NTS Elite APEX 5U Dual EPYC 8-GPU Server

MPN G593-ZD2-AAX1
HPC/AI Server - AMD EPYC™ 9004 - 5U DP NVIDIA HGX™ H100 8-GPU 4-Root Port Powering the Next Generation of Server Architecture and Energy Efficiency

Key capabilities

  • NVIDIA-Certified Systems™ - Data Center Servers
  • NVIDIA HGX™ H100 8-GPU
  • 900GB/s GPU-to-GPU bandwidth with NVIDIA NVLink™ and NVSwitch™
  • Dual AMD EPYC™ 9005/9004 Series Processors
  • 12-Channel DDR5 RDIMM, 24 x DIMMs
  • Dual ROM Architecture
  • 2 x 10Gb/s LAN ports via Intel® X710-AT2
  • 2 x M.2 slots with PCIe Gen3 x4 and x1 interface
  • 8 x 2.5" Gen5 NVMe/SATA/SAS-4 hot-swap bays
  • 4 x FHHL PCIe Gen5 x16 slots
  • 8 x LP PCIe Gen5 x16 slots
  • 5+1 3000W 80 PLUS Titanium redundant power supplies
Part#: APXA015U8H1


Product
NTS Elite APEX 5U Dual EPYC 8-GPU Server
Part number
G593-ZD2-AAX1
Category
Systems › GPU Systems
Form factor
5U rackmount
Typical workloads
AI training, inference, and HPC
Configuration
Engineer-to-order via Configure (CPU, memory, GPU, storage, networking).
Vendor
RackmountNTS (New Tech Solutions, Inc.), Fremont, California, USA.
Facility
ISO-aligned staging, imaging, and burn-in at NTS HQ in Fremont, CA.

NTS Elite APEX 8U Dual AMD EPYC 8-GPU Server


The NTS Elite APEX 8U Dual AMD EPYC 8-GPU Server is an enterprise GPU computing platform designed for demanding AI, deep learning, high-performance computing (HPC), and accelerated data center workloads. Built around a dual AMD EPYC processor architecture and supporting up to eight GPUs, this system provides high-density compute resources for organizations running parallel processing, large-scale model development, and GPU-intensive applications.
Configured and quoted engineer-to-order by New Tech Solutions (RackMountNTS), the system can be tailored to workload requirements. Use Configure to select compatible processors, memory, GPU accelerators, storage, and networking. NTS reviews the proposed configuration for platform compatibility, power delivery, cooling requirements, firmware, and drivers before staging and shipment from Fremont, California.

Key Benefits


  • Eight-GPU computing capacity: Supports configurations with up to eight GPU accelerators for parallel computing, AI development, deep learning, and other GPU-intensive workloads.
  • Dual AMD EPYC processing: Provides multi-core CPU resources to support data preparation, orchestration, virtualization, and demanding enterprise applications alongside GPU acceleration.
  • High-density compute architecture: An 8U rackmount platform accommodates a high-performance GPU configuration for data centers and dedicated AI infrastructure.
  • Scalable memory and storage configuration: Configure memory and storage options to suit model training, large datasets, high-throughput processing, and application requirements, subject to platform support.
  • High-speed networking options: Select compatible Ethernet or InfiniBand connectivity to support data movement, distributed computing, and cluster integration.
  • Engineer-to-order configuration: NTS reviews the complete system configuration, including accelerator compatibility, power, thermal requirements, firmware, and drivers.
  • Deployment and lifecycle support: Receive a staged system with configuration review, serial-tracked acceptance, and documentation to support deployment and ongoing operations.

GPU Availability and Configuration


The NTS Elite APEX 8U platform is intended for configurations with up to eight GPU accelerators. GPU selection depends on the supported accelerator list and the platform's mechanical, electrical, thermal, power, and firmware requirements.
NTS can review the target GPU models and proposed quantities to determine a compatible configuration. GPU options should be confirmed during the BOM review rather than assumed from the platform's maximum GPU capacity. Where applicable, networking and interconnect choices can also be reviewed for the intended multi-GPU or cluster deployment.

