GPUcloud
FRSign up

Custom infrastructure

Custom GPU infrastructure

NVIDIA HGX B200 servers, multi-GPU configurations and HPC clusters designed for your workload. Answer the technical questionnaire: our architecture team designs the infrastructure and replies by email.

Lead times

Some builds are ready in 7 business days; B200 servers and large configurations can take up to a month depending on the build, the hardware and the stock. We confirm the order and our teams work to an agreed schedule.

Answer what you know and choose “Not sure / Recommend” for the rest.

GPU Cloud – Custom GPU Infrastructure Technical Questionnaire

Step 1 of 12

1. GPU & Server Configuration

How many GPUs are needed in total?

1.1 How many GPUs are needed in total?

How many GPUs would you like per server? (4, 8 or other)

1.2 How many GPUs would you like per server? (4, 8 or other)

How many GPU servers are needed?

1.3 How many GPU servers are needed?

Do you need NVLink / NVSwitch between the GPUs?

1.4 Do you need NVLink / NVSwitch between the GPUs?

Must all the GPUs work together or independently?

1.5 Must all the GPUs work together or independently?

Do you need dedicated GPUs with exclusive access?

1.6 Do you need dedicated GPUs with exclusive access?

Do you have specific VRAM requirements?

1.7 Do you have specific VRAM requirements?

Do you need an architecture that allows GPU servers to be added later?

1.8 Do you need an architecture that allows GPU servers to be added later?