Skip to content
MCS · Rigs · Compare

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

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

0 of 41 fields comparable

⇄ swap sides
SpecAWS Trainium2NVIDIA GeForce RTX 5070 Ti
AI Tops—1,406
Bandwidth GB S—896
Base Clock GHZ—2.3
Bf16 Tflops PER Chip667—
Boost Clock GHZ—2.45
Chips PER Trn2 Instance16—
Chips PER Ultraserver64—
Cuda Cores—8,960
Founders Edition—none (partner cards only)
Founders Edition Models Rtx50 Launch—RTX 5090, RTX 5080, RTX 5070
FP8 Dense Petaflops PER Chip1.3—
FP8 Sparse Petaflops PER Chip5.2—
HBM Bandwidth TB S PER Chip2.9—
HBM Capacity GB PER Chip96—
HBM TypeHBM3—
Memory BUS Bits—256
Memory Speed Gbps—28
MSRP (USD)↓ better—749
Neuroncores PER Chip8—
Pcie GEN—5
Required PSU Watts—750
RT Core Tflops—133
RT Cores GEN—4
Status Checked—2026-09-24
Street Price Date—2026-09-01
Street Price Premium PCT US—40
Street Price USD US Typical—1,050
TDP (W)↓ better—300
Tensor Cores GEN—5
Trainium3 GA2025-12-02—
Trainium3 HBM Bandwidth TB S4.9—
Trainium3 Hbm3e GB PER Chip144—
Trn2 Instance FP8 Petaflops20.8—
Trn2 Instance HBM TB1.5—
Trn3 Ultraserver FP8 Pflops362—
Trn3 Ultraserver MAX Chips144—
Ultraserver FP8 Petaflops83.2—
Ultraserver Total HBM Bandwidth TB S185—
Ultraserver Total HBM TB6—
VRAM (GB)—16
Vram Type—GDDR7