NVIDIA L4

Overview

The NVIDIA L4 Tensor Core GPU is a high-performance accelerator designed to deliver exceptional AI inference, video processing, and graphics capabilities. Built on the Ada Lovelace architecture, it offers a balance of power efficiency and computational prowess, making it ideal for data centers, edge deployments, and enterprise applications.

Features

• High Performance: Achieves up to 485 teraFLOPS of FP8 Tensor performance, ensuring rapid processing of AI workloads.
• Energy Efficient: Operates with a maximum thermal design power (TDP) of 72W, suitable for dense deployments.
• Advanced Memory: Equipped with 24GB of GDDR6 memory and 300GB/s memory bandwidth, supporting large datasets and complex models.
• Versatile Connectivity: Features a PCIe Gen4 x16 interface with a 64GB/s interconnect, ensuring high-speed data transfer.
• Comprehensive Software Support: Compatible with NVIDIA AI Enterprise software suite, providing certified frameworks and tools for AI development.
• Virtualization Ready: Supports GPU virtualization technologies, enabling efficient resource allocation in multi-tenant environments.

Specifications

• GPU Architecture: Ada Lovelace
• CUDA Cores: 7424
• Tensor Cores: 240
• RT Cores: 60
• Base Clock Speed: 795 MHz
• Boost Clock Speed: Up to 2040 MHz
• Memory Size: 24GB GDDR6
• Memory Bandwidth: 300GB/s
• Interface: PCIe Gen4 x16
• Form Factor: Low-profile, single-slot
• Maximum Power Consumption: 72W
• Thermal Solution: Passive
• ECC Memory: Yes
• Virtualization Support: Single Root I/O Virtualization (SR-IOV)
• Display Connectors: None
• Graphics APIs: DirectX 12 Ultimate, OpenGL 4.6, Vulkan 1.3
• Compute APIs: CUDA 12.0, OpenCL 3.0