Compute Comparison
vs
All comparisons →

CDNA 3.5

AMD Instinct MI325X 256GB

wins

1

The AMD Instinct MI325X is an upgraded MI300X with 256 GB HBM3e — the highest memory capacity of any single GPU.

Massive context LLM inferenceMemory-bound workloadsLarge open-source models

Blackwell

NVIDIA GB200 NVL72

wins

4

The GB200 NVL72 is NVIDIA's most powerful AI system.

Trillion-parameter trainingAI factory scaleNext-gen LLM inference

Performance comparison

Visual bars — winner highlighted

AMD Instinct MI325X 256GB

Metric

NVIDIA GB200 NVL72

FP16 Tensor TFLOPS (dense)

1.5k T
9.0k T

Bandwidth

6.0k GB/s
16.0k GB/s

VRAM

256 GB
384 GB

TDP (lower=better)

750 W
2.7k W

TFLOPS/watt

1.96
3.33

GB/s per watt

8.00
5.93

Full specification table

SpecAMD Instinct MI325X 256GBNVIDIA GB200 NVL72
ArchitectureCDNA 3.5Blackwell
VRAM256 GB HBM3e384 GB HBM3e (per GPU)
VRAM typeHBM3eHBM3e (per GPU)
Memory bandwidth6,000 GB/s16,000 GB/s
FP16 Tensor TFLOPS (dense)1,470 TFLOPS9,000 TFLOPS
TDP750 W2,700 W
TFLOPS/watt1.96 T/W3.33 T/W
GB/s per watt8.00 GB/s/W5.93 GB/s/W
NVLinkNoYes
Release year20242025

Live cloud pricing

On-demand pricing across providers — updated in real time

Per hourPer GB VRAM/hr
Loading prices…

Power efficiency analysis

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

AMD Instinct MI325X 256GB

FP16 TFLOPS1,470 TFLOPS
TDP750 W
TFLOPS/watt1.960 T/W
GB/s per watt8.00 GB/s/W

NVIDIA GB200 NVL72

FP16 TFLOPS9,000 TFLOPS
TDP2,700 W
TFLOPS/watt3.333 T/W
GB/s per watt5.93 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 AMD Instinct MI325X 256GB for:

  • Massive context LLM inference
  • Memory-bound workloads
  • Large open-source models

Weaknesses / Limitations

  • ROCm software ecosystem lags CUDA maturity
  • Very limited cloud availability vs NVIDIA equivalents
  • No FP8 hardware support in current ROCm stack

Choose NVIDIA GB200 NVL72 for:

  • Trillion-parameter training
  • AI factory scale
  • Next-gen LLM inference

Weaknesses / Limitations

  • Rack-scale system — not a single GPU, requires full NVL72 rack deployment
  • 2,700W system TDP demands dedicated power and liquid cooling infrastructure
  • Extremely limited availability — primarily hyperscaler and large enterprise