NVIDIA L4 GPU: The energy and cost-efficient successor to NVIDIA T4
tl;dr
Characteristics and performance of NVIDIA L4 Tensor Core GPU.
Introduction
At the NVIDIA GTC 2023 keynote, NVIDIA introduced NVIDIA L4 Tensor Core GPU, designed to become NVIDIA T4 successor.
The L4 GPU is a high-performance accelerator created to offer energy-efficient acceleration for a variety of applications, including video processing, AI, visual computing, graphics, virtualization, and other computing tasks. It is powered by the advanced NVIDIA Ada Lovelace architecture and comes in a compact low-profile form factor, making it an affordable and power-efficient solution for delivering high throughput and low latency performance on the edge, data centers and the cloud.
Key upgrades of NVIDIA L4
Generative AI Performance: L4 Tensor Core GPU can deliver up to 2.7X higher generative AI performance compared to the previous GPU generation. This can be used to optimize graphics performance and enable larger image generation.
Better Energy Efficiency: The L4 Tensor Core GPU can deliver up to 99 percent better energy efficiency and lower total cost of ownership compared to traditional CPU-based infrastructure.
Real-time AI Video Pipeline Performance: The L4 Tensor Core GPU can transform video applications with its power, allowing hosting up to 1,040 concurrent AV1 video streams at 720p30 for mobile users. NVIDIA claims that the L4 is capable of delivering 120 times higher AI video performance than CPU-based solutions.
Real-Time Rendering and Ray-Tracing Performance: L4 Tensor Core GPU can deliver over 4X higher rendering performance for AI-based avatars and real-time rendering, as well as over 3X higher ray-tracing performance.
Benchmarks of L4 performances on Generative AI and AI video workloads performed by NVIDIA.
The NVIDIA L4 Tensor Core GPU is a powerful and energy-efficient solution for video, AI, and graphics workloads. The GPU has great potential to replace the very popular Tesla T4.
