Option comparison · GPU Server

Managed or Root, Standard or Advanced: your choice

Our GPU Servers are available in several variants. Managed GPU Server and Root GPU Server use the same hardware platform, with Nvidia GPUs in the GPU Standard or GPU Advanced configurations. This page compares the options in detail, so you can find the right configuration for your budget and workload.

Managed or Root GPU Standard or GPU Advanced 4 Nvidia GPU models
Jump to
Which one when?

Two management models, the same hardware platform

Managed GPU Server

“We don’t want to look after the server ourselves.”

  • For teams without system administration knowledge
  • We take care of installation and configuration, security updates, monitoring and backups
  • One web and one database service are included in the base price
  • Managed GPU Server →
Root GPU Server

“We want our own CUDA, driver and framework stack.”

  • For experienced system administrators
  • Full root access, we only install the operating system
  • Replacement within one business day in the event of a hardware defect
  • Root GPU Server →
Management model

Managed or Root: who manages the server?

Managed GPU Server and Root GPU Server use the same hardware platform. The same performance in a different management model.

Managed GPU ServerRoot GPU Server
Target audienceFor teams without system administration knowledgeExperienced system administrators
Server managementMonitoring, firewall, 24/7 emergency service and support includedYour own responsibility, we only install the operating system
BackupDaily, encrypted and stored at two separate locations, kept for 7 days by default. A longer retention period is available for a fee.Not included, your own responsibility
Operating systemUbuntu LTSUbuntu LTS, other Linux distributions can be self-installed on request via our temporary IP-KVM service
Service Level Agreement (SLA)Basic (included), higher tiers optionalNot available
Also includedOne web and one database serviceNo Managed Services. Optional and charged separately: our On-Demand Services
Price GPU StandardCHF 865.– per month, one-time setup fee CHF 800.–CHF 705.– per month, one-time setup fee CHF 700.–
Price GPU AdvancedCHF 1'409.– per month, one-time setup fee CHF 800.–CHF 1'249.– per month, one-time setup fee CHF 700.–

Prices excluding VAT and excluding the GPU model (see the GPU models table).

Server class

GPU Standard or GPU Advanced: which base fits?

Both server classes are available as Managed GPU Server and as Root GPU Server. They differ in CPU, storage and compatible GPU models.

GPU StandardGPU Advanced
CPUAMD EPYC 7763P CPU, 64 cores / 128 threads, 2.45 GHz base / 3.5 GHz max clockAMD EPYC 9554 CPU, 64 cores / 128 threads, 3.1 GHz base / 3.75 GHz max clock
RAM64 GB DDR4 (max. 2 TB)64 GB DDR5 (max. 2 TB)
Storage2x 960 GB NVMe SSD included, software RAID-1 (expandable up to 4x NVMe and 8x SATA)No drives included in the price (up to 4x NVMe and 4x SATA or 8x SATA configurable)
Compatible GPU modelsNvidia H100 NVL, L40s, RTX 4500 ADANvidia RTX Pro 6000 Blackwell Server Edition
GPU slots24
GPU models

Which Nvidia model fits your requirements?

The GPU is booked in addition to the server class. H100 NVL, L40s and RTX 4500 ADA are available for GPU Standard, the RTX Pro 6000 Blackwell Server Edition exclusively for GPU Advanced.

Nvidia H100 NVLNvidia L40sNvidia RTX 4500 ADANvidia RTX Pro 6000 Blackwell
VRAM94 GB48 GB24 GB96 GB
Memory Bandwidth3,938 GB/s864 GB/s432 GB/s1,792 GB/s
Tensor Cores456568240752
CUDA Cores14,59218,1767,68024,064
TFLOPS (FP32)6091.639.6125
Price per monthCHF 3'999.–CHF 999.–CHF 399.–CHF 1'749.–
One-time fee per GPU cardCHF 3'999.–CHF 999.–CHF 399.–CHF 1'749.–
Available forGPU StandardGPU StandardGPU StandardGPU Advanced

Prices per GPU, excluding VAT, in addition to the chosen server class, identical for Managed and Root GPU Servers. For longer commitments (6 months, 1 year, 2 years or more), contact us for an individual quote. Full, up-to-date pricing in the Price Calculator.

FAQ

Frequently asked questions about the GPU Server options

What does «Managed» mean for a GPU Server?
With a Managed GPU Server, we take care of monitoring, backups, firewall, 24/7 emergency service and support. One web and one database service are also included in the price.
What does «Root» or «unmanaged» mean?
With a Root GPU Server, we install the operating system, everything else (configuration, updates, services) is your responsibility. Root systems are aimed at experienced system administrators.
How many GPUs can I get per server?
The GPU Standard model holds 2 GPUs, the GPU Advanced model 4. On request, we equip your server with up to the maximum number of GPUs for the chosen server class, as well as up to 2 TB RAM, 4x NVMe drives and 8x SATA drives.
Which GPU model can I combine with GPU Advanced?
The Nvidia RTX Pro 6000 Blackwell Server Edition is available exclusively for GPU Advanced, as Managed GPU Server or Root GPU Server. H100 NVL, L40s and RTX 4500 ADA are intended for GPU Standard.
Is GPU access also available on virtual servers?
Managed and Root GPU Servers are always dedicated systems, provisioned exclusively for you. GPU functions are not available on our virtualised servers.
Where are the GPU Servers operated?
Our GPU Servers are operated in the NTT Zürich 1 data centre in Rümlang.
What contract terms apply?
The minimum contract term is 3 months. After that, the contract can be cancelled with one month’s notice.

Technical details in our documentation for Managed Server and Root Server.

One provider

One provider, from rack to app.

From colocation to Root and Managed Products, platforms and services: all with us, in our Swiss data centres. Grow from one layer to the next without switching providers.

One Swiss teamGrow without switching providers

Which GPU fits your workload?

We put the configuration together with you.

Talk to our experts

Do you have questions about our products? No problem. Contact our sales department or go directly to your Cockpit to see all the possibilities at Nine.

Subscribe to our newsletter for the latest updates: