Built on the NVIDIA Ada Lovelace architecture, the NVIDIA L20 is a dual-slot, full-height, full-length PCIe expansion card equipped with 96MB of L2 cache and 48GB of GDDR6 ECC memory, delivering a bandwidth of 864GB/s. It offers 239 TOPS of INT8/FP8 compute performance, with higher precision reduced by half progressively, but does not support FP64. With a power consumption of 275W, it features high performance and scalability.
Boasting powerful computing capabilities and high-density memory capacity, it excels in handling complex cloud inference tasks such as image recognition and natural language processing. Additionally, the L20 is suitable for cloud computing platforms requiring high performance and efficiency, such as data centers and cloud service providers.
Brand | NVIDIA | Model | NVIDIA L20 |
Part Number | NVIDIA L20 | Type | Graphics Card |
Chip Manufacturer | NVIDIA | Chip Model | NVIDIA L20 |
Core Bit Width | 384bit | Graphics Card Slot | PCI Express 4.0 |
Interface | PCIe | Video Memory Capacity | 48GB |
Memory Type | GDDR6 | Memory Bit Width | 384bit |
Core Clock Frequency | 1440MHz（MHz） | Memory Clock Frequency | 2250MHz（MHz） |
Fourth-Generation Tensor Cores
Provide hardware support for structured sparsity and the optimized TF32 format, delivering
out-of-the-box performance gains for faster AI and data science model training. They can leverage DLSS technology to accelerate AI-enhanced graphics features, enabling higher resolutions and
superior performance in select applications.
3rd Generation RT CoreIncreased throughput and concurrent ray tracing and shading capabilities enhance ray tracing
performance, thereby accelerating rendering in product design as well as architecture,
engineering, and construction workflows. Hardware-accelerated motion blur technology can be
utilized to deliver stunning real-time animation effects,
helping to understand the real-world performance of photorealistic designs.
CUDA CoreAccelerated single-precision floating-point (FP32) throughput and higher energy efficiency can
significantly improve performance in workflows such as 3D model development and computer-aided
engineering (CAE) simulation. Enhanced 16-bit Math Capability (BF16) support is available for
mixed-precision workloads.
NVIDIA Omniverse EnterpriseAs the engine for NVIDIA Omniverse™ in the data center, the NVIDIA L20 delivers powerful RTX and
AI capabilities,supporting workloads such as extended reality (XR) and virtual reality (VR) applications,
design collaboration, and digital twins. For extremely complex Omniverse workloads, the NVIDIA
L20 supports accelerated ray tracing and path tracing rendering of materials, enabling physically
accurate simulations and generating photorealistic 3D synthetic data.
Rendering and 3D GraphicsWith the NVIDIA L20 running professional 3D visualization applications, creative professionals can
iterate on more content, render faster,and leverage substantial performance advantages, thereby
improving productivity and accelerating project delivery. Artists and designers can process complex
geometry and high-resolution textures in real time to create photorealistic designs and simulations,
supporting high-fidelity creative workflows.
High-Performance Virtual WorkstationsWhen combined with NVIDIA RTX™ Virtual Workstation (vWS) software*, the NVIDIA L20 delivers
powerful virtual workstations from the data center or cloud to any device. Millions of creative and
technical professionals can access demanding applications from anywhere, achieving not only
exceptional performance comparable to physical workstations but also meeting enhanced security
requirements.
Streaming and Video Content
With three video encoding and three video decoding engines, the NVIDIA L20 takes streaming and video content workloads to the next level. With the addition of AV1 encoding, the L20 delivers breakthrough performance and lower TCO for broadcast streaming, video production, and transcoding workflows.
Technical Support
Service & Support
Company Introduction
