Leaderboard Β· All comparisons Β· LynxBenchAI 26Q3 Index

NVIDIA GeForce RTX 4080 SUPER vs NVIDIA L40S for AI

Device pages: NVIDIA GeForce RTX 4080 SUPER Β· NVIDIA L40S

On the LynxBenchAI 26Q3 Index, the NVIDIA GeForce RTX 4080 SUPER scores 1.01x the NVIDIA L40S overall (Training 0.92x, Inference 1.04x, Compute 1.08x). It leads on Inference and Compute and trails on Training.

Scores on the LynxBenchAI 26Q3 Index

Ratio = the RTX 4080 SUPER's score divided by the L40S's
ScoreRatioRTX 4080 SUPERL40S
Overall (GT)1.01x398.74395.82
Training0.92x291.00316.04
Inference1.04x551.54528.12
Compute1.08x370.70344.33

When to Choose Which

  • Level on training: the two are within 10% (0.92x).
  • Level on inference: the two are within 10% (1.04x).
  • Level on raw compute: the two are within 10% (1.08x).

From the LynxBenchAI 26Q3 Index category scores, which are not raw throughput; a lead under 10% counts as level.

Per Model and Precision

LynxBenchAI score per model and precision; ratio = RTX 4080 SUPER / L40S
ModelPrecisionRatioRTX 4080 SUPERL40S
compute
Dense MatMulBF161.45x352.69243.02
Dense MatMulFP161.42x347.27243.74
Dense MatMulFP321.05x318.83303.04
Dense MatMulFP640.56x4.878.68
Sparse MatMulFP321.04x599.74575.74
Sparse MatMulFP640.86x495.92577.96
inference
DistilBERTFP161.03x921.58898.53
DistilBERTFP81.25x645.88514.72
DistilBERTINT81.16x506.40438.15
DistilGPT2FP161.09x817.72752.18
DistilGPT2FP81.10x504.17457.31
DistilGPT2INT81.09x458.99420.32
Flan-T5 SmallFP161.20x1216.481009.63
Flan-T5 SmallFP81.35x763.76566.59
Flan-T5 SmallINT80.94x472.69503.01
GraphGPS (Peptides)FP160.96x256.57268.17
GraphGPS (Peptides)FP80.93x128.56137.60
GraphGPS (Peptides)INT80.93x127.77138.11
MobileNetV3FP160.88x1205.331375.69
MobileNetV3INT80.82x330.60403.39
training
DistilBERTBF160.96x389.61405.03
DistilBERTFP321.16x571.22492.41
DistilGPT2BF160.92x205.47223.21
DistilGPT2FP321.00x390.78391.29
Flan-T5 SmallBF160.83x218.55263.33
Flan-T5 SmallFP320.90x423.07470.80
GraphGPS (Peptides)BF160.91x168.65184.43
GraphGPS (Peptides)FP320.77x159.03207.50
MobileNetV3BF160.87x215.50248.23
MobileNetV3FP320.86x323.35376.00

Related Comparisons

Each figure is the first device's overall score divided by the second's.

Measure your own machine: pip install lynxbench-ai Β· Install guide β†’

How to Read These Results

A LynxBenchAI score is sustained throughput on a fixed catalogue of PyTorch models, measured after a warm-up at a batch size raised until throughput stops improving, so it is what the machine holds under load, not its best instant. Training, Inference and Compute are three separate scores; GT combines them (Training and Inference weigh 0.4 each, Compute 0.2).

Every run uses the stock, publicly released PyTorch stack, so a result is broadly indicative of the out-of-the-box PyTorch experience on that hardware, and so of custom PyTorch models too. It does not run your own model, and it is not a vendor-tuned peak like an MLPerf submission.

Scores compare only within one Index: the 26Q3 Index fixes the models, precisions, correctness thresholds and scoring, and a later Index may change them. Each test is one timed window, the scores have no 0-100 scale, and a precision a device cannot run scores zero rather than being skipped.

LynxBenchAI is built by TechnoLynx, which the European Commission engaged as an independent expert advisor on AI performance and benchmarking for AI Gigafactories under the EuroHPC Joint Undertaking (case study).

Questions

Is the NVIDIA GeForce RTX 4080 SUPER faster than the NVIDIA L40S for AI?
On the LynxBenchAI 26Q3 Index, the NVIDIA GeForce RTX 4080 SUPER scores 1.01x the NVIDIA L40S overall (Training 0.92x, Inference 1.04x, Compute 1.08x). It leads on Inference and Compute and trails on Training.
Should I choose the NVIDIA GeForce RTX 4080 SUPER or the NVIDIA L40S?
Level on training: the two are within 10% (0.92x). Level on inference: the two are within 10% (1.04x). Level on raw compute: the two are within 10% (1.08x). These follow the LynxBenchAI 26Q3 Index category scores, which are not raw throughput.
Where is the difference between the NVIDIA GeForce RTX 4080 SUPER and the NVIDIA L40S largest?
On the LynxBenchAI 26Q3 Index, the ratio is highest on Dense MatMul at BF16 (Compute), where the NVIDIA GeForce RTX 4080 SUPER scores 1.45x the NVIDIA L40S, and lowest on Dense MatMul at FP64 (Compute), at 0.56x.
What does this result tell me about my own PyTorch model?
Every run uses the stock, publicly released PyTorch stack, so a result is broadly indicative of the out-of-the-box PyTorch experience on that hardware, and so of custom PyTorch models too. It does not run your own model, and it is not a vendor-tuned peak like an MLPerf submission.

The Personal Edition submits your benchmark results to TechnoLynx's servers as part of participating in the public leaderboard. If that isn't acceptable for your use case, contact TechnoLynx about Press, Pro, or Enterprise licenses.

Get Started