Compute Comparison
vs
All comparisons →

Hopper

NVIDIA H200 SXM 141GB

wins

4

The H200 SXM is the SXM5 form factor of the H200, with 141 GB HBM3e and 4.8 TB/s bandwidth.

Very large LLMsMemory-bound inferenceMulti-modal models

Turing

NVIDIA T4

wins

1

The NVIDIA T4 is the most widely deployed inference GPU in the cloud.

Low-cost inferenceEdge AIMulti-tenant serving

Performance comparison

Visual bars — winner highlighted

NVIDIA H200 SXM 141GB

Metric

NVIDIA T4

FP16 Tensor TFLOPS (dense)

989 T
65 T

Bandwidth

4.8k GB/s
320 GB/s

VRAM

141 GB
16 GB

TDP (lower=better)

700 W
70 W

TFLOPS/watt

1.41
0.93

GB/s per watt

6.86
4.57

Full specification table

SpecNVIDIA H200 SXM 141GBNVIDIA T4
ArchitectureHopperTuring
VRAM141 GB HBM3e16 GB GDDR6
VRAM typeHBM3eGDDR6
Memory bandwidth4,800 GB/s320 GB/s
FP16 Tensor TFLOPS (dense)989 TFLOPS65 TFLOPS
TDP700 W70 W
TFLOPS/watt1.41 T/W0.93 T/W
GB/s per watt6.86 GB/s/W4.57 GB/s/W
NVLinkYesNo
Release year20242018

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

NVIDIA H200 SXM 141GB

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

NVIDIA T4

FP16 TFLOPS65 TFLOPS
TDP70 W
TFLOPS/watt0.929 T/W
GB/s per watt4.57 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 H200 SXM 141GB for:

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

Weaknesses / Limitations

  • Very limited cloud availability — fewer providers than H100
  • Premium pricing over H100 with same FP16 TFLOPS
  • 700 W TDP — same power envelope as H100 with no efficiency gain

Choose NVIDIA T4 for:

  • Low-cost inference
  • Edge AI
  • Multi-tenant serving

Weaknesses / Limitations

  • 16 GB VRAM limits model size to ~13B at INT4
  • Low FP16 TFLOPS (65) — not suitable for throughput-sensitive serving
  • Older Turing architecture — no Ada or Hopper efficiency improvements