Skip to main content

AMD · Data center · OAM module

AMD Instinct MI250X 128GB

Exact MI250X reference specifications and AI deployment guidance, separated from FusionPointAI commercial inventory.

Previous generation Reference only
Compare GPU
Neutral technical reference illustration for AMD Instinct MI250X 128GB; not a product photograph
Reference illustration; consult the verified specification table for exact hardware identity. This is not a product photograph.

Verified technical specifications

ManufacturerAMD
Exact SKUMI250X
Product familyAMD MI250X
ArchitectureCDNA 2
GenerationInstinct
Entity typeOAM module
LifecyclePrevious generation
Launch dateThe manufacturer has not published this information.
VRAM128 GB
Memory technologyHBM2e
Memory bandwidth3,200 GB/s
Power560 W
InterfaceAMD platform interface
Form factorAMD · Accelerator · Module
InterconnectInfinity Fabric
MIG / partitioningNot supported
Supported numeric formatsFP64, FP32, BF16, FP16, INT8
Software stackrocm, vllm

AI workload suitability

LLM inference

High memory capacity supports large checkpoints and multi-GPU inference when runtime and topology are certified.

Fine-tuning and training

Designed for sustained accelerator workloads; training suitability depends on interconnect, precision and cluster topology.

Image, video and embeddings

FP16 acceleration supports common generative and vector workloads within the listed VRAM envelope.

Power and cooling

560 W published power requires chassis, airflow and power delivery for this exact form factor.

Multi-GPU use

Infinity Fabric is the recorded interconnect path; topology must still be validated at node level.

Software ecosystem

rocm, vllm

Product timeline

amd-mi250x-128gb @ 7f4ca067f65dae399688591d232bdb1de38d1c9c6efe8d29b254edd6a08dc2bc

Official source

Official manufacturer source

https://www.amd.com/en/products/specifications/accelerators.html

Last verified: 2026-08-24