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

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

Rig A

NVIDIA · workstation

Change

Rig B

AWS · cloud-model

Change

0 of 35 fields comparable

Spec	NVIDIA DGX H200	AWS Trainium2
Bf16 Tflops PER Chip	—	667
Chips PER Trn2 Instance	—	16
Chips PER Ultraserver	—	64
CPU	2x Intel Xeon Platinum 8480C	—
CPU Cores Total	112	—
FP8 Dense Petaflops PER Chip	—	1.3
FP8 Petaflops	32	—
FP8 Sparse Petaflops PER Chip	—	5.2
GPU Count	8	—
GPU Model	H200 SXM	—
HBM Bandwidth TB S PER Chip	—	2.9
HBM Capacity GB PER Chip	—	96
HBM Type	—	HBM3
Internal Storage	8x 3.84TB NVMe U.2	—
MAX Power KW	10.2	—
Network MAX Gbps PER Port	400	—
Network Peak TB S	1	—
Network Ports	8x ConnectX-7 single-port (4x OSFP)	—
Neuroncores PER Chip	—	8
Nvlink PER GPU Bandwidth GB S	900	—
Nvswitch Bandwidth TB S	7.2	—
Nvswitch Count	4	—
OS Storage	2x 1.92TB NVMe M.2	—
System Memory TB	2	—
Total GPU Memory GB	1,128	—
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