Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

The memory usage displayed by nvtop is different from that of nvidia-smi #322

Open
Kelvinlby opened this issue Oct 19, 2024 · 4 comments
Open

Comments

@Kelvinlby
Copy link

I noticed that on my platform (Ubuntu 24.04, RTX3060), when it comes to VRAM utilization, nvtop gives a different number to nvidia-smi
Screenshot from 2024-10-19 14-22-13
Screenshot from 2024-10-19 14-22-16

@Kelvinlby
Copy link
Author

I noticed this when I was writing a C program which needs to get VRAM usage of Nvidia GPUs. I used the following code and get the same result as what nvidia-smi shows. However, I'm not sure whether it takes all VRAM utilization into account and which program (nvtop/nvidia-smi) shows the real utilization.

By the way, in most cases nvtop shows larger VRAM utilization than nvidia-smi.

UINT32 nvidiaUsedVram(void) {
    size_t freeMemory, totalMemory;
    cudaMemGetInfo(&freeMemory, &totalMemory);
    return (totalMemory - freeMemory) / (1024 * 1024);
}

@canytam-krystal
Copy link

I face the same problem. After my os upgraded to 24.04, the dispaly of nvtop is no longer correct even I update the nvtop to 3.0.2-1.

@Syllo
Copy link
Owner

Syllo commented Nov 3, 2024

Hard to say which one is closer to what we can expect as "allocated memory". Nvtop uses the result of nvmlDeviceGetMemoryInfo_v2 coming directly from the driver!

The value nvtop displays is the used memory: Allocated device memory (in bytes).
Perhaps nvidia-smi subtracts the reserved memory? I'm not sure.

@Kelvinlby
Copy link
Author

Kelvinlby commented Nov 4, 2024

Hard to say which one is closer to what we can expect as "allocated memory". Nvtop uses the result of nvmlDeviceGetMemoryInfo_v2 coming directly from the driver!

The value nvtop displays is the used memory: Allocated device memory (in bytes). Perhaps nvidia-smi subtracts the reserved memory? I'm not sure.

Not sure either, since both methods are offered by nvidia.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants