Cisco HCI-GPU-L4 | L4 Tensor Core GPU, 24GB GDDR6, 72W, PCIe Gen4

Q: What Cisco platforms support the HCI-GPU-L4?
A: The HCI-GPU-L4 is designed for Cisco Compute Hyperconverged with Nutanix environments, including the HCIAF220C M7 and HCINX225 M8 series servers. It requires a PCIe Gen4 x16 slot and is compatible with systems managed through Cisco Intersight.

Q: Does the L4 GPU require external power connectors?
A: No, the L4 GPU draws its entire 72W power budget directly from the PCIe slot, eliminating the need for auxiliary power cables. This simplifies installation and enables deployment in space-constrained servers.

Q: What workloads is the L4 GPU optimized for?
A: The L4 excels at AI inference, video transcoding and streaming, virtual desktop infrastructure (VDI), graphics acceleration, and generative AI applications. Its fourth-generation Tensor Cores with FP8 precision support and hardware AV1 encode/decode make it ideal for production inference workloads and video processing at scale.

Q: How does the L4 compare to the previous generation T4?
A: The L4 delivers approximately 2.5X better generative AI performance and 4X higher graphics performance than the T4, while maintaining the same 72W power envelope and single-slot, low-profile form factor. It also adds support for FP8 precision, AV1 encoding, and DLSS 3 technology.

Q: Can multiple L4 GPUs be installed in the same server?
A: Yes, the L4's low 72W power draw and single-slot form factor enable multiple GPUs per server without requiring power distribution or cooling upgrades. However, mixing GPU types within the same Cisco HCI cluster is not supported—all GPUs must be the same model.