Key Applications and Use Cases


  • AI model training and development: GPU-accelerated workloads for training and developing machine learning and deep learning models.
  • Generative AI and large language models: Compute infrastructure for model development, fine-tuning, and other accelerator-intensive AI workloads, depending on GPU memory and software requirements.
  • AI inference: Accelerated inference environments for enterprise AI applications and production workloads.
  • High-performance computing (HPC): Parallel processing and scientific computing applications that benefit from multiple GPUs and high-performance CPU resources.
  • Engineering and scientific simulation: GPU-accelerated workloads used in computational research, simulation, and technical analysis.
  • Data analytics: Large-scale data processing and analytics workloads that use CPU and GPU resources.
  • Rendering and visualization: GPU-intensive rendering, visualization, and media-processing applications.
  • Research and development: Dedicated compute infrastructure for universities, laboratories, research organizations, and enterprise innovation teams.

Engineer-to-Order Configuration and Validation


NTS provides an engineer-to-order process for configuring the system around the intended workload. Use Configure to select CPU, memory, GPU accelerators, storage, and networking options. NTS reviews the proposed BOM for component compatibility, power requirements, cooling, firmware, and driver readiness.
Before shipment, systems are staged and validated at the NTS facility in Fremont, California. Commercial customers receive the same configuration review, staging discipline, serial-tracked acceptance, and lifecycle support path.
Federal, defense, and SLED programs can align the same BOM with SEWP V, ITES-4H, GSA MAS, or cooperative purchasing vehicles when permitted by the applicable contract and statement of work.
Share your workload requirements, target GPU and CPU counts, preferred fabric (Ethernet or InfiniBand), rack power budget, and delivery timeline through contact NTS or Request a Quote. NTS can return a reviewed configuration with lead-time information, dual-source notes when required, and documentation suitable for procurement and contracting review.

Related NTS Resources



After selecting compatible options in Configure, pricing updates in real time. Request a Quote to receive a configuration package with lead times and acceptance documentation aligned to your receiving process. Gold SKUs ship as fixed platforms, while configurable systems remain engineer-to-order and follow the same NTS Fremont staging and validation process.
Whether you are deploying a single AI development node or expanding a multi-rack GPU computing environment, NTS provides a consistent engineering and configuration process across processors, accelerators, storage, and networking. This helps maintain configuration control as infrastructure scales from initial deployment to larger production environments.

Product overview

Highlights and workloads for NTS Elite APEX 5U Dual EPYC 8-GPU Server.

Overview

HPC/AI Server - AMD EPYC™ 9004 - 5U DP NVIDIA HGX™ H100 8-GPU 4-Root Port Powering the Next Generation of Server Architecture and Energy Efficiency
Dimensions (WxHxD, mm)5U
447 x 219.7 x 945
MotherboardMZB3-G43
CPUAMD EPYC™ 9005 Series Processors
AMD EPYC™ 9004 Series Processors

Dual processor, cTDP up to 400W [1]
- At 25°C ambient, cTDP up to 500W

[1] Please refer to the QVL on the support page or contact our sales representatives for verified CPU SKUs.

[Note] If only 1 CPU is installed, some PCIe or memory functions might be unavailable.
Socket2 x LGA 6096
Socket SP5
ChipsetSystem on Chip
Memory24 x DIMM slots
Support DDR5 RDIMM
12-Channel memory per processor

AMD EPYC™ 9005:
Up to 6400 MT/s

AMD EPYC™ 9004:
Up to 4800 MT/s
LANFront:
2 x 10Gb/s LAN (1 x Intel® X710-AT2)
- Support NCSI function

