Accel 96
Accel 96 is a dedicated NVIDIA RTX PRO 6000 Blackwell with 96 GB of GDDR7 ECC memory, on a 24-core Xeon with 512 GB of memory and two 3.84 TB NVMe drives. We deliver it with drivers, container runtime and an inference server already running, so a 70B-parameter model in 4-bit or a 30B model in full precision serves requests the day the server is handed over.
No setup fee. Monthly billing, three-month minimum, then cancel with 30 days notice. Delivery within 48 hours.
- Processor
- Intel Xeon Gold 5412U
- Cores
- 24 cores / 48 threads
- Memory
- 512 GB DDR5 ECC
- Storage
- 2 × 3.84 TB NVMe (mirrored)
- GPU
- NVIDIA RTX PRO 6000 Blackwell, 96 GB
- Network
- 10 Gbit, 20 TB included
- Response
- Engineer within 15 min
Datasheet
Exact configuration as delivered. Variants within a series are quoted individually.
- GPU
- NVIDIA RTX PRO 6000 Blackwell Max-Q, 96 GB GDDR7 ECC, 5th-generation Tensor Cores
- Processor
- Intel Xeon Gold 5412U, 24 cores / 48 threads
- Memory
- 512 GB DDR5 ECC registered
- Storage
- 2 × 3.84 TB NVMe SSD, datacentre edition, software mirror
- Software
- CUDA, NVIDIA container toolkit and vLLM or Triton installed; your choice of model pulled before handover
- Network
- 10 Gbit uplink, 20 TB outbound per month included, then €1 per TB
- Location
- Germany (Falkenstein) or Finland (Helsinki)
- Monitoring
- GPU utilisation, memory, temperature and token throughput on your dashboard
- Backups
- Nightly encrypted snapshots of system and model store to a second site, 30-day retention
- Incident response
- Critical issues answered by an engineer within 15 minutes, around the clock
What Accel 96 is for.
Workloads this configuration has been sized for, and what it adds beyond the standard layer.
Best for
- Serving open-weight LLMs to your own product
- Fine-tuning with LoRA on private data
- Embedding and reranking pipelines at volume
- Rendering, video and 3D pipelines that bill by the hour elsewhere
Included beyond the standard layer
- Inference server configured and load-tested before handover
- Monthly GPU utilisation report with right-sizing advice
Included with every package.
Nothing on this list is an upsell.
Other packages.
- Core 48Core
48 Zen 4 cores, 256 GB and fast NVMe for your main application tier.
- CPU
- 48 cores / 96 threads
- Memory
- 256 GB DDR5 ECC
- Storage
- 2 × 3.84 TB NVMe (mirrored)
€2,540per month - Store 300Store
308 TB of raw disk behind 15 TB of NVMe cache, run as object or block storage.
- CPU
- 32 cores / 64 threads
- Memory
- 256 GB DDR4 ECC
- Storage
- 14 × 22 TB HDD + 2 × 7.68 TB NVMe
€3,290per month - Enterprise 48Enterprise
Core 48 compute in an enterprise chassis with redundant power and four hot-swap NVMe bays.
- CPU
- 48 cores / 96 threads
- Memory
- 256 GB DDR5 ECC
- Storage
- 4 × 3.84 TB NVMe (hot-swap)
€3,860per month - Core 48 MaxCore
512 GB of memory and 30 TB of NVMe for in-memory data and large databases.
- CPU
- 48 cores / 96 threads
- Memory
- 512 GB DDR5 ECC
- Storage
- 2 × 15.36 TB NVMe (mirrored)
€4,120per month - Cluster 3Cluster
Three 48-core nodes on a private network with highly available Kubernetes and Postgres.
- CPU
- 144 cores / 288 threads total
- Memory
- 384 GB DDR5 ECC total
- Storage
- 6 × 1.92 TB NVMe total
€9,850per month
Ready to run Accel 96? Tell us about the workload.
A quote within one business day. €6,780 per month with everything included, no setup fee.