Summary

Half-rack (10-inch width) embed compute systems are purpose-built for applications where physical footprint is the primary constraint: vehicle bays, aircraft compartments, naval platforms, and edge datacenters with fixed rack-unit budgets. The form factor trades off absolute compute density against space efficiency, enabling 2–4 servers to fit in the space of a single full-width (19-inch) system. Three distinct ecosystems serve this market: commercial industrial-grade systems (Neousys, ASUS, ADLINK), military-grade MicroTCA modular platforms (nVent SCHROFF, Pixus, Recab), and purpose-built vertical solutions (Galleon NAS, 7Starlake NAS, Anduril/Klas Voyager).

Key Facts

  • Form Factor: 10-inch width (half of standard 19-inch rackmount); typical heights 1U–3U
  • Thermal Design: Fanless (passive) or ultra-low-airflow; power budgets 50–600W depending on variant
  • Standards: No unified standard (unlike VPX/OpenVPX); loosely follows industrial rackmount conventions (EIA RS-310)
  • Primary Use Cases: Vehicle/aircraft/naval compartment deployment, space-constrained edge POPs, modular multi-node clusters
  • Ecosystem Fragmentation: High — no cross-vendor form-factor compatibility (unlike MicroTCA/VPX backplanes)

What It Is / How It Works

Why Half-Rack?

Standard full-width (19-inch) servers occupy ~152mm width. A 10-inch half-rack is ~254mm width—exactly half the footprint. This enables critical deployments where space is finite and non-negotiable:

  • Vehicle bay mount: A Bradley Fighting Vehicle, Black Hawk helicopter, or patrol boat has a fixed equipment rack (6U–9U height), usually one meter wide. Two half-rack servers fit where one full-width server would. This doubles operational flexibility: a vehicle can carry dual-redundant compute, or a compute + networking unit.
  • Ship compartment: Naval surface combatants (destroyers, frigates) pack electronics into compact combat information centers. Half-rack width reduces floor space, enabling more sensor processing pipelines in the same compartment.
  • Aircraft avionics bay: Narrow fuselage sections (business jets, transport aircraft) require thin, tall server forms. Half-rack width maximizes compute density in constrained vertical spaces.
  • Edge datacenter: A modular edge POP (Point of Presence) designed for rapid deployment uses ruggedized shipping containers or prefab structures. Half-rack systems stack two-per-width, increasing node count per footprint, reducing site size and power-distribution weight.

Platform Tiers

Tier 1: Commercial Semi-Industrial (Neousys SEMIL, ASUS RUC-1000D, ADLINK AXE, Premio JCO)

  • 2U–3U height; fanless or low-airflow cooling; DDR4/DDR5 DRAM
  • Processors: Intel Core Ultra, Xeon E, or Xeon D
  • Temperature range: -20°C to +60°C (some extended to -40°C/+70°C)
  • Network: GigE or dual GigE; occasional 10 GbE option
  • Storage: M.2 NVMe or 2.5" SSD
  • Target: ITS, transportation, base security, non-classified defense-adjacent
  • Price: $5K–$20K per unit

Tier 2: Military-Grade Modular (MicroTCA systems from nVent SCHROFF, Pixus, Recab)

  • 1U–3U height; modular backplane architecture; Advanced Mezzanine Cards (AMCs)
  • Processors: Module-dependent (FPGA, custom AI accelerators, ARM/x86 compute)
  • Temperature range: -40°C to +85°C typical
  • Network: Custom per-module (GigE to 100 GbE backplane)
  • Certifications: VITA 62 (modular power), MIL-STD-810, MIL-STD-461 (EMI)
  • Target: DoD tactical edge, signals processing, classified AI inference
  • Price: $50K–$150K+ (high complexity, low volume)

Tier 3: Vertical Domain Solutions (Galleon XSR NAS, 7Starlake THOR200 NAS, Anduril Menace Compute)

  • 2.5U–3U height; purpose-built for specific mission (NAS, tactical C4, autonomous vehicle compute)
  • Processors: Xeon E/D (NAS), or custom integration (Menace)
  • Storage: 40TB–80TB (NAS tier) or modular GPU expansion (tactical tier)
  • Certifications: MIL-STD-810/901D, FIPS-140-2 (NAS); MIL-STD-461, MilSpec power (tactical)
  • Target: Naval platforms, aerospace, special operations, autonomous systems
  • Price: $25K–$100K+ (domain-specific, moderate volume)

