn flag
Louis

I have been trying for a few hours to install CUDA for tensorflow-gpu. I felt like the thing was up and runnning but I had an error when fitting a simple CNN. I checked my GPU as it seemed to be an out-of-memory error.

nvidia-smi returns :


+---------------------------------------------+
| NVIDIA-SMI 495.29.05    Driver Version: 495.29.05    CUDA Version: 11.5     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  On   | 00000000:01:00.0  On |                  N/A |
| N/A   38C    P8    12W /  N/A |    169MiB /  5938MiB |      3%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

i am on ubuntu 20.04 LTS, on the lattest nvidia-driver-495, and a RTX3060.`

my rig :

enter image description here

sudo lshw -C display returns:

*-display                 
      description: VGA compatible controller
      produit: NVIDIA Corporation
      fabricant: NVIDIA Corporation
      identifiant matériel: 0
      information bus: pci@0000:01:00.0
      version: a1
      bits: 64 bits
      horloge: 33MHz
      fonctionnalités: pm msi pciexpress vga_controller bus_master cap_list rom
      configuration : driver=nvidia latency=0
      ressources : mémoireE/S:fa0-f9f mémoireE/S:fc0-fbf irq:95 mémoire:d0000000-d0ffffff mémoire:fa00000000-fbffffffff mémoire:fc00000000-fc01ffffff portE/S:2000(taille=128) mémoire:d1080000-d10fffff

this tells me my GPU has 6Go of RAM. But my RTX actually has 16G.

Does anyone know where this may come from?

thanks!

65
0 + 0
gpu
Terrance avatar
id flag
Terrance
What is the output of `sudo lshw -C display`? Also, looking at the specs from NVIDIA the RTX3060 comes in either 8GB or 12GB. Not seeing a 16GB version. https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3060-3060ti/
0
Reply
Louis avatar
cn flag
Louis
i edited the post with the `sudo lshw -C display` .
0
Reply
Louis avatar
cn flag
Louis
I also added a pic of my rig. The RTX does actually exist in 16GB version.
0
Reply
Hannu avatar
ca flag
Hannu
Your laptop has 16GB, your RTX 3060 most likely have a separate 6GB dedicated special - possibly higher speed - RAM (note: not the same memory)
0
Reply
Nmath avatar
ng flag
Nmath
What is the problem? The output you posted shows Nvidia drivers and CUDA installed and only 3% of memory in use. This output does not support your narrative of the problem. Please edit your question and include more details including clear reproducible details of the steps you've taken so far
1
Reply
Terrance avatar
id flag
Terrance
Also, how did you install the `nvidia-driver-495` driver? As far as I can see it is in the `proposed` repository meaning that it might contain bugs as it is not officially released to the Ubuntu repos yet. You might have better luck installing the `nvidia-driver-470` since it is the official released version to the Ubuntu repos. Also, checking https://www.notebookcheck.net/NVIDIA-GeForce-RTX-3060-Mobile-GPU-Benchmarks-and-Specs.497453.0.html and https://www.nvidia.com/en-us/geforce/gaming-laptops/ are showing that the mobile version of the RTX 3060 is only 6GB. The 3080 will go to 16GB.
0
Reply
Louis avatar
cn flag
Louis
the problem is that my GPU is supposed have 16Go of ram and only 6 are displayed. Is this normal?
0
Reply
Louis avatar
cn flag
Louis
this driver was installed with CUDA. i will try to re select the 470 and see how it turn out. And yes, sorry for my misunderstanding between RAM and GPU memory.
0
Reply
ธงชาติไทย
Elon Musk
I sit in a Tesla and translated this thread with Ai:

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.