I want to know the number of cores in Colab's assigned GPU. I have used !nvidia-smi -q but it is showing all other details but not showing the number of cores of the GPU.
I have researched about it but didn't find it. Any help regarding this!!
Asked
Active
Viewed 3,295 times
0
Saddam
- 23
- 1
- 5
1 Answers
0
The easy way out would be to run the !nvidia-smi command to get all the GPU information. There seem to be 2 possible options on the cards that you will get after that - K80 or T4, the K80 has 4992 CUDA cores while the T4 has 2560 CUDA cores (Found this using Google).
If you still need scripts to find out the number of cores though, you can find those scripts in this answer (python) and this answer (C). I would still suggest the much simpler way of looking up the CUDA cores for that particular model.
On a sidenote, there are instances where users only get 5% of the GPU ram accessible to them. I have not faced this problem but you should still check the Volatile GPU Util that is shown in the !nvidia-smi command (0% is what should be expected when nothing is running).
veedata
- 1,048
- 1
- 9
- 15