Flopper.io Flopper.io

NVIDIA L4 24GB vs NVIDIA L40S 48GB

The L40S delivers 3.0x the FP8 throughput of the L4 (733 vs 243 TFLOPS dense). The L40S also carries 24 GB more memory (48 GB vs 24 GB).

L4: Ada Lovelace, 2023 L40S: Ada Lovelace, 2023
FP8 dense lead
3.0x
L40S: 733 vs 243 TFLOPS
Memory
24 vs 48 GB
24 GB more for the L40S
Bandwidth
300 GB/s vs 864 GB/s
L40S moves data faster
TDP
72 W vs 350 W
L4 draws 278 W less
See rental pricing from $0.36/hr Download comparison sheet

Share this page

NVIDIA L4 24GB vs NVIDIA L40S 48GB

A post with this link shows the card on the Image tab as its preview.

Post on X Share on LinkedIn
 Printable version →

NVIDIA L4 24GB

Full specs →

What is the NVIDIA L4?

The L4 is a GPU from NVIDIA on the Ada Lovelace architecture, an accelerator for AI training, inference and high-performance computing. It is a PCIe add-in card, so it fits standard servers and workstations. L4 full specifications →

Can I rent the L4 in the cloud?

Yes. Cloud providers tracked by Flopper rent the L4 by the hour, and the pricing page compares their current rates and the price history side by side. L4 rental prices →

How does the L4 compare with other accelerators?

This page sets the L4 against the L40S. Both parts launched in the same year. Both are NVIDIA parts, and both ship as a PCIe card, so they fit the same class of system. Flopper also compares it head to head with the Inferentia2, across every published precision, memory and power. L4 vs Inferentia2 →

Where do the L4 figures come from?

Every L4 figure on this page traces to NVIDIA's own datasheet or specification table. How Flopper verifies specifications →

NVIDIA L40S 48GB

Leads on FP8 Full specs →

What is the NVIDIA L40S?

The L40S is a GPU from NVIDIA on the Ada Lovelace architecture, an accelerator for AI training, inference and high-performance computing. It is a PCIe add-in card, so it fits standard servers and workstations. L40S full specifications →

Can I rent the L40S in the cloud?

Yes. Cloud providers tracked by Flopper rent the L40S by the hour, and the pricing page compares their current rates and the price history side by side. L40S rental prices →

How does the L40S compare with other accelerators?

This page sets the L40S against the L4. Both parts launched in the same year. Both are NVIDIA parts, and both ship as a PCIe card, so they fit the same class of system. Flopper also compares it head to head with the TPU v5e 16GB and ATOM-Max, across every published precision, memory and power. L40S vs TPU v5e 16GB →

Where do the L40S figures come from?

Every L40S figure on this page traces to NVIDIA's own datasheet or specification table. How Flopper verifies specifications →

Specifications

L4
L40S
Architecture
Ada LovelaceAda Lovelace
Launch Year
20232023
Form Factor
PCIePCIe
Memory
24 GB48 GB
Memory Bandwidth
300 GB/s864 GB/s
TDP
72 W350 W
Process Node
4nm4nm

Performance (TFLOPS)

L4
L40S
0.4 TFLOPS 1.4 TFLOPS
30 TFLOPS 92 TFLOPS
60 TFLOPS
120 TFLOPS sparse
183 TFLOPS
366 TFLOPS sparse
121 TFLOPS
242 TFLOPS sparse
362 TFLOPS
733 TFLOPS sparse
121 TFLOPS
242 TFLOPS sparse
362 TFLOPS
733 TFLOPS sparse
243 TFLOPS
485 TFLOPS sparse
733 TFLOPS
1,466 TFLOPS sparse
FP6
No verified data available No verified data available
No verified data available No verified data available
243 TOPS
485 TOPS sparse
733 TOPS
1,466 TOPS sparse

Figures are dense. Where a vendor publishes a figure it sits under the dense one and never decides a winner. Hover any precision for its definition, or read the GPU glossary.

FLOPS by Precision

What actually differs

Both launched in 2023, the L4 on Ada Lovelace and the L40S on Ada Lovelace.

The L40S carries 24 GB more memory (48 GB against 24 GB), so larger models fit on one GPU before you shard. Bandwidth is 300 GB/s against 864 GB/s, which sets the ceiling on decode speed.

Power is 72 W against 350 W, and at FP32 the L4 does more work per watt. Weight the FP8 and FP16 rows for LLM work and the FP64 row for HPC.

New to the terms? The glossary explains FP64, FP32, TF32, BF16, FP16, FP8, INT8, Sparsity and TFLOPS. For how the generations relate, see NVIDIA architectures explained.

Frequently asked questions

Is the L4 faster than the L40S?

At FP8 precision the L40S reaches 733 TFLOPS dense against 243 TFLOPS for the L4. The performance table on this page lists every published precision for both GPUs.

Which should you choose, the L4 or the L40S?

Choose the L40S for training and high-throughput inference: it has 3.0x the FP8 throughput of the L4. It also carries 24 GB more memory, so larger models fit on one GPU before you shard. The L40S draws 278 W more power, so budget for cooling and power delivery before you commit.

Which has more memory, the L4 or the L40S?

The L40S carries 48 GB of memory versus 24 GB for the L4. Memory bandwidth is 300 GB/s for the L4 and 864 GB/s for the L40S.

How much power do the L4 and the L40S draw?

The L4 is rated at 72 W TDP and the L40S at 350 W. On FP32 throughput per watt, the L4 is the more efficient part.

How much do the L4 and the L40S cost to rent?

Live cloud listings tracked by Flopper price the L4 from $0.36 per GPU-hour across 10 providers, and the L40S from $0.60 per GPU-hour across 23 providers. The rental section on this page links to every provider, rate and the price history for each GPU.

Where to Rent

NVIDIA L4 24GB

Rent from

$0.36 / GPU-hr

Lowest of 18 live listings across 10 providers · median $0.80

See all L4 rental prices →

Price history and every provider on the pricing page. Some links are affiliate links.

NVIDIA L40S 48GB

Rent from

$0.60 / GPU-hr

Lowest of 40 live listings across 23 providers · median $1.46

See all L40S rental prices →

Price history and every provider on the pricing page. Some links are affiliate links.

Get Comparison Updates

New GPUs added weekly. Be the first to see how they compare.

Open in Full Compare Tool → L4 full specs L40S full specs Estimate cluster cost in the calculator →

© 2026 Flopper.io - Compare the GPUs Powering AI