General purpose instances featuring full passthrough or an affordable fraction of a NVIDIA L40S GPUs.
The NVIDIA L40S GPU is a cutting-edge solution designed to accelerate AI, graphics, and compute workloads across various industries. Featuring 48 GB of GDDR6 ECC memory, it provides ample capacity for handling massive datasets, enabling efficient training and inference of advanced AI models. Built on NVIDIA's Ada Lovelace architecture, the L40S integrates next-generation CUDA cores, Tensor Cores, and RT Cores to deliver superior performance for AI, machine learning, and deep learning tasks.
| Name | NVMe Storage | Memory | vCPUs | Transfer | Price | Regions |
|---|---|---|---|---|---|---|
G1-gpu-l40s-180m-16c-48g
| 1200 GiB | 180 GB | 16 vCPUs | 10 TB | $2.3397 /hr $1 708 /mo | Sold out |
G1-gpu-l40s-375m-32c-96g
| 2200 GiB | 375 GB | 32 vCPUs | 10 TB | $4.6795 /hr $3 416 /mo | Sold out |
G1-gpu-l40s-750m-64c-192g
| 2600 GiB | 750 GB | 64 vCPUs | 15 TB | $9.3589 /hr $6 832 /mo | Sold out |
G1-gpu-l40s-1500m-128c-384g
| 3400 GiB | 1500 GB | 128 vCPUs | 25 TB | $18.7178 /hr $13 664 /mo | Sold out |
Each NVIDIA L40S unit features 48 GB of GDDR6 ECC memory, with a single physical server capable of hosting up to 8 GPUs. The g value in the plan name indicates the amount of available GDDR6 memory, measured in gigabytes, allocated to each plan. G1 GPU plans are primarily available in the following regions: New Jersey (108), Frankfurt (109), and Bangalore (103).