L40 vs L40S
L40 starts at $0.79/GPU-hour on-demand; L40S starts at $0.97/GPU-hour on-demand. Compare the exact variants, memory and published specifications below. Peak compute figures do not establish application speed.
L40
L40S
Spec sheet
| L40 | L40S | |
|---|---|---|
| Spec variant | L40 | L40S |
| Architecture | Ada Lovelace | Ada Lovelace |
| VRAM (spec variant) | 48 GB | 48 GB |
| FP16 TFLOPS (dense) | 181 | 362 |
| Memory bandwidth | 864 GB/s | 864 GB/s |
| TDP | 300 W | 350 W |
| Interconnect | PCIe Gen4 | PCIe Gen4 |
| Launched | 2022 | 2023 |
Rental price per manufacturer-rated FP16 compute: L40 $4.36 per PFLOP-hour vs L40S $2.68. Formula: cheapest $/hr ÷ dense FP16 PFLOPS (1 PFLOP = 1000 TFLOPS). This uses L40 and L40S, whose datasheets may differ from the variants above. It is not measured job throughput. Specs: L40 manufacturer, L40S manufacturer.
What they cost to rent (2026-10-02)
| L40 | L40S | |
|---|---|---|
| Cheapest on-demand $/hr | $0.79 | $0.97 |
| Cheapest spot/community $/hr | $0.69 | $0.79 |
| Cheapest source | Thunder Compute | Massed Compute |
| Providers offering it | 6 | 16 |
Lowest recorded on-demand rate per configuration; highlighted = cheaper. Spot/marketplace tiers shown on the individual pages.
FAQ
L40 is cheaper: the L40 rents from $0.79 per GPU-hour and the L40S from $0.97 on-demand, updated 2026-10-02.
The largest recorded L40 variant has 48 GB and the largest L40S variant has 48 GB. Other variants may have less memory. They match on maximum VRAM.
Compare more rental options: the screener · price movers · methodology.