Competitive Landscape

vs. MicroTCA (VITA 62)

MicroTCA modular standards enable hot-swap compute modules, true multi-vendor interoperability, and field customization. However, MicroTCA systems are significantly more expensive (~$100K+ for a 3-slot µTCA chassis with compute), and their modular overhead (standardized backplane, inter-module connectors, MCH controller) means lower density than purpose-built systems. MicroTCA is preferred in classified DoD programs where modularity and long-term supportability outweigh cost; commercial/non-classified programs prefer purpose-built half-rack systems.

vs. Full-Width (19-Inch) Rackmount

Full-width servers (2U, 4U standard) achieve better thermal dissipation (larger chassis, more airflow), more expansion slots, and lower cost-per-compute (economy of scale). Half-rack trades off compute density and cost to gain spatial footprint reduction — only justified when space is the binding constraint. For fixed datacenters with unlimited floor space, full-width is always more cost-effective.

vs. Blade/Sled Systems (e.g., NVIDIA HGX)

Modular blade systems (Supermicro SuperBlade, OAM GPU sleds) pack even higher density by eliminating per-node redundancy (shared power, cooling, fabric). Half-rack systems maintain per-node power supplies and thermal isolation, incurring a density penalty. Blades are preferred for homogeneous, high-scale deployments (hyperscaler datacenters); half-rack systems are preferred for heterogeneous, small-cluster, or tactical deployments where modularity and spares commonality matter.

Key Technical Trade-Offs

Thermal Design:

  • Fanless half-rack systems achieve -40°C operation (Neousys, Galleon, 7Starlake) by relying on conduction cooling and passive heat sinks, but limited to 50–150W thermal dissipation.
  • GPUs requiring 300W+ thermal budget demand active cooling (fans) or liquid cooling, reducing fanless feasibility.
  • Result: Half-rack fanless systems support edge inference (NVIDIA Jetson Orin, modest RTX), not heavy compute (H100, H200).

Power Density:

  • Full-width 2U: up to 5–10 kW per unit (shared PSU, shared cooling infrastructure).
  • Half-rack 2U: up to 2–3 kW per unit (smaller PSU, more thermal challenges).
  • Result: Two half-rack units ≠ one full-width unit in raw power capacity. Offset by spatial savings.

Modularity:

  • Half-rack commercial systems (Neousys, ASUS) are largely fixed-configuration (“buy the spec you need”).
  • MicroTCA allows in-field module swaps and future upgrades.
  • Result: Commercial systems are cheaper and simpler; MicroTCA is flexible but expensive.

Ecosystem Players

System Builders

Company Form Factor Ecosystem Market Tier
Neousys 2U half-rack, fanless Industrial/commercial Tier 1 (semi-industrial)
ASUS RUC-1000D Half-rack, fanless Commercial Tier 1 (semi-industrial)
ADLINK AXE 2U–3U half-rack Industrial/defense-adjacent Tier 1–2
Premio 2U–3U half-rack Commercial edge Tier 1
Galleon XSR 3U half-rack Military NAS Tier 3 (vertical)
7Starlake THOR200 2.5U half-rack Military aerospace/submarine Tier 3 (vertical)
nVent SCHROFF 1U–3U µTCA half-rack Military modular Tier 2 (MicroTCA)
Pixus Technologies µTCA chassis Military modular backplane Tier 2 (MicroTCA enabler)

Processor Supply

  • Intel Xeon E / Core / Core Ultra: Dominant in Tier 1 systems
  • Intel Xeon D: Preferred for Tier 3 (extreme-environment) due to low thermal envelope
  • NVIDIA Jetson Orin: Man-portable variant (8–16W); Tier 1 systems
  • NVIDIA RTX Discrete: Tier 1–2 systems; 150–300W typical
  • Custom FPGA / AI Accelerators: Tier 2 (MicroTCA modules)

Power Delivery and DC Distribution

Power budget is a hard constraint in half-rack form factors. The smaller chassis volume limits PSU size, thermal dissipation, and cabling.

