.png){kind=logo}
Eliminate PCIe Slot Starvation in Dense AI & HPC Infrastructure
Unlock flexible PCIe Gen 5 routing, multi-GPU compute, and high-density expansion across 1U/2U servers and edge AI platforms with HighPoint’s modular MCIO ecosystem.
The Challenge: The Physical Limits of Modern AI Infrastructure
Modern AI inference, deep learning clustering, and edge computing demand immense data throughput and multi-accelerator density. However, system architects face hard mechanical and electrical barriers:
Physical Slot Starvation: 1U and 2U server chassis lack sufficient physical PCIe slots to host multiple GPUs, DPUs, and high-speed NICs.
Thermal & Airflow Chokepoints: Rigid PCIe riser cards restrict airflow and force high-power accelerators into dense, hot zones.
BIOS Bifurcation Limitations: Host server platforms often lack granular BIOS bifurcation to split lanes across auxiliary add-in cards.
HighPoint’s 3-Tier PCIe Gen 5 MCIO Expansion Architecture
HighPoint leverages high-density SFF-TA-1016 (MCIO 8i) connectivity and 32 GT/s Gen 5 signaling to route up to 64 GB/s of bandwidth anywhere in the chassis.
.png)
Tier 1: Direct Onboard Slot Expansion
Core Hardware: Onboard MCIO 8i Ports + MCIO-PCIEX16-G5 Passive Bridge Card
How It Works: Converts native MCIO 8i ports on AMD EPYC™ and Intel® Xeon® motherboards into a dedicated PCIe 5.0 x16 expansion slot.
Key Advantage: Zero host slot footprint. Add full-height GPUs, DPUs, or 400GbE NICs without consuming an existing motherboard PCIe slot.
The Hardware Solutions
-
2x MCIO 8i (SFF-1016) Ports (dedicated x8 lanes per port)
-
1x PCIe x16 Slot
-
Provides up to x16 lanes of dedicated downstream bandwidth
Supports any Industry Standard PCIe Device
.png)
Tier 2: Retimer-Based Signal Extension
Core Hardware: Rocket 1624L (Internal) / Rocket 1624M (External) + MCIO-PCIEX16-G5 Bridge Card
How It Works: Uses physical-layer retimers to re-time and equalize PCIe 5.0 signals (32 GT/s) across internal/external cable runs (requires host bifurcation).
Key Advantage: Low-cost signal conditioning. Relocate accelerators to cooler chassis zones or external enclosures with ultra-low latency.
Dual-Slot Expansion (x8)
Versatile Storage & Slot Expansion
-
Up to 4x NVMe Drives
-
2x PCIe Devices
Overcome Signal Degradation
-
Retimer architecture intercepts, re-clocks, & amplifies PCIe Signals
-
Relocate PCIe cards anywhere within the chassis!
The Hardware Solutions
2x Internal MCIO 8i Ports
2x External MCIO 8i Ports
.png)
Tier 3: Independent Switch-Based Multi-Slot Fabric
Core Hardware: Rocket 1624L (Internal) / Rocket 1624M (External) + MCIO-PCIEX16-G5 Bridge Card
How It Works: Uses physical-layer retimers to re-time and equalize PCIe 5.0 signals (32 GT/s) across internal/external cable runs (requires host bifurcation).
Key Advantage: 100% Bifurcation-Free & Direct P2P. Expands a single slot into multiple PCIe x16/x8 slots or up to 32 NVMe drives, enabling Direct Peer-to-Peer GPU-to-storage data offloading.
PCIe Expansion (Rocket 1628A)
The Hardware Solutions
Peer-to-Peer Connectivity Solutions (Rocket 1628A)
Architecture Selection Guide
Which Tier is right for my Workflow?
Solution Tier | Core Products | Bifurcation Required? | Host Slot Consumed? | Max Bandwidth | Ideal Use Case |
|---|---|---|---|---|---|
Tier 1: Direct Onboard | Onboard MCIO + MCIO-PCIEX16-G5 | System Dependent | No (Uses board ports) | PCIe 5.0 x16 (64 GB/s) | Utilizing unpopulated motherboard MCIO ports for zero-footprint slot expansion. |
Tier 2: Retimer Extension | Rocket 1624L/M + MCIO-PCIEX16-G5 | Yes (Host must split lanes) | Yes (1x x16 Slot) | PCIe 5.0 x16 (64 GB/s) | Cost-effective internal cable routing and thermal relocation. |
Tier 3: Switched Fabric | Rocket 1628A/1624A + MCIO-PCIEX16-G5 | No (Hardware Switched) | Yes (1x x16 Slot) | PCIe 5.0 x16 (64 GB/s agg.) | High-density AI multi-slot compute, direct P2P DMA, and up to 32 NVMe drives. |