GPU Computing

BioGrids supports GPU acceleration for computationally intensive bioinformatics applications. NVIDIA GPUs with CUDA provide significant performance improvements for sequence alignment, machine learning, and molecular dynamics.

Requirements

BioGrids requires NVIDIA GPUs with Compute Capability 3.5 or higher. AMD and other GPUs are not currently supported. Check compatibility at NVIDIA CUDA GPUs.

Install recent NVIDIA drivers from the NVIDIA Developer Zone. CUDA libraries are included with BioGrids (versions 5-11 supported). Linux provides better GPU support than macOS.

Available Applications

GPU acceleration is available for BLAST+ sequence alignment, DIAMOND protein alignment, Minimap2 read mapping, and GROMACS molecular dynamics. Machine learning frameworks include GPU-enabled TensorFlow, PyTorch, and AlphaFold. Image analysis tools include CellProfiler and ImageJ/FIJI with GPU plugins.

Installation

Check GPU Status

# Check NVIDIA GPU
nvidia-smi

# Check CUDA version
nvcc --version

# Verify GPU in BioGrids
biogrids-info --gpu-status

Install GPU Software

# Install GPU-enabled versions
biogrids-cli install tensorflow@2.8.0_gpu
biogrids-cli install blast@2.12.0_cuda
biogrids-cli install gromacs@2022.4_cuda

# List available GPU versions
biogrids-cli list --gpu-only

Environment Configuration

# Set GPU visibility (optional)
export CUDA_VISIBLE_DEVICES=0,1

# Configure GPU memory (if needed)
export TF_MEMORY_GROWTH=true

Troubleshooting

Common GPU issues include driver mismatches, memory errors, and performance problems. Verify your driver supports the required CUDA version and check GPU memory usage with nvidia-smi. Monitor GPU utilization during processing to ensure proper acceleration.

For GPU support, contact help@biogrids.org with your GPU model, driver version, CUDA version, and specific error messages.