1 x 10/100/1000 Mbps Management LAN

Rear (MLAN board - CDB66):
1 x 10/100/1000 Mbps Management LAN

[Note] When both MLAN ports are connected with cables, the front MLAN port will be set as the default.
VideoIntegrated in ASPEED® AST2600
- 1 x VGA port
StorageFront hot-swap:
8 x 2.5" Gen5 NVMe/SATA/SAS-4 [1]
- (NVMe from PEX89104)

Internal M.2:
1 x M.2 (2280/22110), PCIe Gen3 x4, from CPU_1
1 x M.2 (2280/22110), PCIe Gen3 x1, from CPU_0

[1] Storage card is required to support SATA and SAS drives.
SASRequire SAS add-in cards
RAIDRequire RAID add-in cards
Modular GPUNVIDIA HGX™ H100 with 8 x SXM5 GPUs
PCIe Expansion SlotsExtension Board CPBG044 x 2:
- 8 x LP x16 (Gen5 x16), from PEX89104

PCIe Cable x 4:
- 2 x FHHL x16 (Gen5 x16), from CPU_0
- 2 x FHHL x16 (Gen5 x16), from CPU_1
Front I/O2 x USB 3.2 Gen1 ports (Type-A)
1 x VGA port
2 x RJ45 ports
1 x MLAN port (default)
1 x Power button with LED
1 x ID button with LED
1 x NMI button
1 x Reset button
1 x Storage activity LED
1 x System status LED
Rear I/OMLAN board - CDB66:
1 x MLAN port
Backplane BoardSpeed and bandwidth:
PCIe Gen5 x4 or SATA 6Gb/s or SAS-4 24Gb/s
Security Modules1 x TPM header with SPI interface
- Optional TPM2.0 kit: CTM010
Power Supply[#1]5+1 3000W 80 PLUS Titanium redundant power supplies [1]

AC Input:
- 100-120V~/ 16A, 50/60Hz
- 200-240V~/ 16A, 50/60Hz

DC Input: (Only for China)
- 240Vdc/ 16A

DC Output:
- Max 1320W/ 100-110V~
+54V/ 24.4A
+12Vsb/ 3A
- Max 1500W/ 111-120V~
+54V/ 27.8A
+12Vsb/ 3A
- Max 2810W/ 200-207V~
+54V/ 52A
+12Vsb/ 3A
- Max 3000W/ 208-240V~ or 240Vdc Input
+54V/ 55.6A
+12Vsb/ 3A

[1] The system power supply requires C19 power cord.
System ManagementASPEED® AST2600 Baseboard Management Controller
GIGABYTE Management Console web interface

Dashboard
HTML5 KVM
Sensor Monitor (Voltage, RPM, Temperature, CPU Status …etc.)
Sensor Reading History Data
FRU Information
SEL Log in Linear Storage / Circular Storage Policy
Hardware Inventory
Fan Profile
System Firewall
Power Consumption
Power Control
Advanced power capping
LDAP / AD / RADIUS Support
Backup & Restore Configuration
Remote BIOS/BMC/CPLD Update
Event Log Filter
User Management
Media Redirection Settings
PAM Order Settings
SSL Settings
SMTP Settings

Build Your Configuration

Processors · memory · storage · networking · power
Select options below — pricing updates in real time as you configure.
  • Real-time pricing
  • SEWP V · ITES-4H quoting
  • U.S. integration & burn-in

Expand each group below to choose options — required fields are marked. Selections update pricing in real time.

Barebone
Processor
Memory
M.2 DRIVE
U.2/U.3 NVMe Drive
(Max Quantity: 8)
GPU
NETWORK ADAPTER
(Max Quantity: 12)
Trusted Platform Module
Power Cables
Cables
OPERATING SYSTEM
NTS AI Stack See what's inside each NTS AI package
Software
NVIDIA Software
Warranty

Why teams choose NTS

Engineer-to-order platforms with U.S. integration, federal contract paths, and workload-aware BOM review — not catalog-only SKUs.

Enterprise GPU rack servers staged for engineer-to-order deployment

Built in the USA

Configuration, burn-in, and rack staging from Fremont, California with nationwide delivery.

