> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nikaplanet.com/llms.txt
> Use this file to discover all available pages before exploring further.

# GDAL and CUDA Installation

> GDAL and CUDA are pre-installed on all ephemeral VM kernels for geospatial analysis and GPU acceleration

GDAL and CUDA are pre-installed on all ephemeral VM kernels that your notebooks have direct access to. No manual installation required.

## Pre-Installed Software

### Available Configurations

All VM kernels come with:

* **GDAL 3.9.1**: Geospatial Data Abstraction Library for raster and vector data
* **CUDA 12.2**: NVIDIA CUDA for GPU acceleration
* **Python 3.10/3.11**: Multiple Python versions available

### Kernel Selection

When starting your machine, you can choose from:

* **Python 3.10, GDAL 3.9.1 CUDA 12.2**
* **Python 3.11, GDAL 3.9.1 CUDA 12.2**
* **Python 3.10, GDAL 3.9.1 CUDA 12**

<img src="https://mintcdn.com/nika/pMnlw4L-PpXjCsj1/guides/data-analysis/gdal-cuda-installation.jpg?fit=max&auto=format&n=pMnlw4L-PpXjCsj1&q=85&s=fe4b4ab0f42fa576d20f40c7070b042f" alt="GDAL CUDA Pre-installed" width="343" height="216" data-path="guides/data-analysis/gdal-cuda-installation.jpg" />

## Ready to Use

### No Installation Required

* **Pre-configured**: All geospatial libraries are ready to use
* **GPU Ready**: CUDA drivers and libraries are installed
* **Immediate Access**: Start analyzing geospatial data right away

### Direct Access

* **Workspace Integration**: Direct access to workspace files
* **Large Dataset Support**: Optimized for TB-PB sized geospatial data
* **Performance Optimized**: GPU acceleration for raster processing

## Usage Examples

### GDAL Operations

```python theme={null}
import rasterio
import geopandas as gpd

# Read raster data
with rasterio.open('/workspace/data/satellite.tif') as src:
    data = src.read()

# Read vector data
gdf = gpd.read_file('/workspace/data/shapefile.shp')
```

### CUDA Acceleration

```python theme={null}
import cupy as cp
import numpy as np

# GPU-accelerated array operations
gpu_array = cp.array(np.random.random((1000, 1000)))
result = cp.linalg.svd(gpu_array)
```

## Next Steps

Now that you have GDAL and CUDA ready:

1. **Start Analysis**: Begin your geospatial analysis immediately
2. **Load Data**: Access your workspace data files
3. **Use GPU**: Leverage CUDA for accelerated processing
4. **Scale Up**: Work with large geospatial datasets efficiently

Happy analyzing!