WOLF-263L
Block Diagram
The block diagram shows two NVIDIA RTX 5000 Blackwell GPUs connected to a central PCIe Gen4 switch, with each GPU providing a PCIe x16 path and up to 4 video outputs. The ConnectX-7 EN SmartNIC connects to the same switch and routes two 100G KR4 Ethernet interfaces to P1. The diagram also shows configurable PCIe options on P2 and P5, rear video routing through P3/P4/P6, and SOSA 10.6.4 configurations where P3 and P6 are depopulated.
Overview

Dual NVIDIA Blackwell GPUs with ConnectX-7 and 100GbE
The VPX6U-BW5000E-DUAL-VO-CX7 combines two NVIDIA RTX™ 5000 Blackwell embedded GPUs, an NVIDIA ConnectX-7 EN SmartNIC, and a configurable PCIe Gen4 switch in a rugged 6U VPX payload module. The design provides high-performance embedded computing (HPEC), artificial intelligence (AI), high-bandwidth networking, video output, and sensor-processing capability in a single VPX module.
Each NVIDIA RTX 5000 Blackwell GPU connects to the onboard PCIe Gen4 switch through a PCIe x16 interface and includes 24 GB of GDDR7 ECC memory. The ConnectX-7 EN connects to the PCIe switch through two PCIe x16 interfaces and provides two 100G KR4 Ethernet interfaces routed through VPX connector P1.
This architecture provides high-bandwidth data paths between the GPUs, network interface, VPX PCIe fabric, and onboard PCIe devices, with support for NVIDIA GPUDirect RDMA and RoCE.
Features
WOLF-263L is a rugged 6U VPX module that integrates two NVIDIA RTX 5000 Blackwell embedded GPUs with a ConnectX-7 EN SmartNIC and configurable PCIe Gen4 switch. Each GPU includes 24 GB GDDR7 ECC memory and supports high-throughput AI/HPEC processing, while ConnectX-7 provides dual 100GbE KR4 networking with GPUDirect RDMA and RoCE. The module supports up to 8 DisplayPort outputs with HDMI options and a configurable 210W to 350W power envelope.
GPU Features
Dual NVIDIA RTX 5000 Blackwell GPUs:
10496 CUDA Cores per GPU
320 Tensor Cores per GPU (5th Gen)
80 Ray Tracing Cores per GPU (4th Gen)
24 GB GDDR7 256-bit ECC VRAM per GPU
Blackwell GPU parallel processing:
CUDA Toolkit 13, Compute Capability 12.0
CUDA-X AI and CUDA-X HPEC libraries
OpenCL 3.0, DirectX 12 Ultimate, OpenGL 4.6, OpenGL ES 3.2, Vulkan 1.4
NVENC (9th Gen) and NVDEC (6th Gen) with H.264, H.265, and AV1 support
Connectivity / System Management
NVIDIA ConnectX-7 EN SmartNIC
Two 100G KR4 Ethernet interfaces routed through VPX P1
NVIDIA GPUDirect RDMA and RoCE support
Configurable PCIe Gen4 switch
Each GPU connects to the PCIe switch through PCIe x16
ConnectX-7 connects to the PCIe switch through two PCIe x16 interfaces
Configurable PCIe options on P2 and P5: x16 or x8 + x8
IPMI system management, Tier 3
Mechanical / Open Systems Architecture
Rugged conduction cooled (CC), air cooled (AC), or liquid flow through (LFT)
Operating temperature:
CC: -40°C to +70°C standard
AC: -40°C to +60°C standard
Vibration (sine): 10G peak (5–2000 Hz)
Shock: 40G peak for conduction cooled, 30G peak for air cooled
Weight: AC approx. 1.5 kg, CC approx. 2.5 kg, LFT approx. 1.7 kg
ANSI/VITA 48 and VITA 65 (VPX-REDI, OpenVPX)
OpenVPX 10.2.7 Payload Slot Profile, user defined for video
OpenVPX and SOSA Aligned 10.6.4 / 10.6.3 Payload Slot Profiles, P3/P6 depopulated with no rear video
Specifications
Key configuration and performance parameters for WOLF-263L 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
This module is preliminary / EAU. Contact Sales for ordering numbers and configuration options.
