GPU servers, from training to inference
Dedicated GPU servers on a flat monthly rate.
Overview
GPU servers give you GPU capacity for training, inference and similar workloads.
Choose from a range of GPUs to match your workload.
GPU instances
GPU instances use MIG (Multi-Instance GPU) to share one GPU's resources among several users.
GPU instances are on the way. We haven't set a launch date yet.
Dedicated GPU servers
A dedicated GPU server gives you the whole GPU. Nothing is shared with other users, so there's no performance loss.
For machine learning and inference workloads only. This product can't be sublet to third parties.
Prices are in Korean won (KRW) and do not follow the currency setting in the header. VAT not included.
Frequently asked questions
It depends on the GPU model. Most servers are ready within 1 to 7 days.
Dedicated GPU servers normally come with one GPU each. If you need more, contact sales and we'll work out a multi-GPU build.
Yes. If the standard build doesn't fit, contact sales and we'll adjust it.