WOLF-3576
Overview
NVIDIA RTX 2000 Ada XMC with 3 Video Outputs
The XMC-AD2000E-VO module integrates an NVIDIA RTX™ 2000 (AD107) Ada embedded GPU on a single-width XMC mezzanine, providing a compact GPU accelerator for embedded visualization, video encode/decode, and AI inference workloads.
Host data is delivered to the GPU over a PCIe Gen4 x8 interface, processed on the Ada architecture CUDA/Tensor/RT cores, and then routed out through up to three simultaneous video outputs. The default output type is DisplayPort++ 1.4, with configuration options to support HDMI or DVI through component population changes.
The module includes 8 GB of GDDR6 (128-bit) graphics memory with up to 256 GB/s bandwidth, and supports a configurable operating power range intended for thermally constrained environments while still enabling higher performance points when platform power and cooling allow.
Block Diagram
The block diagram shows the Ada AD107 GPU connected to the XMC P15 connector via PCIe Gen4 x8, with three GPU display links routed through optional DP++/DVI circuitry to XMC P16 as VID1/VID2/VID3, alongside GDDR6 frame buffer and board power regulation.
Features
WOLF-3576 is a rugged XMC module that integrates an NVIDIA RTX 2000 Ada (AD107) embedded GPU to provide embedded graphics, encoding, and AI inference with up to 3 simultaneous video outputs. It supports DisplayPort++ (default), HDMI, and DVI output options and includes 8 GB GDDR6 128-bit memory. Host connectivity is provided over a PCIe Gen4 x8 interface.
Video Output
- Up to 3 simultaneous video outputs
- DisplayPort++ 1.4 (default)
- 4K @ 120 Hz / 5K @ 60 Hz, 10-bit color depth
- Options for HDMI or DVI via component population changes
Connectivity / Software
- PCIe Gen4 x8 host interface
- NVIDIA GPUDirect RDMA support
- Linux and Windows drivers available
- Host CPU support requirement: Intel E/S/H/H35 class, AMD H/HS class
Mechanical / Ruggedization
- Rugged conduction cooled (CC)
- Operating temperature: -40°C to +70°C standard (operational max: TBD)
- Vibration (sine): 10G peak (5–2000 Hz)
- Shock: 40G peak (conduction cooled)
- Dimensions (CC): 143.75 mm x 74 mm x 13.5 mm
- VITA 46.9 VPX I/O mapping patterns supported: X12d, X8d, X16s
Specifications
Key configuration and performance parameters for WOLF-3576 are summarized below for quick comparison. Values shown reflect the standard module configuration; contact WOLF if you need a tailored build or a specific payload profile.
Configuration Guide
Configuration options for the XMC-AD2000E-VO include:
- Display interface population (DP++ default; HDMI/DVI options)
- Conformal coatings
- XMC 1.0 / XMC 2.0 / XMC+ (ANSI/VITA 88) via component population changes
- Air or conduction cooled builds
- Maximum power configuration options
- Variant locking
Contact WOLF to determine the appropriate configuration for your system.
Related Products
- NVIDIA RTX A2000E (GA107)
- 2560 CUDA Cores, 80 Tensor Cores
- 8 GB GDDR6 VRAM
- PCIe Gen 4 x8
- N/A
- Module power 26–50 W
- NVIDIA RTX 500 Blackwell (GB203)
- 1792 CUDA Cores, 56 Tensor Cores
- 6 GB GDDR7 VRAM
- PCIe Gen 4 x8 (Gen 5 capable)
- N/A
- Module power 30–60 W
FAQ
Choose WOLF-3576 when you need an XMC GPU module with stronger Ada-generation compute capability than an RTX A2000-class module, while staying within a compact and thermally constrained mezzanine form factor.
This module is typically selected when:
- You need RTX 2000 Ada GPU performance for embedded graphics, encode/decode, and AI inference
- You need up to 3 simultaneous video outputs
- You want 8 GB GDDR6 memory with a 128-bit interface
- You need PCIe Gen4 x8 host connectivity
- You want a higher-performance option than WOLF-3476, but do not need a larger VPX GPU module
If your priority is the lowest power Blackwell XMC option with 6 GB memory, WOLF-3696 may be a better fit. If your platform is already qualified around Ampere A2000, WOLF-3476 may be simpler for sustainment.
Confirm the display plan early because XMC video output depends heavily on carrier routing and connectorization.
WOLF-3576 supports:
- Up to 3 simultaneous video outputs
- DisplayPort++ 1.4 by default
- 4K at 120 Hz or 5K at 60 Hz with 10-bit color depth
- HDMI or DVI options through component population changes
Integration checks to confirm early:
- Whether your carrier routes all 3 display links from XMC P16
- Whether the system needs DisplayPort++, HDMI, DVI, or a mix of outputs
- Whether the display outputs are routed to a front panel, rear I/O, or a carrier breakout
- Whether downstream displays require specific resolution, refresh rate, or color-depth support
Locking the display interface mix early helps avoid carrier rework and late-stage cabling changes.
Confirm these early for predictable bring-up:
- PCIe interface: PCIe Gen4 x8 to the XMC connector
- Carrier compatibility: confirm support for the required XMC connector option, such as XMC 1.0, XMC 2.0, or XMC+
- I/O mapping: confirm the carrier supports the required VITA 46.9 mapping pattern, such as X12d, X8d, or X16s
- Driver support: Linux and Windows drivers are available
- Host CPU support: confirm the platform meets the NVIDIA host CPU support requirement, such as Intel E/S/H/H35 class or AMD H/HS class
- GPUDirect RDMA intent: if you plan to use peer-to-peer or low-latency data movement, validate the full carrier, host, and endpoint topology
For XMC GPU modules, the biggest integration risks are carrier PCIe capability, display routing, and host platform compatibility.
The key ordering-time decisions are usually driven by carrier compatibility, cooling, and display routing:
- Display interface population: DisplayPort++ default, with HDMI and DVI options by component population changes
- XMC connector type: XMC 1.0, XMC 2.0, or XMC+ depending on carrier requirements
- Cooling build: conduction cooled or air cooled depending on chassis and mechanical design
- Maximum power configuration: choose a power setting in the 25W to 50W range that matches platform thermal headroom
- Conformal coating: supports environmental robustness for deployed systems
- Variant locking: supports configuration control for qualified or repeat-build programs
Contact WOLF to determine the appropriate configuration for your system.