Intel Gaudi2
vs
Nvidia L40S
Two data center GPUs, one Gaudi (2nd Gen) and one Ada Lovelace: 96 GB HBM2e against 48 GB GDDR6, 2.8x the memory bandwidth on the Gaudi2, and RoCE 600 GB/s on the Gaudi2 where the L40S is PCIe only.
At a glance
Memory
- Gaudi2
- 96 GB
- L40S
- 48 GB
2x more on Gaudi2
Memory bandwidth
- Gaudi2
- 2,450 GB/s
- L40S
- 864 GB/s
2.8x higher on Gaudi2
Cheapest on-demand
- Gaudi2
- $1.03 /GPU/hr
- L40S
- $0.38 /GPU/hr
2.7x more expensive on Gaudi2
Price comparison
Median price per GPU per hour by billing type, with the cheapest listing under it.
| Billing | Gaudi2 | L40S | Difference |
|---|---|---|---|
| On-demand | $1.21 /GPU/hr from $1.03 | $1.50 /GPU/hr from $0.38 | 19% cheaper on Gaudi2 |
| Reserved | -- /GPU/hr from $0.58 (12 mo) | $1.11 /GPU/hr from $0.60 (12 mo) | |
| Spot | No spot listings | $1.11 /GPU/hr from $0.74 |
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 reserved on the Gaudi2: 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
Gaudi2 vs L40S price history
Weekly median on-demand price per GPU per hour
Specs comparison
| Spec | Gaudi2 | L40S | Difference |
|---|---|---|---|
| Memory | |||
| Capacity | 96 GB HBM2e | 48 GB GDDR6 | 2x more on Gaudi2 |
| Bandwidth | 2,450 GB/s | 864 GB/s | 2.8x higher on Gaudi2 |
| Bus width | 384-bit | ||
| Compute · dense | |||
| FP8 | 865 TFLOPS | 733 TFLOPS | 1.2x higher on Gaudi2 |
| FP16 / BF16 | 432 TFLOPS | 362.1 TFLOPS | 1.2x higher on Gaudi2 |
| INT8 | 733 TOPS | ||
| FP32 | 91.6 TFLOPS | ||
| Platform | |||
| Interconnect | RoCE 600 GB/s | PCIe only | RoCE on Gaudi2 only |
| Board power | 600 W | 350 W | 1.7x higher on Gaudi2 |
| Released | Q2 2022 | Q3 2023 | 15 months newer on L40S |
| Precision | |||
| Supported formats | FP8FP16BF16TF32FP32INT8 | FP8FP16BF16TF32FP32INT4INT8 | |
Compute: Vendor peak figures, dense and per GPU. Not measured throughput.
Source: Gaudi2 datasheet and L40S datasheet.
Where to rent them
Providers listing these GPUs, those with both first, so you can switch without moving clouds.
| Provider | Gaudi2 /GPU/hr | L40S /GPU/hr |
|---|---|---|
Runcrate
|
On-Demand $1.21 | On-Demand $0.97 |
Sesterce
|
On-Demand $1.21 | On-Demand $0.97 |
Cyfuture AI
|
On-Demand $1.03 | -- |
Lium
|
-- | On-Demand $0.38 |
Novita
|
-- | On-Demand $0.55 |
EmpirioLabs AI
|
-- | On-Demand $0.80 |
GPU.ai
|
-- | On-Demand $0.80 |
Packet·ai
|
-- | On-Demand $0.92 |
Massed Compute
|
-- | On-Demand $0.97 |
3 providers list the Gaudi2 and 35 the L40S. Every listing is on the Gaudi2 page and the L40S page.
Similar GPUs
Other comparisons in the same class.
-
Nvidia H100
From $1.45 /GPU/hr Compare -
Nvidia A100
From $0.80 /GPU/hr Compare -
AMD MI300X
From $2.39 /GPU/hr Compare
Our data for Intel Gaudi2 was last updated on Sept. 7, 2026, and for Nvidia L40S on Sept. 7, 2026.