QuData

Rent MI250X by the Hour by QuData Ai

Explore the specifications and benchmarks of the MI250X. Compare hardware configurations and performance metrics to optimize your flagship HPC workloads with AMD's maximum GPU architecture.

GPU Description

Explore the specifications and benchmarks of the MI250X. Compare hardware configurations and performance metrics to optimize your flagship HPC workloads with AMD's maximum GPU architecture.

GPU Characteristics

GPU typeMI250X
GPU connectionOAM
GPU model nameAMD Instinct MI250X
Cooling systemLiquid Cooling
GPU memory128GB HBM2e
Power Limit (W)560.0
GPU temperature (°C)95
GPU Clock Speed (MHz)1700
Memory Clock Speed (MHz)800
Memory bandwidth3277 GB/s

GPU Configuration

Stream Processors14080
Matrix of the core220
Support for FP16/INT8Yes
ArchitectureCDNA 2
VirtualizationYes

Benchmarks

ffmpeg~12ms
Coremark (iterations per second)~135000-175000
Llama2 Inference (tokens per second)~325-425
Tensorflow Mixed Training~20.5-27.8
Stable Diffusion (images/min)~205-275
HPC Computing (TFLOPS FP64)~47.9

Providers with MI250X