The Radeon Instinct™ MI25 accelerator, designed with the most advanced Next-Gen “Vega” GPU architecture, is the ultimate training accelerator for large scale machine intelligence and Deep Learning, along with being an optimized open compute workhorse for single-precision HPC-class system workloads.

The MI25 delivers leadership in FP16 and FP32 performance in a passively-cooled single GPU Server card with 24.6 TFLOPS of FP16 and 12.3 TFLOPS of FP32 peak performance through its 64 compute units with 4,096 stream processors.1

The Radeon Instinct MI25’s powerful compute engine and advanced memory architecture, combined with AMD’s ROCm open software platform and ecosystem, provides a powerful, flexible heterogeneous compute solution that allows datacenter designers to meet the challenges of a new era of compute and Machine Intelligence.

Specification

GPU Architecture:                         "Vega"

Stream Processors:                      4,096

Performance:

Half-Precision (FP16)                    24.6 TFLOPS

Single-Precision (FP32)                12.3 TFLOPS

Double-Precision (FP64)              768 GFLOPS

GPU Memory:                               16GB HBM2

Memory Bandwidth:                    Up to 484 GB/s

ECC:                                                 Yes

Bus Interface:                                PCIe® Gen 3 x16

MxGPU Capability:                        Yes

Board Form Factor:                       Full-Height, Dual-Slot

Length:                                            10.5”

Thermal Solution:                          Passively Cooled

Standard Max Power:                   300W TDP

OS Support:                                    Linux® 64-bit

ROCm Software Platform:            Yes

Programing Environment:

ISO C++, OpenCL™, CUDA (via AMD’s HIP conversion tool) and Python5 (via Anaconda’s NUMBA)