Nvidia NVLink Bridge 2-Slot for 3090 A100 H100 A30 A40 A800 A5000 A5500 A6000 900-53651-2500-000 P3651

This is the 2 slot version for when there is no empty slots between 2 slot cards. If you have one or more empty slots between the cards or the cards are 3 slot this NVLink will not work. See the attached images showing the card layout.

NVLink 3.0 For any brand of RTX Ampere models:

• 3090
• A30
• A40
• A100 (Requires three NVLinks)
• A800
• A4500
• A5000
• A5500
• A6000
• H100  (Requires three NVLinks)

Part number 900-53651-2500-000

Model: P3651

This is the same as PNY part number: NVLAMP-2SLOT-BSP and RTXA6000NVLINK-KIT

This is the same as PNY UPC: 7834635987

This is the same as Supermicro Part number: GPU-NVTNVLINK-A100

Up to 400 GB/s for NVIDIA A800 40GB Active with 2 NVLink bridges Up to 112 GB/s all other GPUs with single NVLink bridge. 

NVIDIA® NVLink™ is the world's first high-speed GPU interconnect offering a significantly faster alternative for multi-GPU systems than traditional PCIe-based solutions. Connecting two NVIDIA® graphics cards with NVLink enables scaling of memory and performance1 to meet the demands of your largest visual computing workloads.

NVLink Overview and How It Works

NVIDIA’s NVLink is a high-bandwidth, energy-efficient interconnect technology designed to facilitate faster data exchange between the CPU and GPU, as well as between multiple GPUs. 

High-Bandwidth Connection

NVLink offers significantly higher bandwidth than the older PCI Express (PCIe) connections. This allows for faster data transfers between the components that are connected with NVLink.

Multi-GPU Support

NVLink is particularly useful for setups with multiple GPUs (like in high-performance computing or data centers). It allows for direct GPU-to-GPU communication, which can reduce the need to send data back and forth through the system’s main memory.

Hardware Bridges

NVLink is used with a physical bridge that connects two GPUs together. This bridge facilitates the high-speed communication between the GPUs.

Unified Memory

In certain high-end systems, NVLink can support unified memory access, where the CPU and GPU can share a single memory address space. This can simplify programming and increase performance for certain tasks.

Scalability

NVLink is designed to be scalable, meaning it can be used in configurations ranging from workstations with just a couple of GPUs to large supercomputers with thousands of GPUs.

Compatibility

Not all NVIDIA GPUs support NVLink. It’s essential to ensure that both the GPUs and the motherboard (or system) support NVLink if you plan to use this technology.

PNY and other manufacturers design their NVLink-compatible GPUs to NVIDIA’s specifications, ensuring they can take full advantage of the NVLink features when paired with compatible hardware and software.

If you’re thinking about utilizing NVLink in your setup, make sure you have the appropriate hardware (GPUs, bridges, motherboard) and that your software applications can benefit from the enhanced GPU-to-GPU communication.