Representative configuration options include:
Cooling configuration: conduction cooled, air cooled, or liquid flow through
Slot profile / backplane mapping
Rear video routing versus SOSA-aligned P3/P6 depopulated configurations
Display output options, including DisplayPort and HDMI
PCIe configuration options on P2 and P5
Default power threshold / maximum module power configuration
Conformal coating options
FAQ
Choose WOLF-263L when you need dual Blackwell RTX 5000 GPU performance plus integrated high-bandwidth networking on the same 6U VPX module.
This module is typically selected when:
- You need two RTX 5000 Blackwell GPUs in one 6U VPX payload slot
- You need ConnectX-7 networking with dual 100GbE KR4 interfaces
- Your application benefits from GPUDirect RDMA and RoCE for low-latency data movement
- You need up to 8 DisplayPort outputs with HDMI options
- You want a single module that combines HPEC, AI processing, video output, and high-speed networking
If your system does not need integrated ConnectX-7 networking, WOLF-2638 may be the simpler dual-GPU option.
ConnectX-7 adds high-bandwidth network connectivity directly into the dual-GPU processing architecture.
In practice, it helps when:
- Large sensor or network datasets need to move into GPU-accelerated pipelines
- The system needs two 100G KR4 Ethernet interfaces routed through VPX P1
- RoCE is used for efficient Ethernet fabric communication
- GPUDirect RDMA is used to reduce CPU overhead and avoid unnecessary memory copies
- The application depends on low-latency movement between the network, PCIe switch, and GPUs
This makes WOLF-263L a strong fit for distributed AI, sensor fusion, and high-throughput HPEC systems where data movement is as important as GPU compute.
Confirm these early to avoid backplane, cooling, and I/O routing issues:
- Cooling method: conduction cooled, air cooled, or liquid flow through
- Power envelope: confirm the chassis can support the selected 210W to 350W module power configuration
- Slot profile and backplane mapping: confirm OpenVPX 10.2.7, 10.6.4, or 10.6.3 compatibility as required
- SOSA-aligned configurations: confirm whether P3 and P6 are depopulated and whether rear video is available
- Ethernet routing: confirm dual 100G KR4 interfaces through VPX P1
- PCIe routing: confirm configurable PCIe Gen4 options on P2 and P5, including x16 or x8 + x8
- Video routing: confirm whether the system needs up to 8 DisplayPort outputs, HDMI options, or a no-rear-video SOSA configuration
- System management: confirm IPMI Tier 3 integration and monitoring expectations
For WOLF-263L, the highest-risk integration areas are cooling capacity, backplane profile alignment, 100GbE routing, and rear video availability.
The key ordering-time choices are mainly driven by cooling, slot profile, networking, and rear I/O requirements:
- Cooling configuration: conduction cooled, air cooled, or liquid flow through
- Slot profile and backplane mapping: determines whether the module is built for OpenVPX video routing or SOSA-aligned P3/P6 depopulated configurations
- Rear video routing: determines whether up to 8 DisplayPort outputs are available or whether the build removes rear video for SOSA alignment
- Display output configuration: DisplayPort and HDMI options should be locked early
- PCIe configuration: P2 and P5 options may be configured as x16 or x8 + x8
- Network configuration: confirms dual 100G KR4 routing through P1 and the intended ConnectX-7 fabric behavior
- Power threshold: choose a 210W to 350W configuration that matches chassis thermal capability
- Conformal coating: supports environmental robustness for deployed systems
This module is preliminary / EAU. Contact Sales for ordering numbers and configuration options.
