Compute Comparison
vs
All comparisons →

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

Ada Lovelace

NVIDIA L40S

wins

1

The L40S is NVIDIA's inference-optimized Ada Lovelace GPU. 48GB GDDR6 and 362 TFLOPS FP16 make it a strong choice for production inference at lower cost than H100.

InferenceVideo AIMulti-modal workloads

Performance comparison

Visual bars — winner highlighted

NVIDIA GB300 SXM6 288GB

Metric

NVIDIA L40S

FP16 Tensor TFLOPS (dense)

2.8k T
362 T

Bandwidth

8.0k GB/s
864 GB/s

VRAM

288 GB
48 GB

TDP (lower=better)

1.2k W
350 W

TFLOPS/watt

2.33
1.03

GB/s per watt

6.67
2.47

Full specification table

SpecNVIDIA GB300 SXM6 288GBNVIDIA L40S
ArchitectureBlackwell UltraAda Lovelace
VRAM288 GB HBM3e48 GB GDDR6
VRAM typeHBM3eGDDR6
Memory bandwidth8,000 GB/s864 GB/s
FP16 Tensor TFLOPS (dense)2,800 TFLOPS362 TFLOPS
TDP1,200 W350 W
TFLOPS/watt2.33 T/W1.03 T/W
GB/s per watt6.67 GB/s/W2.47 GB/s/W
NVLinkYesNo
Release year20252023

Live cloud pricing

On-demand pricing across providers — updated in real time

Per hourPer GB VRAM/hr

NVIDIA GB300 SXM6 288GB

No live pricing data available.

NVIDIA L40S

Canopy Wave
$0.99/hr
TheAI Cloud
$0.79/hr
HIGHRESO
$1.09/hr
E2E Networks
$1.20/hr
Lightning AI
$2.14/hr
Hugging Face
$1.80/hr
FluidStack
$5.20/hr
AceCloud
$1.83/hr

Salad: Lowest priority tier — preemptible

Power efficiency analysis

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

NVIDIA GB300 SXM6 288GB

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

NVIDIA L40S

FP16 TFLOPS362 TFLOPS
TDP350 W
TFLOPS/watt1.034 T/W
GB/s per watt2.47 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 GB300 SXM6 288GB

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

NVIDIA L40S

Best cloud price$0.38/hr
TFLOPS per $952.6 T/$
GB/s per $2274 GB/s/$

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 GB300 SXM6 288GBNVIDIA L40S
7B LLM fine-tune8× GPU, 24 h—$73
70B LLM fine-tune64× GPU, 72 h—$1.8k
Inference server1× GPU, 730 h—$277
Image gen batch4× GPU, 8 h—$12

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 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

Choose NVIDIA L40S for:

  • Inference
  • Video AI
  • Multi-modal workloads

Weaknesses / Limitations

  • GDDR6 bandwidth (864 GB/s) is lower than HBM-based GPUs — bottleneck for memory-bound LLM serving
  • No NVLink — cannot scale across cards for very large models
  • Less suitable for large-scale training than H100 or A100