EPYC 9654 + NVIDIA accelerator
FUSION
Whole GPUs, passed through
Inference, fine-tuning, rendering, transcoding
from€132.30/mo
€189−30% annual
01 — Sizes
Pick a size
Six lines, one price list, every location. Switch families later without rebuilding: we migrate the volume live.
Fleet discounts stack on top: 3% from three servers, 6% from five, 10% from ten, applied to the whole invoice.
G1 · RTX 4000 Ada
- 8 vCPU Dedicated physical cores
- 32 GB RAM
- 500 GB NVMe Gen5
- 40 TB · 10 Gbps
Geekbench 6 multi
18,400 · 620k IOPS · 13,100 MB/s
€132.30/mo
€189Save €680.40
G2 · RTX 4090 24 GB
- 12 vCPU Dedicated physical cores
- 64 GB RAM
- 1 TB NVMe Gen5
- 60 TB · 10 Gbps
Geekbench 6 multi
27,600 · 780k IOPS · 13,900 MB/s
€265.30/mo
€379Save €1,364.40
G3 · L40S 48 GB
- 16 vCPU Dedicated physical cores
- 128 GB RAM
- 2 TB NVMe Gen5
- Unmetered · 25 Gbps
Geekbench 6 multi
36,200 · 940k IOPS · 14,600 MB/s
€524.30/mo
€749Save €2,696.40
G4 · H100 80 GB NVL
- 24 vCPU Dedicated physical cores
- 256 GB RAM
- 4 TB NVMe Gen5
- Unmetered · 25 Gbps
Geekbench 6 multi
52,800 · 1180k IOPS · 14,900 MB/s
€1,323/mo
€1,890Save €6,804
02 — What you get
Full specification
Full PCIe passthrough, no vGPU slicing
CUDA 12 and ROCm images ready
NVLink pairs on request
Annual commitment, thirty percent off
Single-core, Geekbench 6
Geekbench 6 single-core, median of 40 runs on production nodes.
03 — Where
FUSION · 8 locations
Whole GPUs, passed through
04 — Alternatives