Compute Comparison
vs
All comparisons →

CDNA 3.5

AMD Instinct MI325X 256GB

wins

4

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

Ampere

NVIDIA RTX 3080 10GB

wins

1

The RTX 3080 is an entry-level option for AI workloads with 10GB GDDR6X.

Entry-level AISmall model inferenceHobbyist compute

Performance comparison

Visual bars — winner highlighted

AMD Instinct MI325X 256GB

Metric

NVIDIA RTX 3080 10GB

FP16 Tensor TFLOPS (dense)

1.5k T
45 T

Bandwidth

6.0k GB/s
760 GB/s

VRAM

256 GB
10 GB

TDP (lower=better)

750 W
320 W

TFLOPS/watt

1.96
0.14

GB/s per watt

8.00
2.38

Full specification table

SpecAMD Instinct MI325X 256GBNVIDIA RTX 3080 10GB
ArchitectureCDNA 3.5Ampere
VRAM256 GB HBM3e10 GB GDDR6X
VRAM typeHBM3eGDDR6X
Memory bandwidth6,000 GB/s760 GB/s
FP16 Tensor TFLOPS (dense)1,470 TFLOPS45 TFLOPS
TDP750 W320 W
TFLOPS/watt1.96 T/W0.14 T/W
GB/s per watt8.00 GB/s/W2.38 GB/s/W
NVLinkNoNo
Release year20242020

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 RTX 3080 10GB

FP16 TFLOPS45 TFLOPS
TDP320 W
TFLOPS/watt0.141 T/W
GB/s per watt2.38 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 RTX 3080 10GB for:

  • Entry-level AI
  • Small model inference
  • Hobbyist compute

Weaknesses / Limitations

  • 10 GB VRAM — only fits very small quantized models
  • Older Ampere architecture with lower TFLOPS than Ada equivalents
  • Not suitable for fine-tuning or production inference workloads