Cloudmersive Private Cloud AI Server is the common AI base platform used by Cloudmersive AI APIs. It is required when utilizing most Cloudmersive AI APIs.

Cloudmersive Private Cloud AI Server has specific hardware requirements to run:

• CPU: 4 Cores Minimum
• RAM: 128 GB Minimum
• GPU: 1 NVIDIA L40 or L40S 48 GB GPU RAM Minimum
• Disk: 500 GB SSD
• Operating System: Linux - Red Hat Enterprise Linux (RHEL) 10, Debian 11, or Ubuntu Server 24.04

Supported NVIDIA GPU families include L40/L40S, RTX 6000 / RTX PRO 6000, A100, H100, and H200.

For faster performance, customers can consider increasing the GPU to:

• GPU: 1 NVIDIA RTX 6000 / RTX PRO 6000 48 GB+ GPU RAM

For fastest performance and throughput, customers can increase the GPU to:

• GPU: 1 NVIDIA A100 80 GB, H100 80 GB+, or H200 141 GB GPU RAM

These guidelines also apply to Cloudmersive Managed Instance.

Cloud Deployment Guidelines

Microsoft Azure

• Standard_NC144ds_xl_RTXPRO6000BSE_v6 — 144 vCPU, 516 GB RAM, 1× RTX PRO 6000 96 GB
• Standard_NC40ads_H100_v5 — 40 vCPU, 320 GiB RAM, 1× H100 94 GB
• Standard_ND96isr_H200_v5 — 96 vCPU, 1,850 GiB RAM, 8× H200 141 GB
• Standard_NC24ads_A100_v4 — 24 vCPU, 220 GiB RAM, 1× A100 80 GB

Amazon Web Services

• g6e.4xlarge — 16 vCPU, 128 GiB RAM, 1× L40S 48 GB
• g7e.4xlarge — 16 vCPU, 128 GiB RAM, 1× RTX PRO 6000 96 GB
• p5.4xlarge — 16 vCPU, 256 GiB RAM, 1× H100 80 GB
• p5e.48xlarge / p5en.48xlarge — 192 vCPU, 2 TiB RAM, 8× H200 141 GB

Google Cloud Platform

• g4-standard-48 — 48 vCPU, 180 GB RAM, 1× RTX PRO 6000 96 GB
• a3-highgpu-1g — 26 vCPU, 234 GB RAM, 1× H100 80 GB
• a3-ultragpu-8g — 224 vCPU, 2,952 GB RAM, 8× H200 141 GB
• a2-ultragpu-1g — 12 vCPU, 170 GB RAM, 1× A100 80 GB

Oracle Cloud Infrastructure

• BM.GPU.L40S.4 — 112 OCPUs (~224 vCPU), 1,024 GB RAM, 4× L40S 48 GB
• BM.GPU.RTXPRO.8 — 144 OCPUs (~288 vCPU), 3,072 GB RAM, 8× RTX PRO 6000 96 GB
• BM.GPU.H100.8 — 112 OCPUs (~224 vCPU), 2,048 GB RAM, 8× H100 80 GB
• BM.GPU.H200.8 — 112 OCPUs (~224 vCPU), 3,072 GB RAM, 8× H200 141 GB
• BM.GPU.A100-v2.8 — 128 OCPUs (~256 vCPU), 2,048 GB RAM, 8× A100 80 GB

On-Premises Server Guidelines

Dell

• PowerEdge R660 (1U)
• PowerEdge R760 (2U)
• PowerEdge R760xa

HPE

• ProLiant DL320 Gen11 (1U)
• ProLiant DL380 Gen11 (2U)
• ProLiant DL380 Gen10 Plus (2U)
• ProLiant DL580 Gen10 (4U)

Frequently Asked Questions

Are GPUs from other manufacturers (e.g. AMD, Google TPU, etc.) supported?

Not at this time. Currently Cloudmersive Private Cloud requires NVIDIA GPUs due to CUDA and other architectural features.

Are multiple GPUs supported?

Yes, you can either scale up by having multiple GPUs in one server, or scale out by having multiple servers each with 1 GPU. We recommend a symmetrical deployment, i.e. all servers have the same hardware configuration.

Can different size GPUs be used for pre-production and production?

Yes, you can use more cost-efficient GPUs in pre-production (e.g. L40) and higher power GPUs in production (e.g. RTX 6000, A100, H100, or H200).