Skip to content
Enterprise hardware · New & pre-owned supply
sales@aster-rackit.comEnglish / Global
Quote basket 0Configure system →
☰
GPUs & acceleratorsAMD
Instinct MI300X

Request exact-unit photography.
No unverified model image substituted.

manufacturer model

Instinct MI300X

AMD Instinct MI300X — GPUs & accelerators

AMDGPUs & accelerators45 technical fields
Family
Instinct
Series
Instinct MI300 Series
Form Factor
Servers
Launch Date
12/06/2023

OEM catalog fields reviewed · remaining checks open

Price and availability: Request quotation

Download editable RFQ checklist ↓

A technical record does not certify stock, condition, current orderability or a factory-approved BOM.

Identity & purchasing scope

AMD Instinct MI300X — the identity baseline for comparing AMD gpus & accelerators offers. Field notes: family Instinct, series Instinct MI300 Series, form factor Servers. Evidence labels separate reviewed from unverified values. Use the quote basket to collect this record, then confirm the final configuration with the sales team.

This record identifies Instinct MI300X in the AMD gpus & accelerators catalog. Its classification is manufacturer model. Preserve the full identifier when comparing offers: suffixes, revisions and assembly contents are part of the purchasing question.

ASTER RACKIT reference for AMD Instinct MI300X. AMD's official specification table identifies this model and the individually sourced technical fields below. Compare the platform, power, memory and interface requirements before requesting a quotation. Use the technical record to prepare an inquiry rather than as evidence that an item is available, tested, licensed or covered by a particular warranty. Inherited fast-shipping and stock language is not carried into this catalog's commercial fields.

Inherited search aliases: Instinct MI300X · AMD Instinct MI300X. These are lookup aids, not certified equivalents.

Recorded technical specifications

Each field states its evidence status. Inherited and title-derived values remain unverified; reviewed OEM facts do not certify an offered configuration.

FieldRecorded valueSource snapshot
FamilyInstinctoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
SeriesInstinct MI300 Seriesoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Form FactorServersoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Launch Date12/06/2023oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
GPU ArchitectureCDNA3oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
LithographyTSMC 5nm | 6nm FinFEToem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Stream Processors19,456oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Matrix Cores1216oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Compute Units304oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Engine Clock2100 MHzoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Eight-bit Precision (FP8) Performance (E5M2, E4M3)2.61 PFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Eight-bit Precision (FP8) Performance with Structured Sparsity (E5M2, E4M3)5.22 PFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Half Precision (FP16) Performance1.3 PFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Half Precision (FP16) Performance with Structured Sparsity2.61 PFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Single Precision (TF32 Matrix) Performance653.7 TFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Single Precision (TF32) Performance with Structured Sparsity1.3 PFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Single Precision Matrix (FP32) Performance163.4 TFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Single Precision (FP32) Performance163.4 TFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Double Precision Matrix (FP64) Performance163.4 TFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Double Precision (FP64) Performance81.7 TFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak INT8 Performance2.6 POPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak INT8 Performance with Structured Sparsity5.22 POPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak bfloat161.3 PFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak bfloat16 with Strutured Sparsity2.61 PFLOPsoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Transistor Count153oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
External Power Connectors54V UBBoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Typical Board Power (TBP)750W Peakoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Last Level Cache (LLC)256 MBoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Dedicated Memory Size192 GBoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Dedicated Memory TypeHBM3oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Infinity CacheYesoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Memory Interface8192oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Memory Clock5.2 GHzoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Memory Bandwidth5.3 TB/soem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Memory ECC SupportYes (Full-Chip)oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
GPU Form FactorOAM Moduleoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Bus TypePCIe 5.0 x16oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Infinity Fabric Links8oem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Peak Infinity Fabric Link Bandwidth128 GB/soem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
CoolingPassive OAMoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
RAS SupportYesoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Page RetirementYesoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Page AvoidanceYesoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
SR-IOVYesoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z
Supported TechnologiesAMD CDNA 3 Architecture, AMD ROCm - Ecosystem without Borders, AMD Infinity Architectureoem_public_catalog
oem_catalog_verified · 2026-09-12T08:51:40.010Z

Configuration & purchase checks

  1. Keep the full accelerator name, memory capacity, form factor and edition.
  2. Do not treat workstation, PCIe server, SXM and OAM products as interchangeable hardware.
  3. Review exact server qualification, risers, auxiliary power, airflow and cooling together.
  4. List interconnect bridges, cables and server enablement kits as separate bill-of-materials requirements.
  5. Request labels, connectors and included accessories. Marketing performance figures are not a benchmark of the offered machine.

Adding an item prepares a local RFQ line. It does not create a price offer, reserve inventory or submit an order.

Read the category procurement guide →

Typical deployment contexts

Common project contexts where buyers request this gpus & accelerators. Contexts inform the inquiry — they are not compatibility claims.

01

AI training cluster build-out

02

Inference infrastructure expansion

03

GPU virtualization rollout

Frequently asked questions

What is the Instinct MI300X?

The Instinct MI300X is a recorded AMD gpus & accelerators identity (manufacturer model). Recorded highlights: Family Instinct; Series Instinct MI300 Series.

Are the specifications on this page verified?

The record carries 45 technical fields, each with an evidence label. Reviewed fields come from OEM catalog evidence; unverified fields stay visibly marked and should be confirmed before purchase.

How do I get a price for the Instinct MI300X?

Add the record to the quote basket and send the inquiry — ASTER RACKIT confirms price, condition and availability in a written quotation.

Does an image on this page show the actual offered unit?

