AMD Instinct MI210's 2nd Gen CDNA Architecture Uses AI

GigaIO & AMD: Facilitating increased computational effectiveness, scalability, and quicker AI workload deployments

Previously, 32 GPUs required four servers with eight GPUs each, plus the price and latency of connecting all that gear

Computing for double precision (FP64) on PCIe form factors for mainstream HPC and AI workloads with the AMD Instinct MI210 accelerator

By utilising the AMD ROCm software ecosystem in conjunction with MI210 accelerators

The MI210 accelerator accelerates deep learning training with 181 teraflops peak theoretical FP16 and BF16 performance

Using AMD EPYC CPUs and AMD Instinct MI210 accelerators, AMD is delivering performance, efficiency, and overall system throughput for HPC and AI

For deep learning training, the AMD Instinct MI210 accelerator provides 181 teraflops of peak FP16 and bfloat16 floating-point performance