NovaOtterBack to Solutions

Recommended Starting Point

CPUDual Intel Xeon or AMD EPYC

GPU4x or 8x NVIDIA A100, H100, H200, or RTX

MemoryUp to 8 TB

CoolingDatacenter air cooling

Configuration Choices

Processor

Select CPU capacity around GPU count, memory bandwidth, and scheduler needs.

2x 16-core AMD EPYCEfficient dual-socket baseline for a GPU server.+NZ$02x 64-core AMD EPYCMore CPU throughput for data staging and multi-user queues.+NZ$9,4092x 96-core AMD EPYCHigh-core-count CPU path for larger server workloads.+NZ$11,7772x 128-core AMD EPYCMore CPU headroom for preprocessing and concurrent jobs.+NZ$16,4192x Intel Xeon GoldIntel server CPU option for teams standardizing on Xeon platforms.+NZ$2,2402x Intel Xeon PlatinumHigh-end Intel CPU path for dense datacenter deployments.+NZ$37,020

Graphics Card

Choose accelerator count and memory class around training, inference, or simulation requirements.

4x NVIDIA A100 40 GBBase HBM GPU configuration for shared AI compute.+NZ$08x NVIDIA A100 40 GBDouble the A100 count for larger scheduled workloads.+NZ$87,1724x NVIDIA H100 94 GBHigher-memory H100 path for modern training and inference.+NZ$171,0148x NVIDIA H100 94 GBEight-GPU H100 profile for large shared queues.+NZ$516,3704x NVIDIA H200 141 GBH200 option for larger VRAM and memory bandwidth needs.+NZ$242,4618x NVIDIA H200 141 GBMaximum H200 capacity for demanding datacenter workloads.+NZ$568,764

Memory

Size ECC memory relative to GPU VRAM and host-side preprocessing.

4x 16 GB DDR5 ECC/REGEntry host-memory configuration.+NZ$04x 32 GB DDR5 ECC/REGMore RAM for data loaders and shared users.+NZ$4,2928x 64 GB DDR5 ECC/REGBalanced memory profile for multi-GPU jobs.+NZ$11,58812x 96 GB DDR5 ECC/REGLarge host-memory profile for heavier preprocessing.+NZ$70,78812x 128 GB DDR5 ECC/REGMaximum listed memory class for dense server jobs.+NZ$115,163

Operating System

Choose the deployment image for AI frameworks, orchestration, or your own imaging process.

Ubuntu AI software stackLinux stack prepared for CUDA, containers, model serving, and schedulers.+NZ$0Windows ServerWindows Server license path for datacenter administration.+NZ$3,946Windows 11 ProDesktop-style Windows environment for mixed tooling.+NZ$376No operating systemFor teams that image systems internally.+NZ$0

Primary SSD

Use fast local storage for the OS, models, checkpoints, and active datasets.

4 TB PCIe 4.0 NVMe SSDBase high-speed system and working-data drive.+NZ$07.68 TB PCIe 4.0 NVMe SSDMore local room for models and project data.+NZ$3,79115.36 TB PCIe 4.0 NVMe SSDLarge local dataset and checkpoint drive.+NZ$10,72330.72 TB PCIe 4.0 NVMe SSDVery large local NVMe capacity.+NZ$21,42161.44 TB PCIe 4.0 NVMe SSDHigh-capacity NVMe path for large data staging.+NZ$46,664

Additional SSD

Add a secondary storage tier when local datasets exceed the primary drive.

Not includedUse only the primary SSD.+NZ$015.36 TB PCIe 4.0 NVMe SSDAdditional high-speed local dataset volume.+NZ$13,17630.72 TB PCIe 4.0 NVMe SSDLarger secondary NVMe capacity.+NZ$23,87461.44 TB PCIe 4.0 NVMe SSDVery large secondary NVMe storage tier.+NZ$44,123122.88 TB PCIe 4.0 NVMe SSDMaximum listed local storage expansion.+NZ$81,863

RAID Controller

Use a RAID controller only when redundancy or specific storage topology requires it.

Not includedRun SSDs without an added RAID controller.+NZ$04-port NVMe RAID controllerHardware RAID option for up to four NVMe SSDs.+NZ$1,5528-port NVMe RAID controllerLarger NVMe RAID topology for storage-heavy builds.+NZ$5,0868-port SAS/SATA RAID controllerRAID controller for SATA or SAS storage arrays.+NZ$2,755

Network

Plan fabric speed around shared storage, cluster scheduling, and east-west traffic.

10 Gb dual-port EthernetIntegrated baseline network for the server.+NZ$025 Gb dual-port EthernetBetter fit for shared storage and datacenter uplinks.+NZ$344100 Gb dual-port InfiniBandHigh-throughput fabric for GPU clusters and storage nodes.+NZ$3,400200 Gb dual-port InfiniBandFaster fabric path for dense multi-node workloads.+NZ$3,502400 Gb InfiniBandHighest listed network option for cluster-scale deployments.+NZ$4,834

Trusted Platform Module

Add hardware-backed platform security if your deployment policy requires it.

Not includedNo additional TPM module.+NZ$0TPM 2.0 moduleHardware security module for supported fleet policies.+NZ$146

UPS System

Add rack power protection when the server will run outside an existing UPS-backed rack.

Not includedUse existing rack or facility power protection.+NZ$01500 W rackmount UPSEntry rackmount UPS for a smaller power envelope.+NZ$2,3183000 W rackmount UPSHigher-capacity rackmount UPS option.+NZ$4,5926000 W UPS systemLarger UPS option for heavier GPU configurations.+NZ$9,990

Warranty/Support

Choose the support window for the server lifecycle.

3 year limited warrantyBase service plan for labor and part replacement.+NZ$04 year limited warrantyExtended support window for production deployments.+NZ$7,5965 year limited warrantyLongest listed support term for critical systems.+NZ$11,394

Why this configuration path

Rack GPU server for centralized training and inference.

Focused on 4x/8x GPU and NVLink-class server options.

Useful when multiple users need scheduled GPU capacity.