No. Images are reference material only. Request exact-unit photography with serial numbers before committing to an order.

Platform relationships & upgrade review

No explicit target-platform relationship is provided for this record. Brand, socket, capacity and visual resemblance alone must not be used to invent a compatibility claim.

Related records below are category neighbors. They are not approved replacement parts.

Actual-unit evidence to request

01

Card or module label

02

Host and power connectors

03

Bracket or carrier

04

Cooling construction

05

Interconnect accessories

No server boot-screen image is attached to this record. Request product photographs and any diagnostic or entitlement record as separate evidence.

Facility reference scenes from our warehouse and test benches. They show the environment and handling standard for this equipment family — they are not photographs of the exact offered unit.

At acceptance, compare the delivered unit, labels and accessories with the approved list. Retain source revisions and receiving records for later maintenance.

Source trace & documentation

Remaining review: complete platform compatibility · complete mechanical integration requirements · all omitted source fields · current stock availability · production lifecycle beyond any explicit launch date

Source fileSource recordEvidence status
How evidence quality is classified →

Frequently asked questions

Is the item currently available?

This is not a live stock feed. Confirm full identifier, condition, quantity, price and delivery scope in a written quotation.

Does this record certify compatibility?

No. Inherited fields and platform associations require review against the exact destination system and current manufacturer documentation.

What should I send in an RFQ?

Include the full identifier, quantity, destination, target platform, condition, accessories and evidence needed for acceptance.

Is a configuration number another server model?

No. Models, ordering configurations, options and reference labels are distinct record types. Ask which actual purchasing unit will be supplied.

Related category records

Browse category →
AMDmanufacturer model
Instinct MI350P

AMD Instinct MI350P — GPUs & accelerators

Family
Instinct
Series
Instinct MI350 Series
Form Factor
Servers

OEM catalog fields reviewed · remaining checks open

View details →
AMDmanufacturer model
Instinct MI350X

AMD Instinct MI350X — GPUs & accelerators

Family
Instinct
Series
Instinct MI350 Series
Form Factor
Servers

OEM catalog fields reviewed · remaining checks open

View details →
AMDmanufacturer model
Instinct MI455X

AMD Instinct MI455X — GPUs & accelerators

Family
Instinct
Series
Instinct MI400 Series
Launch Date
7/23/2026

OEM catalog fields reviewed · remaining checks open

View details →
AMDmanufacturer model
Instinct MI300A

AMD Instinct MI300A — GPUs & accelerators

Family
Instinct
Series
Instinct MI300 Series
Form Factor
Servers

OEM catalog fields reviewed · remaining checks open

View details →
AMDmanufacturer model
Instinct MI355X

AMD Instinct MI355X — GPUs & accelerators

Family
Instinct
Series
Instinct MI350 Series
Form Factor
Servers

OEM catalog fields reviewed · remaining checks open

View details →
AMDmanufacturer model
Instinct MI325X

AMD Instinct MI325X — GPUs & accelerators

Family
Instinct
Series
Instinct MI300 Series
Form Factor
Servers

OEM catalog fields reviewed · remaining checks open

View details →
01

GPU & AI infrastructure — the next decisions

AMD / GPU & AI infrastructure

Instinct MI300X

Distinguish PCIe cards from integrated accelerator baseboards.

This guide adds the decisions, documents and next steps that support this page.

GPU platforms
GPU & AI infrastructure illustrative environment
02

Choose the right route

Training

Start with model size, precision, parallelism and dataset movement.

GPU platforms

Inference

Define request concurrency, latency target and model-memory footprint.

AI systems

GPU expansion

Check slot width, risers, auxiliary power and supported cooling.

GPU checklist
03

Details that change the configuration

Accelerator architecture

Distinguish PCIe cards from integrated accelerator baseboards.

Fabric & data path

Plan host networking, GPU interconnect and storage access separately.

Rack power & cooling

Request platform power requirements and facility cooling review.

04

From requirement to handover

  1. 01

    Define AI workload

  2. 02

    Size accelerator memory

  3. 03

    Review fabric & cooling

  4. 04

    Confirm supported BOM

A planning sequence, not a record of completed work.

05

Explore the relevant catalog

AMD

Recorded catalog identities

Counts show records in the site catalog. They do not represent stock, available quantities or verified compatibility.

Snapshot: 2026-09-15

06

Compare the buying requirements

GPU & AI infrastructure comparison brief
DecisionRecord in your briefEvidence to retain
Accelerator architectureDistinguish PCIe cards from integrated accelerator baseboards.Complete identifier and source revision
Fabric & data pathPlan host networking, GPU interconnect and storage access separately.Platform document or written clarification
Rack power & coolingRequest platform power requirements and facility cooling review.Offered-unit details and agreed scope
Open the comparison workspace
07

Build your readiness checklist

Your preparation progress

Tick items as you prepare them. This is your personal checklist, not a hardware validation.

0 / 4 prepared0

Saved in this browser when local storage is available.

08

A closer look at the hardware

Instinct MI300X — Gpu Node Interior facility reference
Gpu Node Interior Explore
Instinct MI300X — Nvlink Rack Scale facility reference
Nvlink Rack Scale Explore
1 / 2

Reference-based generated scenes; actual-unit photos are available only through an explicit unit-specific request. Related models are separate catalog records, not substitutes for the item on this page.

09

Keep the supporting material together

GPU checklist

Retain the supporting reference with your project brief.

Open resource
10

Turn the brief into the next step

AMD / GPU & AI infrastructure

Keep the requirement specific.

Workload and precision defined · Accelerator model and quantity named · Platform support documents requested · Power and cooling review assigned.