Contract-native quoting

ITES-4H, SEWP V, and OMNIA Partners paths with CLIN-ready BOM structure.

Engineer-to-order

Choose CPU, memory, up to 8 gpus, storage, network, and redundant power per workload.

Rack-ready delivery

Serial capture for CMDB, integration validation, and specialist turn-up support.

Solutions & Deployment

Optimized reference architectures, workload deployment patterns, and technical specifications for the NTS Elite APEX 5U Dual EPYC 8-GPU Server

Platform at a Glance

5U server chassis staged for factory burn-in and rack-ready deployment

5U Chassis

Custom engineer-to-order 5u chassis featuring rigorous factory burn-in, thermal validation, and seamless rack-ready deployment.
Enterprise server PCIe expansion with NICs, HBAs, and accelerator cards

Enterprise Expansion

High-bandwidth PCIe Gen 4/5 expansion slots supporting high-speed NICs, SAS/SATA/NVMe HBAs, and dedicated hardware accelerators.
U.S. integration facility validating server systems for federal and SLED programs

U.S. Integration

Secure U.S.-based configuration, multi-point validation, and compliant logistics tailored for federal, defense, and SLED programs.
Procurement specialist preparing federal and SLED contract quotes

Contract Quoting

Streamlined procurement via ITES-4H, NASA SEWP V, and SLED cooperative contracts with pre-negotiated pricing.

Delivery & Integration Lifecycle

From initial BOM design to production-ready deployment, our dedicated engineering team manages the complete lifecycle of your NTS Elite APEX 5U Dual EPYC 8-GPU Server.
NTS engineers staging server racks with structured cabling before datacenter deployment
  • Build & Burn-In
    Precision factory assembly, custom firmware baselining, rigorous stress testing, and comprehensive asset documentation prior to shipment.
  • Rack Integration
    Structured cabling, optimized power distribution, and comprehensive in-rack validation aligned with your specific data center standards.
  • Image & Tag
    Custom OS/hypervisor imaging, host-name provisioning, and automated serial/asset tag capture for seamless CMDB and ITAM integration.
  • Deliver & Support
    Secure, coordinated logistics combined with post-delivery engineering support to ensure rapid, hassle-free site turn-up.
  • GPU Readiness
    Cluster networking checks and workload smoke tests before handoff.
GPU server architecture with accelerators, host compute, and high-speed I/O

System Architecture

NTS Elite APEX 5U Dual EPYC 8-GPU Server is engineer-to-order for GPU Systems workloads — built for sustained accelerator duty cycles, in-rack serviceability, and network fabric headroom as your cluster scales.
  • Right-size CPU, memory, and I/O in an engineer-to-order BOM — validated and burn-tested in Fremont before ship
  • Add NICs, HBAs, and accelerators through PCIe expansion without chassis constraints
  • Deploy AMD Instinct / APU layouts with validated power, cooling, and service paths
  • Scale east-west and cluster traffic on 10/25/100GbE-class networking without platform upgrades
  • Match storage latency to workload — NVMe, SAS, or SATA tiers in the same chassis
  • Keep mission workloads online with redundant power and cooling for 24×7 operation
  • Image, monitor, and troubleshoot remotely from day one via IPMI / BMC
Federal procurement specialist reviewing contract compliance documentation alongside secure server integration

