Compare

Compare Enterprise Chips

Search and compare NVIDIA, AMD, and Intel server chips side-by-side - pick up to 8 chips at a time and the smart table auto-highlights the best value in every row

Step 1

Pick chips

Choose up to 8 to compare — the table below updates instantly

Brand

101 chips - showing 24 of 101

Step 2

Comparing 3/8 chips

NVIDIA B200 Tensor Core GPU

NVIDIA H200 Tensor Core GPU

NVIDIA H100 Tensor Core GPU

Specification

Pre-Order

NVIDIA B200 Tensor Core GPU

Blackwell

In Stock

NVIDIA H200 Tensor Core GPU

Hopper

In Stock

NVIDIA H100 Tensor Core GPU

Hopper

Memory

384GB HBM3e

141GB HBM3e

80GB HBM3

Memory Bandwidth

10 TB/s

4.8 TB/s

3.35 TB/s

Compute

Tensor Corestie

528

CUDA Corestie

18432

FP8 TFLOPS

4500

1979

FP16 TFLOPS

2250

989

TF32 TFLOPS

1125

494.5

FP64 TFLOPStie

Connectivity

Interconnect

NVLink 5.0, 1.8TB/s

NVLink 4.0, 900GB/s

Physical

TDP

1000W

700W

Form Factor

SXM6

SXM5

Cooling

Liquid Cooled

Launch Date

H2 2024

Q2 2024

Q1 2024

Manufacturing Process

4nm

Category Leaders

Memory

NVIDIA B200 Tensor Core GPU

Leads on Memory Bandwidth

Compute

NVIDIA B200 Tensor Core GPU

Leads on FP8 TFLOPS

Step 3

Visual Comparison

At a glance overview of who leads on each spec

Each corner is a spec — the bigger the shape, the more powerful the chip. A shape hugging the center means no data for that spec. Hover a chip to focus its outline.

What this tells you

Wider shape = strong across more specs — a balanced pick.
Long spike in one direction = specialized for that workload.
Trophy in bars view = the best value for that specific metric.

Tip: pair this with the table below to see exact numbers — the visuals show relative strength, the table shows the raw specs.

Browse All Categories Browse All Products

Chip Configurator

Not sure which chip fits your workload?

Answer 3 quick questions and our AI-powered configurator will recommend the optimal NVIDIA, AMD or Intel chip for your AI training, inference, HPC or professional workload.

Try the Configurator

Workload-matched picks
Budget & performance aware
Backed by certified engineers

How it works

1Pick your workload
2Set budget & performance
3Get matched chips + quote