A data center GPU against a consumer card, one Ada Lovelace and one Blackwell: 48 GB GDDR6 against 32 GB GDDR7 and 2.1x the memory bandwidth on the RTX 5090.
At a glance
Memory
- L40S
- 48 GB
- RTX 5090
- 32 GB
1.5x more on L40S
Memory bandwidth
- L40S
- 864 GB/s
- RTX 5090
- 1,792 GB/s
2.1x higher on RTX 5090
Cheapest on-demand
- L40S
- $0.55 /GPU/hr
- RTX 5090
- $0.35 /GPU/hr
36% cheaper on RTX 5090
Price comparison
Median price per GPU per hour by billing type, with the cheapest listing under it.
| Billing | L40S | RTX 5090 | Difference |
|---|---|---|---|
| On-demand | $1.50 /GPU/hr from $0.55 | $0.65 /GPU/hr from $0.35 | 2.3x more expensive on L40S |
| Reserved | $1.19 /GPU/hr from $0.60 (12 mo) | $0.57 /GPU/hr from $0.21 (1 mo) | 2.1x more expensive on L40S |
| Spot | $0.90 /GPU/hr from $0.40 | -- /GPU/hr from $0.11 |
Heads up: A provider's own page may quote a different figure, for example on tax, region or a promotion. A monthly-only plan shows a derived hourly rate. Medians are across providers, one vote each, so a large catalog does not outweigh a small one. No median for spot on the RTX 5090: we only show one when at least 3 providers list the GPU on that billing type. Verify before provisioning. More on how we price.
Price history
L40S vs RTX 5090 price history
Weekly median on-demand price per GPU per hour
Specs comparison
| Spec | L40S | RTX 5090 | Difference |
|---|---|---|---|
| Memory | |||
| Capacity | 48 GB GDDR6 | 32 GB GDDR7 | 1.5x more on L40S |
| Bandwidth | 864 GB/s | 1,792 GB/s | 2.1x higher on RTX 5090 |
| Bus width | 384-bit | 512-bit | 1.3x wider on RTX 5090 |
| Compute · dense | |||
| FP4 | Not supported | 1,676 TFLOPS | RTX 5090 only |
| FP8 | 733 TFLOPS | 419 TFLOPS | 1.7x higher on L40S |
| FP16 / BF16 | 362.1 TFLOPS | 209.5 TFLOPS | 1.7x higher on L40S |
| INT8 | 733 TOPS | 838 TOPS | 1.1x higher on RTX 5090 |
| FP32 | 91.6 TFLOPS | 104.8 TFLOPS | 1.1x higher on RTX 5090 |
| Platform | |||
| Interconnect | PCIe only | PCIe only | |
| Board power | 350 W | 575 W | 1.6x higher on RTX 5090 |
| Released | Q3 2023 | Q1 2025 | 2 years newer on RTX 5090 |
| Precision | |||
| Supported formats | FP8FP16BF16TF32FP32INT4INT8 | FP4FP8FP16BF16TF32FP32INT8 | |
Compute: Vendor peak figures, dense and per GPU. Not measured throughput.
Source: L40S datasheet and RTX 5090 datasheet.
Where to rent them
Providers listing these GPUs, those with both first, so you can switch without moving clouds.
| Provider | L40S /GPU/hr | RTX 5090 /GPU/hr |
|---|---|---|
Vast.ai
|
On-Demand $0.74 | On-Demand $0.40 |
GPU.ai
|
On-Demand $0.79 | On-Demand $0.40 |
Runpod
|
On-Demand $0.79 | On-Demand $0.69 |
Runcrate
|
On-Demand $0.97 | On-Demand $0.77 |
Novita
|
On-Demand $0.55 | -- |
Packet·ai
|
On-Demand $0.92 | -- |
Sesterce
|
On-Demand $0.97 | -- |
Massed Compute
|
On-Demand $0.97 | -- |
HyperAI
|
-- | On-Demand $0.35 |
Lium
|
-- | On-Demand $0.48 |
Salad
|
-- | On-Demand $0.50 |
Nova Cloud
|
-- | On-Demand $0.63 |
35 providers list the L40S and 20 the RTX 5090. Every listing is on the L40S page and the RTX 5090 page.
Similar GPUs
Other comparisons in the same class.
-
Nvidia H100
80 GB HBM3 · 3.9x the bandwidth of the L40S
From $1.30 /GPU/hr Compare -
Nvidia A100
80 GB HBM2e · 2.2x the bandwidth of the L40S
From $0.43 /GPU/hr Compare
-
Nvidia L40
48 GB GDDR6 · released Q4 2022
From $0.46 /GPU/hr Compare
-
Nvidia RTX PRO 6000
96 GB GDDR7 · workstation card
From $0.66 /GPU/hr Compare
-
Nvidia RTX 4090
24 GB GDDR6X · 44% less bandwidth than the RTX 5090
From $0.23 /GPU/hr Compare
Our data for Nvidia L40S was last updated on Sept. 14, 2026, and for Nvidia RTX 5090 on Sept. 14, 2026.