Skip to main content

AMD Instinct MI300X (CDNA 3)

Overview

The AMD Instinct MI300X, launched in early 2024, is based on the CDNA 3 architecture with 3D chiplet packaging. 192GB HBM3 memory (~2.4× that of the H100), delivering 1,307 TFLOPS FP16 compute. It is a direct competitor to the NVIDIA H100 / H200.

Core Specifications

ItemSpecification
ArchitectureCDNA 3
Process NodeTSMC 5nm + 6nm (hybrid)
Transistor Count146 billion
Memory192 GB HBM3
Memory Bandwidth5,300 GB/s
FP3265.3 TFLOPS
FP6432.6 TFLOPS
FP16/BF16 Matrix1,307 TFLOPS
FP8 Matrix2,614 TFLOPS
INT8 Matrix2,614 TOPS
TDP750 W
InterconnectInfinity Fabric 4.1 (448 GB/s), PCIe 5.0
Form FactorOAM (UBB 8-card)

MI300X vs H100 / H200

MetricMI300XH100H200
Memory192 GB80 GB141 GB
Memory TypeHBM3HBM3HBM3e
Bandwidth5.3 TB/s3.35 TB/s4.8 TB/s
FP16 Compute1,307 TFLOPS989 TFLOPS989 TFLOPS
FP8 Compute2,614 TFLOPS3,958 TFLOPS3,958 TFLOPS
TDP750 W700 W700 W
Price~$15,000~$30,000~$33,000

Key advantage: Large memory, lower price, significant ROCm 7 performance improvements.

Vendor Information

ItemDetails
ManufacturerAMD, Inc.
Official Websitehttps://www.amd.com
Product Pagehttps://www.amd.com/en/products/accelerators/instinct/mi300x.html
Drivershttps://www.amd.com/en/support/accelerators/amd-instinct.html
ROCmhttps://rocm.docs.amd.com
ROCm Version5.7+ (6.x recommended)

Use Cases

  • LLM training and inference (192GB memory advantage)
  • MoE models (70B+ fits on a single card)
  • HPC and AI mixed workloads
  • Cost-sensitive customers