Bare-metal servers with dedicated NVIDIA GPUs for model training, inference, rendering and other GPU-bound workloads — no shared VRAM, no noisy neighbours.
| Model | GPU | vCPU / RAM | Storage | Price | |
|---|---|---|---|---|---|
| GPU-T4Inference & light training | 1× NVIDIA T4 16GB | 8 vCPU / 32 GB | 500 GB NVMe | $189/mo | Order |
| GPU-A10Fine-tuning & rendering | 1× NVIDIA A10 24GB | 16 vCPU / 64 GB | 1 TB NVMe | $389/mo | Order |
| GPU-A100Model training at scale | 1× NVIDIA A100 80GB | 32 vCPU / 128 GB | 2 TB NVMe | $1,249/mo | Order |
| GPU-2xA100Multi-GPU training | 2× NVIDIA A100 80GB | 64 vCPU / 256 GB | 4 TB NVMe | $2,349/mo | Order |
Full CUDA/cuDNN support, pre-imaged with common ML frameworks on request.
Dedicated VRAM means predictable latency, without contention from other tenants.
GPU-accelerated rendering pipelines for 3D, VFX and video transcoding workloads.
Hourly or monthly billing options for teams that need burst capacity, not a permanent cluster.