AMD Instinct MI300A Full Specs

Summary

14,592 Shaders 2.1GHz

128GB HBM35.33TB/s

Peak·FP32·FP16 122.6 TFLOPS

Form Factor APU SH5 Socket

TDP 760W

Highlights

Peak OPS

Peak OPS FP32 FP16 Tensor FP8-32 FP8-32 Tensor Sparse Tensor FP16-32 FP16-32 Tensor Sparse Tensor FP32 FP64 Tensor FP64 Tensor BF16 BF16 Tensor Sparse Tensor TF32 Tensor INT8 Memory Bandwidth

Instinct MI300A3.92 POPSINT8 Tensor Sparse

x1

Top Score3.35 PFLOPS FP4 Tensor Sparse

x0.85

Desktop1xGeForce RTX 5090

Theoretical Performance

Clock Speed2100MHz UC·Base·Boost·OC

Peak OPS 3.92 POPSINT8 Tensor Sparse

Tensor FP8-32 1.96 PFLOPS FP8-32 Tensor Sparse 3.92 PFLOPS

FP16 122.6 TFLOPS Tensor FP16-32 980.6 TFLOPS FP16-32 Tensor Sparse 1.96 PFLOPS

Tensor BF16 980.6 TFLOPS BF16 Tensor Sparse 1.96 PFLOPS

Tensor TF32 490.3 TFLOPS

FP32 122.6 TFLOPS Tensor FP32 122.6 TFLOPS

FP64 61.29 TFLOPS Tensor FP64 122.6 TFLOPS

Tensor INT8 1.96 POPS

Texture Rate 1.92 TTexel/s

Core Config

Shaders 14,592 Shaders

TMUs 912 TMUs

Tensor Cores 912 T-Cores

CUs 228 CUs

Clocks

Base Clock 1GHz

Boost Clock 2.1GHz

Cache

L1 32KB/CU

L2 Cache 24MB shared

L3 Cache 256MB shared

Memory

128GB HBM3

Memory Bus 8192-bit

Memory Speed 5.2GT/s

Memory Bandwidth 5.33TB/s

ECC Yes

Power & Thermals

TDP 760W

API Support

OpenCL 3

GFX 9.4

Decoder

Decoder 3x VCN 4.0.3

Codec JPEG VP9 AVC (H.264) HEVC (H.265) AV1

Card

Form Factor APU SH5 Socket

PCIe PCIe 5.0 x16

Cooling Passive

Multi-GPU Multi-GPU Support Supported Multi-GPU Type Infinity Fabric

Chip

Manufacturer AMD

Chip Designer AMD

Architecture CDNA 3

Family Instinct

Branding

Codename Aqua Vanjaram

Chip Variant Aqua Vanjaram XT

Market Segment Server

Release Date Dec 6, 2023

Fabrication

Foundry TSMC

Other Foundries TSMCIO

Fabrication Node N5 N6IO

Die Size 690mm² 1480mm²IO

Transistor Count 78B 48.8BIO

Transistor Density 113 MTr/mm² 32.97 MTr/mm²IO

Images