Rocket 7638D PCIe Gen5 x16 Switch Adapter
Powering Next-Generation AI, HPC, Scientific Research, and Media Production
The Rocket 7638D PCIe Gen5 Switch Adapter is a game-changing connectivity solution designed to eliminate bandwidth bottlenecks in high-demand computing environments. A dedicated 48-lane Gen5 PCIe switch chipset enables each adapter to intelligently allocate bandwidth to ensure hosted GPUs and NVMe storage operates at their maximum performance potential.
The unique hybrid 1x external CDFP-CopprLink port and 2x internal MCIO 8i ports configuration enables the Rocket 7638D to establish a direct, high-speed data pathway between hosted GPUs and NVMe storage arrays. The result is ultra-fast rendering, seamless real-time data processing, and accelerated workflows—making it the ideal choice for mission-critical applications in AI/ML, HPC, scientific research, and professional media production.
Key Features:
Proven PCIe Gen5 x16 Switch Architecture
Powered by Broadcom’s PEX 89048 48-lane PCIe Gen5 Switch Chipset
32 Lanes of Dedicated Downstream Bandwidth: Allocates x16 lanes exclusively for external GPU + x16 lanes for NVMe storage
Eliminates compute and storage contention for maximum throughput
Enterprise-Grade Connectivity
1x External CDFP-CopprLink Port for GPU expansion: Complies with PCI-SIG CopprLink standard, ensuring signal integrity at Gen5 speeds
2x Internal MCIO 8i Ports for scalable NVMe storage arrays: Up to 16 Drives
Performance Without Compromise
GPU and NVMe operate on independent pathways
Up to 64GB/s & 12 Million IOPS
Ultra-low latency, high-bandwidth throughput for AI training, 8K media editing, scientific simulations, and HPC modeling
Cross-Platform Compatibility
Universal support for x86 (Intel/AMD) & ARM-based platforms
Driverless deployment using native NVMe drivers across all major OS platforms
Smart Diagnostics & Custom Firmware
Integrated firmware-level monitoring suite
Real-time health, connectivity, and bandwidth utilization tracking
Flexible Firmware Customization services to meet unique workflow & application requirements
Industry Leading PCIe Switching Architecture
The Rocket 7638D is powered by industry proven PCIe Gen5 Switching Architecture, which delivers a full x16 lanes of PCIe bus bandwidth, and is capable of transmitting data at a staggering 60 GB/s, in either direction, with minimal latency! This level of sustained transfer speed is essential for data intensive applications including AI, ML, Scientific modeling and Engineering, and 3D design and animation workflows, all of which increasingly rely on high end GPUs and today’s fastest NVMe storage to maintain optimal performance.
Convert an MCIO 8i port into a Full-Sized PCIe 5.0 x16 Slot
The MCIO-PCIEX16-G5 Expansion bridge card can transform industry standard MCIO ports into PCIe Gen5 x16 slots. Each card can accept up to two MCIO connections and provide up to x16 lanes to the hosted PCIe device. This capability significantly boosts system connectivity, effectively overcoming the PCIe limitations of conventional motherboard design.
Rocket 7638D
Why Choose the Rocket 7638D?
HighPoint’s Rocket 7638D sets a new standard for high-speed interconnect solutions. Unlike traditional adapters and expansion boards, which force GPUs and PCIe storage devices to share bandwidth, the Rocket 7638D’s innovative switching architecture ensures both operate at peak performance—boosting rendering speeds, accelerating AI/ML pipelines, and optimizing real-time data analysis.
The unique CDFP-CopprLink connectivity and PCI-SIG certified cabling solution guarantees that maximum Gen5 bandwidth is always on tap, optimizing signal integrity and eliminating bottlenecks for mission critical applications. For professionals in data centers, research labs, AI clusters, and media studios, the Rocket 7638D provides the scalability, reliability, and performance to address today’s most challenging workflows.
Applications and Use Cases
HighPoint’s groundbreaking Rocket 7638D external PCIe Switch adapter is engineered to address today’s most demanding AI/ML and HPC workflows. Powered by a Broadcom 48-lane PCIe Gen5 switch and enterprise-grade CopprLink CDFP cabling, the Rocket 7638D delivers up to 128GB/s of bi-directional bandwidth, providing a direct link between the hosted GPUs and NVMe storage arrays.
Artificial Intelligence & Machine Learning (AI/ML)
Faster training and inference with seamless GPU-to-NVMe integration
HPC
(High- Performance Computing)
Optimized throughput for complex simulations, climate modeling, and analytics
Scientific
Research
Accelerated workflows for imaging, data acquisition, and real-time processing
Media Editing &
Content Creation
Smooth 8K video editing, VFX rendering, and data-intensive production pipelines
.png){kind=logo}
