AxForge offering · Request
Systems with four NVIDIA RTX 3090 cards, 24 GB GDDR6X each, for workloads that need dedicated EU-hosted GPU capacity without the newest silicon. Add a public request for dedicated capacity. Registered demand is used to prioritize new capacity, and machines are sized to the workloads that asked for them.
Currently request-only. Add your request to the public forum; this is not a booking or a promised launch.

Why AxForge
| 24 GB at ~936 GB/s | Capacity that newer consumer cards at this price point don't match: 20 to 35B quantized models on a single card, or four cards for dev, test and volume work. |
|---|---|
| Still ahead of newer 16 GB cards | On the same-workload llama.cpp CUDA scoreboard the RTX 3090 decodes at 158.2 t/s with 5,175 t/s prompt processing, about 2.1× an RTX 3060. Community measurements, not ours. |
| A modern 35B MoE on one card | A community benchmark ran Qwen3.6-35B-A3B on a single RTX 3090: ~140 t/s fully in VRAM, ~89 t/s with full long context via partial CPU offload. On 24 GB, the offload configuration decides the result. |
Specifications
| System | Four NVIDIA RTX 3090 cards in one dedicated machine, the budget tier |
|---|---|
| Memory | 24 GB GDDR6X per card |
| Availability | Request capacity: register your requirement in the console |
| Tenancy | Dedicated machine, your traffic only, full SSH access |
| Price per GPU-hour | €0.62 /GPU-hour, launch price |
| Region | European region, placement confirmed with your request. |
| Community decode | 158.2 t/s single-card (llama.cpp CUDA scoreboard, Llama 2 7B Q4_0). Labelled, not ours |
RTX 3090 catalogue reference: specifications and requirements. List prices in EUR.
How it works
| 1 | Confirm capacity and terms in the console before the rental steps below. A public request records demand; it does not reserve or provision a machine. |
|---|---|
| 2 | Sign in to AxForge. |
| 3 | Choose RTX 3090 and describe your requirement: model sizes, traffic, term. |
| 4 | AxForge scopes the 4-GPU configuration with you; deployment needs a separate agreement; a public request is not a booking. |
| 5 | Add your SSH key and connect when the machine is yours. |
| 6 | End the rental when your workload is complete. |
Request GPU or sign in to start.
FAQ
Capacity and terms must be confirmed before a booking. Request dedicated capacity for this GPU through the public forum. Registered demand is used to prioritize new capacity, and machines are sized to the workloads that asked for them.
It is our budget tier: 24 GB GDDR6X per card is real capacity for smaller models, dev and test nodes, and workloads that do not need the newest silicon, all on a dedicated EU-hosted machine.
Four NVIDIA RTX 3090 cards with 24 GB GDDR6X each, rented as one dedicated machine serving only your traffic.
Capacity and terms must be confirmed before a booking. It is the entry tier for dedicated EU-hosted hardware at AxForge: the launch price is listed per GPU-hour, and deployment needs a separate agreement; a public request is not a booking. Register your requirement and AxForge scopes the configuration and price with you in EUR. If 128 GB of unified memory matters more than raw decode speed, look at the DGX Spark.
Capacity and terms must be confirmed before a booking. In an AxForge European region. GPU rentals run on our machines in Stockholm (eu-se-1) or Málaga, Spain (eu-es-1); placement would be agreed before any booking.
No. Your model, your traffic, our hardware. Prompts are never persisted. Only operational metadata is kept for billing and operations.