Compute Comparison
vs
All comparisons →

CDNA 4

AMD Instinct MI355X

wins

0

The AMD Instinct MI355X is AMD's CDNA 4 flagship.

Large LLM inferenceAgent workloadsMemory-bound AI

Blackwell Ultra

NVIDIA GB300 SXM6 288GB

wins

3

The GB300 SXM6 (Grace Blackwell Ultra) is a CPU+GPU superchip with 288 GB HBM3e and 8 TB/s memory bandwidth.

Frontier AI trainingInference at scaleMemory-bound LLM serving

Performance comparison

Visual bars — winner highlighted

AMD Instinct MI355X

Metric

NVIDIA GB300 SXM6 288GB

FP16 Tensor TFLOPS (dense)

2.5k T
2.8k T

Bandwidth

8.0k GB/s
8.0k GB/s

VRAM

288 GB
288 GB

TDP (lower=better)

1.4k W
1.2k W

TFLOPS/watt

1.79
2.33

GB/s per watt

5.71
6.67

Full specification table

SpecAMD Instinct MI355XNVIDIA GB300 SXM6 288GB
ArchitectureCDNA 4Blackwell Ultra
VRAM288 GB HBM3e288 GB HBM3e
VRAM typeHBM3eHBM3e
Memory bandwidth8,000 GB/s8,000 GB/s
FP16 Tensor TFLOPS (dense)2,500 TFLOPS2,800 TFLOPS
TDP1,400 W1,200 W
TFLOPS/watt1.79 T/W2.33 T/W
GB/s per watt5.71 GB/s/W6.67 GB/s/W
NVLinkNoYes
Release year20252025

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 MI355X

FP16 TFLOPS2,500 TFLOPS
TDP1,400 W
TFLOPS/watt1.786 T/W
GB/s per watt5.71 GB/s/W

NVIDIA GB300 SXM6 288GB

FP16 TFLOPS2,800 TFLOPS
TDP1,200 W
TFLOPS/watt2.333 T/W
GB/s per watt6.67 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 MI355X for:

  • Large LLM inference
  • Agent workloads
  • Memory-bound AI

Weaknesses / Limitations

  • ROCm ecosystem less mature than CUDA — some frameworks require porting
  • Very limited cloud availability compared to NVIDIA equivalents
  • 750W TDP — high power draw comparable to H100 SXM

Choose NVIDIA GB300 SXM6 288GB for:

  • Frontier AI training
  • Inference at scale
  • Memory-bound LLM serving

Weaknesses / Limitations

  • Very limited cloud availability — still ramping supply
  • Extremely high cost — enterprise/hyperscaler pricing only
  • Requires specialized NVLink-C2C infrastructure