Skip to content
AI GPU ladder

AMDAI GPU · APU (SH5 socket)

MI300A

AMD Instinct MI300A is a CDNA 3 + Zen 4 data-center accelerator with 128 GB HBM3 (unified) and 5.3 TB/s of memory bandwidth.

Ladder rank
#6
AI index (H100 SXM = 100)
125.2
Series
Instinct MI300

Specifications

Dense throughput, no sparsity
Architecture CDNA 3 + Zen 4
Form factor APU (SH5 socket)
Process TSMC 5 nm + 6 nm
FP16 / BF16 tensor (dense) 980.6 TFLOPS
FP8 tensor (dense) 1,961.2 TFLOPS
FP4 tensor (dense) —
Memory 128 GB HBM3 (unified)
Memory bandwidth 5.3 TB/s
Power 550 W
Launch Jan 2023

AI index

Tensor × bandwidth

100 × √((980.6 ÷ 989.4) × (5.3 ÷ 3.35)) = 125.2

The AI index is the geometric mean of dense 16-bit tensor throughput and memory bandwidth, each relative to the H100 SXM. It is a spec-sheet comparison, not a measured benchmark.

Sources

2 references