Dolphin PCIe Chassis / backplanes

PCIe Expansion Chassis & Backplanes

Add high-power, full-size or specialized PCIe cards without replacing the host platform or forcing every device into the computer enclosure.

Chassis formats
2U · 4U
PCIe support
Gen1 through Gen5
Device capacity
Up to 9 cards
Power
Up to 4800 W on eBox 4 Pro
eBox 2 eBox 4 eBox 4 Pro Passive backplanes Active backplane Copper / optical uplinks
Technical context: Expansion chassis must be qualified with the selected host adapter, endpoint power, cooling, slot width, BAR demand, reset and hot-plug requirements. Mechanical capacity does not guarantee endpoint compatibility. The eBox 4 Pro and MXB585 are published as preliminary products; confirm availability, final configuration and regulatory status before design-in.

System Capabilities

Expansion chassis provide additional slots, power delivery, cooling and service access for PCIe cards that do not fit comfortably inside the host.

Additional card capacity

Add GPUs, FPGAs, frame grabbers, NICs, NVMe or instrumentation when host slots are insufficient.

Dedicated power and cooling

Place high-power devices in an enclosure designed around card spacing, auxiliary power and airflow.

Independent service access

Access add-in cards and cooling components without opening or replacing the host computer.

Configurable expansion scale

Choose passive or switched backplanes and a chassis size matched to card count and bandwidth requirements.

Technical Features

Key platform elements and integration considerations are outlined below.

eBox 2

A 2U, 19-inch expansion chassis for PCIe 1.0–5.0 devices with an 850 W ATX supply, front-to-rear cooling and mounting for one or two passive backplanes. Supported optical configurations can place compatible I/O up to 100 metres from the host.

eBox 4

A 4U, 19-inch ATX-based platform with up to eight add-in cards, a standard 1000 W supply and front-to-rear cooling. One or two passive 2-slot, 3-slot or 5-slot backplanes can be configured; model-specific optical links extend to as much as 200 metres.

eBox 4 Pro

A switched PCIe 5.0 platform built around one or two MXB585 active backplanes. Published configurations support up to nine installed PCIe cards in total, including one or two uplink/target cards; this provides up to eight downstream x16 add-in-device positions depending on topology. Balanced dual 2400 W supplies provide 4800 W maximum output or redundancy up to 2400 W.

Passive backplanes

PCIe 5.0 passive backplanes provide one x16 endpoint slot, two x8 endpoint slots or four x4 endpoint slots behind a dedicated x16 uplink slot. Each backplane requires a compatible transparent target/uplink adapter.

Active switched backplanes

The MXB585 active PCIe 5.0 backplane supports flexible x16 slot allocation, multiple uplink arrangements and transparent or NTB operation. Firmware, chassis configuration and software determine the available multi-host functions.

Power and thermal engineering

Full-length GPUs and FPGAs can exceed slot power and cooling assumptions. Auxiliary power connectors, airflow direction, cable obstruction and sustained workload must be checked together.

Specifications

Technical parameters and system-integration requirements are listed below. Chassis, power and uplink availability must be confirmed for the final ordered configuration. The eBox 4 Pro and MXB585 remain preliminary in the current published documentation.

PCIe Expansion Chassis & Backplanes architecture and integration parameters
Parameter Unit Value / Description
CHASSIS OPTIONS
eBox 2 2U 19-inch chassis, up to eight add-in cards, 850 W ATX supply and one or two passive backplanes; supported optical links up to 100 m.
eBox 4 4U 19-inch chassis, up to eight add-in cards, standard 1000 W ATX supply and one or two passive backplanes; supported optical links up to 200 m.
eBox 4 Pro 4U active PCIe 5.0 platform based on one or two MXB585 backplanes; up to nine installed cards including uplink/target cards, with up to eight downstream x16 device positions depending on configuration. Dual 2400 W supplies provide 4800 W maximum or redundancy up to 2400 W.
Backplane choices Passive two-, three- and five-slot layouts, plus active switched Gen5 platforms.
CARD & LINK FIT
Card mechanics Full-height or low-profile, card length, single/double width and rear I/O clearance.
Host link Transparent or NTB adapter, PCIe generation, x4/x8/x16 width and copper or optical cable.
Power W Slot power plus auxiliary 6/8/12VHPWR requirements and startup transients.
Cooling Front-to-rear airflow, device fan orientation, pressure drop and thermal headroom.
OPERATIONS
Enumeration BIOS MMIO resources and endpoint reset behaviour affect large multi-card configurations.
Hot add Not inherent to a chassis. Supported SmartIO Hot-Add workflows require a compatible NTB host adapter, eXpressWare release, operating system and endpoint driver.
Peer-to-peer Requires compatible endpoints, root complex, switch configuration, BIOS and IOMMU settings.
Management Active platforms provide Ethernet monitoring and management; passive systems rely on adapter diagnostics.

Applications

The architecture is most useful where native PCIe access, sustained data movement or tightly coupled multi-node communication materially affects the system.

GPU and AI expansion

Install full-length accelerators in a dedicated powered chassis while the host remains a compact server, workstation or embedded controller.

FPGA and frame-grabber systems

Increase accelerator and acquisition-card capacity without replacing the controlling computer or redesigning its motherboard.

High-speed data acquisition

Place digitizers and protocol interfaces near sensor or test racks while centralising processing and storage.

NVMe and storage expansion

Build direct PCIe paths to external NVMe devices or storage backplanes, subject to host resource and peer-to-peer compatibility.

Medical imaging equipment

Separate high-power image acquisition and reconstruction hardware from operator systems while engineering medical safety independently.

Production and laboratory test

Improve card access, replacement, cooling and cable routing in systems that change frequently or exceed internal host capacity.