AMD Instinct GPUs

AMD Instinct GPU’s

The time of exascale has come.

Researchers are now able to solve grand challenges, once thought insurmountable, thanks to enormous computational power coupled with HPC and AI.

Why AMD Instinct™?

AMD Instinct™ accelerators are engineered from the ground up for this new era of computing, supercharging HPC and AI workloads to propel new discoveries.

Powered by the AMD CDNA architecture, AMD Instinct™ MI100 accelerators deliver a giant leap in compute and connectivity, offering a nearly 3.5x performance boost for HPC (FP32 matrix) and a nearly 7x boost for AI workloads (FP16) compared to AMD’s prior generation accelerators.

Solutions Powered by AMD Instinct™

Combined with the award winning AMD EPYC™ processors and AMD Infinity Fabric™ technology, AMD Instinct™ GPU powered systems provide scientists and researchers platforms that propel discoveries today and prepare them for exascale tomorrow.

Speak with one of our system engineers today

Extraordinary Under the Hood

At the heart of the AMD Instinct™ products is AMD CDNA architecture-breaking new ground for the convergence of HPC and AI in the era of Exascale.

Powered by compute cores designed from the ground-up to deliver nearly 1.8x FP64 performance efficiency than previous generation and all-new Matrix Cores technology delivering a giant leap in HPC and AI performance, AMD CDNA drives computational discoveries to new heights.

Technologies

AMD CDNA™ Architecture

AMD CDNA architecture redefines computing for the modern era. Engineered from the ground up with compute cores designed for the Exascale era, AMD CDNA architecture delivers nearly 1.7x higher FP64 performance within the same power budget compared to previous generation AMD accelerators.

AMD ROCm™ – Open, Flexible and Portable

When it comes to solving the world’s most profound challenges, scientists and researchers need the most powerful and accessible tools at their fingertips. With the ROCm™ open software platform built for flexibility and performance, the machine learning and HPC communities can now gain access to an array of different open compute languages, compilers, libraries and tools designed from the ground up to meet their most demanding needs- helping to accelerate code development and solve the toughest challenges in the world today.

AMD Infinity Architecture

AMD Infinity Architecture, introduced with the 2nd Gen AMD EPYC™ Processors, empowers system builders and cloud architects alike to unleash the very latest in server performance without sacrificing power, manageability, or the ability to help secure their organization’s most important assets, its data.

Model Specifications

AMD Instinct™ MI100 Accelerator

Compute Units120
Stream Processors7,680
FP16 Performance184.6 TFLOPs
FP32 Performance23.1 TFLOPs
FP64 Performance11.5 TFLOPs
Memory32 GB
Mem. TypeHBM2
Peak Mem.1228.8 GB/s

AMD Radeon Instinct™ MI50 Accelerator (32GB)

Compute Units60
Stream Processors3840
FP16 Performance26.5 TFLOPs
FP32 Performance13.3 TFLOPs
FP64 Performance6.6 TFLOPs
Memory32 GB
Mem. TypeHBM2
Peak Mem.1024 GB/s

Radeon Instinct™ MI25 Accelerator

Compute Units64
Stream Processors4096
FP16 Performance24.6 TFLOPs
FP32 Performance12.29 TFLOPs
FP64 Performance768 GFLOPs
Memory16 GB
Mem. TypeHBM2
Peak Mem.484 GB/s