NVIDIA Ada Lovelace Architecture for Professional Visualization

Third-generation RT Cores deliver up to 2X the ray tracing performance over the previous generation, delivering groundbreaking performance for photorealistic rendering. Enhanced RT Cores combined with new Shader Execution Reordering (SER) technology dynamically reorders inefficient workloads, dramatically improving shader performance to accelerate end-to-end ray traced image rendering performance.

Specialized for deep learning matrix multiply and accumulate math operations, the new fourth-generation Tensor Cores accelerate more data types and support Fine-Grained Structured Sparsity to deliver more than 2X throughput for tensor matrix operations compared to the previous generation. The updates to the Tensor Cores will also accelerate new FP8 precision modes while independent floating-point and integer data paths allow more efficient execution of workloads using a mix of computation and addressing calculations.

The Ada-based CUDA® cores bring double-speed processing for single-precision floating point (FP32) operations over the previous generation GPUs. This provides significant performance gains for graphics workflows like 3D model development and compute workflows like desktop simulation for computer-aided engineering (CAE).

New Ada GPUs support PCIe Gen 4.0, providing 2X the bandwidth of PCIe Gen 3.0. This improves data transfer speeds from CPU memory for data-intensive tasks such as AI and data science. Faster PCIe performance also accelerates GPU direct memory access (DMA) transfers, enabling faster video data transfers from GPUDirect® for video-enabled devices and faster I/O with GPUDirect Storage.