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 · gpu

Change

0 of 38 fields comparable

Spec	AWS Trainium2	NVIDIA GeForce RTX 5090 Founders Edition
AI Tops	—	3,352
Bandwidth GB S	—	1,792
Base Clock GHZ	—	2.01
Bf16 Tflops PER Chip	667	—
Boost Clock GHZ	—	2.41
Chips PER Trn2 Instance	16	—
Chips PER Ultraserver	64	—
Cuda Cores	—	21,760
Displayport Count	—	3
Displayport Version	—	2.1b
FP8 Dense Petaflops PER Chip	1.3	—
FP8 Sparse Petaflops PER Chip	5.2	—
HBM Bandwidth TB S PER Chip	2.9	—
Hbm3e Capacity GB PER Chip	96	—
Hdmi Count	—	1
Hdmi Version	—	2.1b
Length MM	—	304
Memory BUS Bits	—	512
MSRP (USD)↓ better	—	1,999
Neuroncores PER Chip	8	—
Nvdec Count	—	2
Nvdec GEN	—	6
Nvenc Count	—	3
Nvenc GEN	—	9
Pcie GEN	—	5
Required PSU Watts	—	1,000
RT Core Tflops	—	318
RT Cores GEN	—	4
Slot Width	—	2
TDP (W)↓ better	—	575
Tensor Cores GEN	—	5
Trn2 Instance FP8 Petaflops	20.8	—
Ultraserver FP8 Petaflops	83.2	—
Ultraserver Total HBM Bandwidth TB S	185	—
Ultraserver Total HBM TB	6	—
VRAM (GB)	—	32
Vram Type	—	GDDR7
Width MM	—	137