Cisco HCIX-GPU-A16 | NVIDIA A16 GPU, Quad-GPU, 64GB GDDR6, 250W Passive

Q: What is the primary use case for the HCIX-GPU-A16?
A: The HCIX-GPU-A16 is optimized for high-density virtual desktop infrastructure deployments. Its quad-GPU design enables IT teams to deliver GPU-accelerated virtual desktops and workstations to knowledge workers running Office 365, web browsers, video conferencing, and light CAD applications. A single card can support up to 64 concurrent users depending on the vGPU profile selected.

Q: Does this GPU require NVIDIA vGPU software licensing?
A: Yes, the HCIX-GPU-A16 requires NVIDIA vGPU software to partition and allocate GPU resources to virtual machines. Organizations must license either NVIDIA Virtual PC for virtual desktops or NVIDIA RTX Virtual Workstation software for graphics-intensive workloads. Licensing is per-GPU or per-concurrent-user depending on the edition selected.

Q: Which Cisco platforms support the HCIX-GPU-A16?
A: The HCIX-GPU-A16 is qualified for Cisco UCS C-Series rack servers and Cisco HyperFlex HX-Series hyperconverged nodes. It requires a PCIe 4.0 x16 slot, adequate power supply capacity (minimum 250W available), and proper PCIe riser configuration. The GPU must be ordered from Cisco as it contains a unique SBIOS identifier required by CIMC and UCS Manager.

Q: Can multiple HCIX-GPU-A16 cards be installed in the same server?
A: Yes, multiple A16 cards can be installed in a single server if sufficient PCIe slots, power capacity, and cooling are available. However, GPU mixing is not supported—all installed GPUs must be the same model. Two A16 cards in a 2U server can support over 100 concurrent virtual desktop users.

Q: What is the difference between the A16 and other NVIDIA data center GPUs?
A: The A16 is specifically engineered for maximum VDI user density with its quad-GPU architecture, delivering four independent 16GB GPUs on a single card. Unlike compute-focused GPUs such as the A100 or graphics workstation GPUs like the RTX A6000, the A16 prioritizes concurrent user sessions over raw per-GPU performance, making it ideal for virtual PC workloads rather than AI training or 3D rendering.