Compute Comparison
vs
All comparisons →

Blackwell

NVIDIA B200 180GB

wins

4

The B200 is NVIDIA's Blackwell flagship — 4.5 PFLOPS FP16 and 8 TB/s memory bandwidth make it the most powerful GPU available for AI training at scale.

Next-gen LLM trainingTrillion-parameter modelsAI factories

Hopper

NVIDIA H200 141GB

wins

1

The H200 upgrades the H100 with 141GB HBM3e memory and 4.8 TB/s bandwidth — a massive boost for memory-bound workloads like large LLM inference.

Very large LLMsMemory-bound inferenceMulti-modal models

Performance comparison

Visual bars — winner highlighted

NVIDIA B200 180GB

Metric

NVIDIA H200 141GB

FP16 TFLOPS

4.5k T
989 T

Bandwidth

8.0k GB/s
4.8k GB/s

VRAM

1.8k GB
1.4k GB

TDP (lower=better)

1.0k W
700 W

TFLOPS/watt

4.50
1.41

GB/s per watt

8.00
6.86

Full specification table

SpecNVIDIA B200 180GBNVIDIA H200 141GB
ArchitectureBlackwellHopper
VRAM180 GB HBM3e141 GB HBM3e
VRAM typeHBM3eHBM3e
Memory bandwidth8,000 GB/s4,800 GB/s
FP16 TFLOPS4,500 TFLOPS989 TFLOPS
TDP1,000 W700 W
TFLOPS/watt4.50 T/W1.41 T/W
GB/s per watt8.00 GB/s/W6.86 GB/s/W
NVLinkYesYes
Release year20242024

Live cloud pricing

On-demand $/hr across providers — updated in real time

Loading prices…

Power efficiency analysis

TFLOPS/watt and GB/s/watt — critical for data center TCO

NVIDIA B200 180GB

FP16 TFLOPS4,500 TFLOPS
TDP1,000 W
TFLOPS/watt4.500 T/W
GB/s per watt8.00 GB/s/W

NVIDIA H200 141GB

FP16 TFLOPS989 TFLOPS
TDP700 W
TFLOPS/watt1.413 T/W
GB/s per watt6.86 GB/s/W

TFLOPS/watt measures compute efficiency — how much AI throughput you get per watt of power consumed. For data centers with PUE of 1.2–1.5, a 10% improvement in TFLOPS/watt translates directly to lower electricity costs and cooling requirements. GB/s/watt measures memory bandwidth efficiency, which is the binding constraint for memory-bound LLM inference workloads.

When to choose each GPU

Choose NVIDIA B200 180GB for:

  • Next-gen LLM training
  • Trillion-parameter models
  • AI factories

Choose NVIDIA H200 141GB for:

  • Very large LLMs
  • Memory-bound inference
  • Multi-modal models