Typical power envelopes by tier:

Tier Power Budget Notes
Tier 1 (commercial fanless, Neousys/Premio) 50–150 W Intel Core or Xeon E; Jetson Orin class GPU
Tier 1 (semi-industrial GPU, ASUS RUC-1000G) 300–600 W Discrete NVIDIA RTX 40-series; requires fan cooling
Tier 2 (MicroTCA, nVent SCHROFF) 50–500 W Module-dependent; VITA 62 power supply standard
Tier 3 (vertical, Galleon XSR, 7Starlake THOR200) 100–300 W Xeon D with NVMe storage; fanless possible

Vehicle bus power standards:

Half-rack systems deployed in ground vehicles must accept 24–28V DC power from the vehicle bus (MIL-STD-1275 for military ground vehicles). Aircraft platforms typically supply 28V DC (MILSPEC aircraft) or 115V AC (transport aircraft with conversion). Shipboard platforms may supply 115/220V AC (MIL-STD-1399). Power supply selection is a first-order integration challenge for deployed systems, and some half-rack vendors (PowergridM, Behlman) supply purpose-built 1U half-rack UPS/conditioners designed to sit in the same bay as the compute unit.

48V DC distribution:

Higher-performance half-rack systems increasingly use 48V internal distribution to reduce cable losses at high current. The VITA 62 power supply standard supports 48V outputs. PowergridM’s stackable UPS line includes 48V-capable configurations for tactical power conditioning; output voltages are configurable from 2V to 48V. 48V internal buses allow thinner cables and lower I²R losses, important in chassis where cable routing is tightly constrained.

Power redundancy:

Most Tier 1 commercial half-rack systems do not offer redundant PSUs — the chassis is too small. Redundancy is achieved at the system level (two units, failover). MicroTCA Tier 2 systems support redundant power modules (two VITA 62 PSUs per shelf) per the standard.

Ruggedness and Environmental Ratings

MIL-STD-810H compliance: The current revision of the DoD environmental test standard. Vendors claiming “MIL-STD-810” should be qualified: which revision (H is current), and which test methods?

Key MIL-STD-810H methods relevant to half-rack edge compute:

  • Method 501/502: High/low temperature storage and operation. Relevant range for vehicle deployment: operating at -32°C to +63°C (Category C3).
  • Method 514: Vibration. Different profiles for ground transport, wheeled vehicle, tracked vehicle, and aircraft. Tracked vehicle spectra are far more severe than wheeled.
  • Method 516: Shock. Drop testing (1m operational shock) and bench handling. Naval shock (Method 522 / MIL-STD-901D) is a separate, far more demanding standard.
  • Method 507: Humidity (damp heat, 40°C / 85% RH sustained)
  • Method 510: Sand and dust ingestion — test for IP-rated sealed enclosures

IP ratings (IEC 60529):

Rating Protection Application
IP54 Dust-limited / splash-resistant Light outdoor use; Neousys SEMIL standard configuration
IP65 Dust-tight / water jet resistant Vehicle exterior, maritime deck exposure
IP67 Dust-tight / 30-minute immersion (1m) Amphibious, exposed deployment; Neousys SEMIL-1300 and Galleon
IP69K Steam jet cleaning Extreme wash-down industrial

Operating temperature ranges:

Configuration Range Notes
Standard commercial server 0°C to +40°C Controlled datacenter
Semi-industrial (Neousys, Premio, ADLINK) -20°C to +60°C Typical fanless AI edge server
Extended rugged (Neousys SEMIL-1300, Galleon XSR) -40°C to +70°C Vehicle bay, exposed rack
MIL-grade (Crystal Group, VPX Tier 2) -40°C to +71°C or +85°C Full MIL-STD-810H qualification

Cost premium of ruggedization:

Ruggedization adds material cost (aluminum alloy vs. steel chassis, conformal coating on PCBs, MIL-spec connectors), qualification cost (running MIL-STD-810 test campaigns costs $50K–$250K per product), and production overhead (lower volume, US sourcing requirements for DoD programs). As a rough multiplier: a commercial Tier 1 half-rack unit priced at $5K–$20K may have a ruggedized/qualified equivalent at $25K–$100K for the same processing capability. Full VPX Tier 2 systems with MIL-STD-810H/901D qualification can run $100K–$300K.

Sources