Your Cart

0 items

Your cart is empty.

AMD CDNA 2 Architecture

AMD MI210

Double-precision HPC performance in a standard PCIe card.

45.3

TFLOPS FP64 Matrix

64GB

HBM2e Memory

1.6 TB/s

Memory Bandwidth

300W

PCIe Dual-Slot

Overview

Overview

The AMD Instinct MI210 brings exascale-class double-precision performance to mainstream rack servers in a standard dual-slot PCIe card, so HPC workloads that depend on FP64 do not need a specialised chassis.

MI210 vs. A100 PCIe

Both cards draw 300W in a PCIe form factor. The MI210 leads in FP64; the A100 leads in mixed precision.

MI210A100 PCIe 80GB
ArchitectureAMD CDNA 2NVIDIA Ampere
Memory64GB HBM2e80GB HBM2e
Bandwidth1.6 TB/s2.0 TB/s
FP6422.6 TFLOPS9.7 TFLOPS
FP64 Matrix45.3 TFLOPS19.5 TFLOPS
TDP300W300W
Compare all GPUs in our matrix

Compatible Servers

Dell PowerEdge Servers That Support the MI210

11 servers in our catalog are configurable with this accelerator.

Use Cases

What the MI210 Does Best

Scientific Computing & HPC

FP64 performance ahead of the A100 PCIe makes the MI210 one of the strongest double-precision accelerators in a standard PCIe form factor — climate modelling, molecular dynamics, CFD and genomics all benefit.

AI Training & Mixed Precision

181 TFLOPS of FP16/BF16 for deep learning training, so HPC-first sites can run AI convergence workloads on the same hardware.

Open Ecosystem & ROCm

AMD's ROCm stack is open-source with upstream PyTorch and TensorFlow support, and HIP conversion tooling helps port existing CUDA applications.

Specifications

AMD MI210

GPU Memory
64 GB HBM2e
ECC
Y
Bandwidth
1638 GB/sec
Max Power
300W
Host Interface
PCIe Gen4x16
Interconnect
Infinity Fabric
Slot Width
DW
Form Factor
FHFL
Aux Cable
CPU 8 pin
Workload
HPC/ML

Looking for MI210-based acceleration?

Tell us your workload and we'll help you build the right configuration.