NovoServe has recently secured a large allocation of NVIDIA RTX PRO 6000 GPUs, widely considered the most capable accelerator for modern professional workloads. We currently hold multiple units in stock and ready for deployment, with physical hardware fully configurable to meet your specific architectural demands.
The RTX PRO 6000 delivers 96GB of memory per card, making it ideal for hosting large language models entirely in memory. Operating on dedicated bare metal ensures full data privacy, strict sovereignty, and zero shared hypervisor overhead. Our month-to-month contracts remove the financial risk of long-term lock-in while giving you absolute physical control over your infrastructure.
You can build your GPU servers with single, dual, or even quad RTX Pro 6000 GPU cards. As you can fully configure your GPU bare metal servers with NovoServe, below are some of the recommended configs for you:
Processor: 2x Intel Xeon Gold 6122 (Total: 40 Cores, 1.8 GHz)
Memory: 128 GB RAM
Storage: 2x 480GB SSD (OS) + 2x 1920GB NVMe (Data)
GPU: 1x NVIDIA RTX PRO 6000 96GB Max Q (Total: 96GB GPU Memory)
Included Uplink: 2 Gbps Dedicated Unmetered
Data Center: Amsterdam, NL
Monthly List Price: €1,700.00 EUR
This entry-level high-performance configuration delivers a massive 96GB of VRAM on a single physical GPU. It provides the exact memory footprint required to easily host large language models in memory, including unquantized 70B parameter deployments. Operating on entirely dedicated hardware guarantees you receive raw bare-metal performance with zero shared hypervisor overhead. For service providers, this chassis serves as a reliable foundation to build and sell virtual private servers, custom VMs, or dedicated AI agents to end-users.
Processor: 2x AMD EPYC 7702 (Total: 128 Cores, 2.0 GHz)
Memory: 256 GB RAM
Storage: 2x 480GB SSD (OS) + 2x 1920GB NVMe (Data)
GPU: 2x NVIDIA RTX PRO 6000 96GB Max Q (Total: 192GB GPU Memory)
Included Uplink: 5 Gbps Dedicated Unmetered
Data Center: Amsterdam, NL
Monthly List Price: €3,500.00 EUR
Scaling parallel computing requires hardware architectures that prevent storage delays before data ever reaches the GPU. Dual AMD EPYC 7702 processors deliver 128 physical cores, while dual 1920GB NVMe drives ensure fast I/O read rates match your processing throughput. You gain 192GB of total GPU memory, establishing this machine as a highly capable node for continuous inference tasks and dense multi-tenant virtualization. Backed by a dedicated 5 Gbps unmetered uplink, you can ingest massive active datasets continuously without triggering unexpected bandwidth overage penalties.
Processor: 2x AMD EPYC 7702 (Total: 128 Cores, 2.0 GHz)
Memory: 512 GB RAM
Storage: 2x 480GB SSD (OS) + 2x 1920GB NVMe (Data)
GPU: 4x NVIDIA RTX PRO 6000 96GB Max Q (Total: 384GB GPU Memory)
Included Uplink: 10 Gbps Dedicated Unmetered
Data Center: Amsterdam, NL
Monthly List Price: €6,000.00 EUR
At 384GB of pooled GPU memory, this unit is a tier-one production node built for complex AI inference and managing heavy East-West AI agent traffic. The hardware is paired with a dedicated 10 Gbps unmetered network connection to answer the core traffic asymmetry challenges outlined in recent Cisco AI network research. As generative flow volumes expand rapidly, pushing terabytes of data through public cloud environments incurs massive financial egress penalties. This fully dedicated setup provides predictable, flat-rate economics for relentless, high-bandwidth compute operations.
According to the 2026 Cisco report, "AI Impact on Wide Area Networks," bandwidth is rapidly becoming the primary challenge in modern AI infrastructure. The network acts as the digital spinal cord connecting data streams to compute nodes, meaning connection latency directly harms agentic AI inference speeds. NovoServe bypasses this bottleneck entirely utilizing a massive 18 Tbps network uplink capacity in the Netherlands. You can read more about why connectivity is the definitive constraint in our latgest article: Connectivity is the new challenge in mordern AI infrastructure.
Do you want build your GPU bare metal servers with RTX Pro 6000 GPU cards? Speak to our infrastructure experts, and let's discuss your build.
