AI & GPU Servers
Dedicated GPU-server options for LLM, image generation, RAG and data-processing workloads. GPU model, inventory, driver/framework versions and capacity are confirmed with a representative benchmark.
Plans
Billing period, discount and delivery model are stated on each plan card.
- NVIDIA RTX A4000 (16 GB GDDR6)
- AMD Ryzen 9 5950X (16c/32t)
- 128 GB DDR4 ECC RAM
- 2× 1.92 TB NVMe SSD
- CUDA + PyTorch + TF preinstalled
- NVIDIA RTX A5000 (24 GB GDDR6)
- AMD Threadripper PRO 5965WX (24c/48t)
- 256 GB DDR4 ECC RAM
- 4× 1.92 TB NVMe SSD
- LLM/SDXL fit verified by benchmark
- NVIDIA L40S (48 GB GDDR6, Ada)
- AMD EPYC 7543 (32c/64t)
- 256 GB DDR4 ECC RAM
- 4× 3.84 TB NVMe SSD · 10 Gbps
- LLM and multimodal workloads · benchmark required
- NVIDIA A100 PCIe (40 GB HBM2e)
- AMD EPYC 7763 (64c/128t)
- 512 GB DDR4 ECC RAM
- 8× 3.84 TB NVMe SSD · 25 Gbps
- MIG (Multi-Instance GPU) support
- NVIDIA H100 PCIe (80 GB HBM3)
- AMD EPYC Genoa (96c/192t)
- 1 TB DDR5 ECC · 100 Gbps
- Transformer Engine + FP8
- Large-model workloads · benchmark and topology design
Why AIOR AI Servers?
Total cost between hourly cloud GPU and monthly dedicated GPU depends on utilisation, data transfer, storage, management and commitment terms. The GPU model and software image available from AIOR are stated in the quote after current inventory and project compatibility are confirmed; not every model is assumed to be in stock at all times.
Which GPU for which workload?
GPU selection cannot be made from model parameter count alone. Weight precision and quantisation, context length, KV cache, batch size, training/inference method, multi-GPU topology and latency target all change memory needs. The candidate workload is benchmarked first, then the appropriate GPU, VRAM and scaling plan are selected.
Plan Comparison
| Feature | AI Workstation | AI Pro | AI Studio (L40S) | AI Enterprise (A100) | AI Top (H100) |
|---|---|---|---|---|---|
| GPU | |||||
| GPU model | RTX A4000 | RTX A5000 | L40S | A100 | H100 |
| GPU memory | 16 GB GDDR6 | 24 GB GDDR6 | 48 GB GDDR6 | 40 GB HBM2e | 80 GB HBM3 |
| CPU & RAM | |||||
| CPU | Ryzen 9 5950X | TR PRO 5965WX | EPYC 7543 | EPYC 7763 | EPYC Genoa |
| vCore | 16c/32t | 24c/48t | 32c/64t | 64c/128t | 96c/192t |
| RAM (ECC) | 128 GB | 256 GB | 256 GB | 512 GB | 1 TB DDR5 |
| Network & Storage | |||||
| Storage | 2× 1.92 TB NVMe | 4× 1.92 TB NVMe | 4× 3.84 TB NVMe | 8× 3.84 TB NVMe | 16× 3.84 TB NVMe Gen4 |
| Port speed | 1 Gbps | 1 Gbps | 10 Gbps | 25 Gbps | 100 Gbps |
Frequently Asked Questions
Hourly or monthly rental?
What's the minimum GPU rental period?
Which frameworks come preinstalled?
I want multiple GPU nodes?
Which LLM models can I run?
Ready to get started?
Requests accepted 24/7 · standard support Mon–Fri 09:00–18:00 · critical on-call only under the applicable product or written SLA