GVirtuS gets a major boost with RDMA! 

We’re excited to present a breakthrough from the CLEVER Project: a novel RDMA over InfiniBand Communicator that significantly accelerates GPGPU virtualization using GVirtuS — enabling smarter, faster, and more efficient AI workloads across remote GPUs ⚡🧠 

Traditional TCP/IP-based GPU virtualization is plagued by context switches and latency. With the new RDMA Communicator, we slash those overheads and enable up to 82% faster performance on real HPC systems. 

🧪 Key Results: 

  • 35–55% performance gain over TCP/IP 
  • 10x fewer context switches = lower CPU load 
  • Seamless integration into GVirtuS plug-in architecture 
  • Tested on real CUDA workloads: Matrix Multiplication & SAXPY 

💡 The RDMA-enhanced GVirtuS now delivers high-speed CUDA offloading with: 

  • Pre-registered memory regions 🧠 
  • Polling-based completions for low latency 
  • Real-time optimized buffer management 

Read Zenodo paper to see how the RDMA Communicator stacks up in matrix multiplication across different communication methods 👇 
(Lower = Better Execution Time

🌐 Powered by the CLEVER Project: https://www.cleverproject.eu 

Read zenodo paper: https://zenodo.org/records/14717622