Nvidia expands CUDA support to RISC-V and enhances CUDA-X libraries

Nvidia is expanding its software ecosystem by targeting RISC-V architecture and enhancing its CUDA-X suite of GPU-accelerated libraries. To support RISC-V, Nvidia is establishing specific requirements for server-grade CPUs, including adherence to RVA23 standards, reliability, availability, and serviceability (RAS) features, and PCIe coherency. These requirements aim to ensure efficient software performance and avoid the limitations of a “lowest common denominator” approach.

Simultaneously, Nvidia is growing its CUDA-X toolbox, which includes between 400 and 900 pre-optimized libraries for AI, high-performance computing, and engineering. Recent updates at the DAC 2026 conference introduced new solver libraries such as cuISS, cuDSS, and cuEST for quantum chemistry. These tools have demonstrated significant performance gains, including up to an 11x speedup for computational engineering tools and a 20x gain in computational lithography for clients like Samsung. Additionally, Cloudera has integrated Nvidia’s cuDF library to accelerate Apache Spark workloads by up to four times.