Procurement & Compliance

Procure NTS Elite APEX 5U Dual EPYC 8-GPU Server through SEWP V, ITES-4H, and SLED cooperative contracts. Benefit from CLIN-ready BOM structures, full compliance documentation, and secure delivery coordinated by our Fremont integration specialists.
  • Acquire federal, DoD, and public-sector quotes precisely mapped to your contract CLINs, funding cycles, and periods of performance
  • Access complete Trade Agreements Act (TAA) compliance and country-of-origin documentation for seamless contracting officer approval
  • Streamline agency acquisition reviews with detailed Energy Star ratings and power-efficiency metrics
  • Maintain strict property accountability with pre-captured serial numbers, MAC addresses, and custom asset tags for audit-ready tracking
  • Ensure compliance with FIPS 140-3 and NIST SP 800-53 security standards through pre-award engineering consultations
  • Coordinate secure, scheduled delivery windows directly with your ACO or COR, arriving fully integrated and rack-ready
SEWP V ITES-4H SLED Cooperatives CLIN-ready BOM Fremont, CA
Contract vehicles for this SKU are listed under Available through next to the product image. Explore SEWP V, ITES-4H, GSA, and SLED procurement paths →

Ideal Use Cases

Enterprise deployment scenarios where NTS Elite APEX 5U Dual EPYC 8-GPU Server delivers optimal price-performance, high reliability, and simplified serviceability.
GPU cluster training large language models at scale in an NTS data center

LLM Training

Fine-tune and train large language models at scale.
Multi-GPU HGX server tray for deep learning training on NTS hardware

Deep Learning

Multi-GPU training for CV, NLP, and GenAI workloads.
High-density HPC compute cluster running scientific simulations on NTS systems

HPC Simulations

Scientific computing and high-density parallel jobs.
NTS GPU server delivering real-time AI inference and batch processing

AI Inference

Real-time inference and batch data processing.
Computer vision lab training and deploying models on NTS GPU servers

Computer Vision

Training and deployment for vision pipelines.
NTS GPU rack powering generative AI, diffusion, LLM serving and RAG infrastructure

Generative AI

Diffusion, LLM serving, and RAG infrastructure.

Common workload profiles

Typical deployment patterns aligned to NTS Elite APEX 5U Dual EPYC 8-GPU Server capacity, I/O profile, and service tiers.
Multi-node HGX GPU cluster with high-bandwidth fabric and NTS branding for distributed training

Distributed training

Multi-node gradient sync across high-bandwidth GPU fabric.
Open GPU tray and dataset staging with NTS plaque for foundation model fine-tuning

Model fine-tuning

Adapt foundation models on domain-specific datasets.
Dense GPU inference rack with NVMe scratch and NTS branding for batch scoring

Batch inference

High-throughput scoring with optimized batch sizing.
GPU server beside flash array for vector search and embedding retrieval on NTS hardware

Vector search

Embedding indexes and similarity retrieval at scale.
Hybrid HPC and AI GPU chassis with NTS plaque for coupled simulation pipelines

Simulation + AI

Coupled HPC solvers with GPU-accelerated pipelines.
GPU render farm nodes with fast scratch I/O and NTS branding for frame batching

Render farms

Frame batching with fast scratch and asset I/O.

Frequently Asked Questions

Common technical and procurement topics for NTS Elite APEX 5U Dual EPYC 8-GPU Server. Additional questions are handled by our specialists and knowledge base.

Need architecture guidance or a formal quote?

Service & Trust

Coverage, procurement, and domain expertise for AI Solutions platforms — backed by U.S. integration and contract-native quoting.
  • NTS Fremont integration floor — rack staging and burn-in with NTS branding on the rack
    Built & Supported in the USA Factory-aligned configuration, validation, burn-in, and rack integration from Fremont, CA — with nationwide logistics for federal and commercial programs.
  • Nationwide U.S. logistics and data center delivery support with NTS-branded equipment crates
    Nationwide Coverage Delivery and post-sales engineering across all U.S. regions — defense, civilian agency, higher education, and commercial data center.
  • Contract-native quoting workstation with CLIN-aligned BOMs beside an NTS-branded secure server
    Contract-Native Quoting CLIN-aligned BOMs for ITES-4H, SEWP V, and SLED cooperative vehicles — view all contract paths.
  • NTS solution architects consulting on an AI and HPC rack with NTS branding on the chassis
    AI Solutions Expertise Dedicated solution architects for AI, HPC, virtualization, storage, and edge — from first quote through rack turn-up.