Specifications
Model
NVIDIA L4
Ordering Part Number
900-2G193-0000-000
Alternate SKU
900-2G193-0000-001
GPU Architecture
NVIDIA Ada Lovelace
CUDA Cores
7,680
Tensor Cores
240 (4th generation)
RT Cores
60 (3rd generation)
GPU Memory
24 GB GDDR6
Memory Bus
192-bit
Memory Bandwidth
300 GB/s
FP32 Performance
30.3 TFLOPS
INT8 Tensor Performance
485 TOPS
Host Interface
PCIe 4.0 x16
Form Factor
Half-height, half-length (HHHL), single slot, passive
Max Power
72 W
Video Engines
2x NVENC / 3x NVDEC, AV1 encode and decode
Cooling
Passively cooled (server airflow)
Workloads
AI inference, video transcoding, virtual desktop, graphics, simulation
Overview
The NVIDIA L4 (900-2G193-0000-000) is a universal, energy-efficient accelerator built on the NVIDIA Ada Lovelace architecture. With 24 GB of GDDR6 at 300 GB/s, 7,680 CUDA cores and 240 fourth-generation Tensor Cores, it delivers 30.3 TFLOPS FP32 and up to 485 TOPS INT8 in a 72 W envelope — roughly 120x the AI inference performance of a CPU at 99% lower power per inference.
Its half-height, half-length, single-slot passively cooled form factor fits dense 1U and 2U servers and edge cabinets where a full-height card will not. The card carries dedicated NVENC/NVDEC engines with AV1 encode and decode support, making it as useful for video pipelines as for LLM and vision inference.
QS Compute supplies the L4 in the factory-sealed 900-2G193-0000-000 and 900-2G193-0000-001 SKUs with volume pricing and 15-day sample lead time for qualified projects.
Key Benefits
72 W single-slot envelope — up to 24x more AI inference performance than CPU at the same power; 24 GB GDDR6 fits larger models and more concurrent streams; AV1 encode/decode cuts video bitrate ~30% without quality loss; PCIe 4.0 x16 with vGPU support for virtualised workloads.
Applications
LLM and vision inference at the edge and in the data center; live video transcoding and cloud gaming; NVIDIA RTX Virtual Workstation (vWS) and VDI; simulation, data science and analytics; AI-augmented security and smart-city analytics.
Request a Quote — NVIDIA L4 24GB — 900-2G193-0000-000 PCIE 4.0 INFERENCE, VIDEO & GRAPHICS GPU
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →