The Radeon Instinct MI50 compute card is designed to deliver high levels of performance in areas such as deep learning, high-performance computing (HPC), cloud computing, and graphics rendering systems. This new accelerator is designed with optimized deep learning operations, excellent double precision performance, and ultra-fast HBM2 memory offering memory bandwidth speeds of 1 TB/s.

Expand your datacenter server designs with AMD's Infinity Fabric Link technology, which can be used for direct connection of up to 2 clusters of 4 GPUs in a single server with speeds up to 5.75x the speed of PCIe® 3.0.

Achieve fast, reliable, and accurate results on large-scale systems with the Radeon Instinct MI50, which is equipped with full ECC and RAS capabilities. Combine this finely balanced and ultra-scalable solution with our open ROCm ecosystem, which includes the optimized MIOpen Radeon Instinct libraries supporting frameworks like TensorFlow, PyTorch, and Caffe 2, and you have a solution ready for the next era of computing and machine intelligence.

• Based on "Vega 7nm" technology with 60 supercharged compute units (3840 Stream Processors)
• Up to 53 TOPS INT8 performance for inference workloads
• Up to 26.5 TFLOPS FP16 and 13.3 TFLOPS FP32 performance for training workloads
• Up to 6.6 TFLOPS Double Precision for HPC
• 16 GB or 32 GB ultra-fast HBM2 ECC memory with memory bandwidth up to 1 TB/s
• The first PCIe® Gen 4 x16 capable GPU
• AMD Infinity Fabric Link – up to 184 GB/s peer-to-peer GPU communication speeds
• ROCm Open Ecosystem