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

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

Rig A

AWS · cloud-model

Change

Rig B

NVIDIA · workstation

Change

0 of 45 fields comparable

Spec	AWS Trainium2	NVIDIA DGX B200
Bf16 Tflops PER Chip	667	—
Bluefield3 Dpus	—	2
Chips PER Trn2 Instance	16	—
Chips PER Ultraserver	64	—
CPU Base GHZ	—	2.1
CPU Count	—	2
CPU MAX Boost GHZ	—	4
CPU Model	—	Intel Xeon Platinum 8570
FP4 Pflops Dense	—	72
FP4 Pflops Sparse	—	144
FP8 Dense Petaflops PER Chip	1.3	—
FP8 Pflops	—	72
FP8 Sparse Petaflops PER Chip	5.2	—
Gpus	—	8
HBM Bandwidth TB S	—	64
HBM Bandwidth TB S PER Chip	2.9	—
HBM Capacity GB PER Chip	96	—
HBM Type	HBM3	—
Height MM	—	444
Internal Storage TB	—	30.72
Length MM	—	897.1
MAX Power KW	—	14.3
Networking Osfp Ports	—	4
Neuroncores PER Chip	8	—
Nvswitch Count	—	2
OS	—	NVIDIA DGX OS / Ubuntu
OS Storage TB	—	3.8
PER Port Gbps MAX	—	400
Rack Units	—	10
System Memory GB Default	—	2,048
System Memory GB MAX	—	4,096
Total CPU Cores	—	112
Total Hbm3e GB	—	1,440
Total Nvlink Bandwidth TB S	—	14.4
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	—
Width MM	—	482.2