Compute Comparison
vs
All comparisons →

Ampere

NVIDIA A100 40GB

wins

1

The A100 40GB is the more affordable sibling of the A100 80GB.

ML trainingInferenceCost-effective HPC

Blackwell Ultra

NVIDIA GB300 SXM6 288GB

wins

4

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

NVIDIA A100 40GB

Metric

NVIDIA GB300 SXM6 288GB

FP16 Tensor TFLOPS (dense)

312 T
2.8k T

Bandwidth

1.6k GB/s
8.0k GB/s

VRAM

40 GB
288 GB

TDP (lower=better)

400 W
1.2k W

TFLOPS/watt

0.78
2.33

GB/s per watt

3.89
6.67

Full specification table

SpecNVIDIA A100 40GBNVIDIA GB300 SXM6 288GB
ArchitectureAmpereBlackwell Ultra
VRAM40 GB HBM2e288 GB HBM3e
VRAM typeHBM2eHBM3e
Memory bandwidth1,555 GB/s8,000 GB/s
FP16 Tensor TFLOPS (dense)312 TFLOPS2,800 TFLOPS
TDP400 W1,200 W
TFLOPS/watt0.78 T/W2.33 T/W
GB/s per watt3.89 GB/s/W6.67 GB/s/W
NVLinkYesYes
Release year20202025

Live cloud pricing

On-demand pricing across providers — updated in real time

Per hourPer GB VRAM/hr

NVIDIA A100 40GB

Paperspace
$7.90/hr
Replicate
$0.90/hr
E2E Networks
$1.98/hr
Northflank
$1.42/hr
Lightning AI
$2.19/hr
Guanxingyun
$3.58/hr
FluidStack
$5.60/hr
Sesterce
$1.10/hr

NVIDIA GB300 SXM6 288GB

No live pricing data available.

Power efficiency analysis

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

NVIDIA A100 40GB

FP16 TFLOPS312 TFLOPS
TDP400 W
TFLOPS/watt0.780 T/W
GB/s per watt3.89 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.

Performance-per-dollar

TFLOPS/$ and GB/s/$ at current best cloud prices

NVIDIA A100 40GB

Best cloud price$0.29/hr
TFLOPS per $1075.9 T/$
GB/s per $5362 GB/s/$

NVIDIA GB300 SXM6 288GB

Best cloud priceN/A
TFLOPS per $—
GB/s per $—

TFLOPS/$ = FP16 TFLOPS ÷ best on-demand $/hr. Higher is better — more compute per dollar. GB/s/$ = memory bandwidth ÷ best on-demand $/hr. For memory-bound inference workloads (most LLM serving), GB/s/$ is often a better predictor of cost-efficiency than TFLOPS/$. Prices are live from the Compute Comparison API and update in real time.

Workload Cost Comparison

Total job cost = best on-demand $/hr × GPU count × hours

WorkloadScaleNVIDIA A100 40GBNVIDIA GB300 SXM6 288GB
7B LLM fine-tune8× GPU, 24 h$56—
70B LLM fine-tune64× GPU, 72 h$1.3k—
Inference server1× GPU, 730 h$212—
Image gen batch4× GPU, 8 h$9—

Based on current best on-demand prices. Spot pricing typically 40–70% lower. Actual cost depends on provider, region, and availability.

When to choose each GPU

Choose NVIDIA A100 40GB for:

  • ML training
  • Inference
  • Cost-effective HPC

Weaknesses / Limitations

  • 40 GB VRAM is a hard ceiling — cannot run models that require 80 GB
  • Lower bandwidth than A100 80GB (1,555 vs 2,039 GB/s)
  • Older Ampere architecture — lower TFLOPS than H100

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