Side-by-side, with the numbers that matter.

Two rigs, one table. Higher numbers win in cobalt; TDP and price flip.

Rig A

AMD · accelerator

Change

Rig B

AWS · cloud-model

Change

0 of 37 fields comparable

Spec	AMD Instinct MI300X	AWS Trainium2
Bf16 Matrix Tflops Dense	1,307.4	—
Bf16 Tflops PER Chip	—	667
Chips PER Trn2 Instance	—	16
Chips PER Ultraserver	—	64
Compute Units	304	—
Form factor	OAM	—
Fp16 Matrix Tflops Dense	1,307.4	—
Fp32 Vector Tflops	163.4	—
Fp64 Matrix Tflops	163.4	—
Fp64 Vector Tflops	81.7	—
FP8 Dense Petaflops PER Chip	—	1.3
FP8 Matrix Tflops Dense	2,614.9	—
FP8 Matrix Tflops Sparse	5,229.8	—
FP8 Sparse Petaflops PER Chip	—	5.2
HBM (GB)	192	—
HBM Bandwidth TB S PER Chip	—	2.9
HBM Capacity GB PER Chip	—	96
HBM Type	HBM3	HBM3
INT8 TOPS (dense)	2,614.9	—
Matrix Cores	1,216	—
Memory bandwidth (TB/s)	5.3	—
Neuroncores PER Chip	—	8
Pcie GEN	5	—
Process Node NM	5	—
TBP Watts	750	—
Tf32 Matrix Tflops Dense	653.7	—
Tf32 Matrix Tflops Sparse	1,307.4	—
Trainium3 GA	—	2025-12-02
Trainium3 HBM Bandwidth TB S	—	4.9
Trainium3 Hbm3e GB PER Chip	—	144
Trn2 Instance FP8 Petaflops	—	20.8
Trn2 Instance HBM TB	—	1.5
Trn3 Ultraserver FP8 Pflops	—	362
Trn3 Ultraserver MAX Chips	—	144
Ultraserver FP8 Petaflops	—	83.2
Ultraserver Total HBM Bandwidth TB S	—	185
Ultraserver Total HBM TB	—	6