900-2G193-0000-001

• NVIDIA L4 24GB, Ada Lovelace AD104 GPU on 5 Nm
• 24 GB GDDR6, 192-Bit bus, 12.5 Gbps, 300 GB/S Bandwidth
• 240 Tensor cores, 60 RT cores; 795 MHz core, 2040 MHz Boost Clock
• 72 W TDP, No Supplementary Power connectors, 250 W Recommended PSU
• Single-Slot PCIe Form factor, PCIe 4.0 x16, Passive Cooling
• Professional class, Launched 2023 Q1

The NVIDIA 900-2G193-0000-001 is an L4 24GB accelerator card built on the 5 nm Ada Lovelace AD104 GPU, launched in Q1 2023 under the Tesla brand designation. It pairs 24 GB of GDDR6 memory on a 192-bit bus at 12.5 Gbps for 300 GB/s of bandwidth with 240 Tensor cores and 60 RT cores, clocking from 795 MHz base to 2040 MHz boost. The single-slot PCIe card uses a PCIe 4.0 x16 interface, is passively cooled, and draws just 72 W TDP with no supplementary power connectors, on a 250 W recommended system supply. 900-2G193-0000-001 is a professional-class accelerator.

A passively cooled single-slot accelerator has no fan of its own and relies on server chassis airflow, which is what allows several of them to be packed into one node; combined with a 72 W TDP and slot-only power, that makes performance per watt and per rack unit the defining characteristic rather than peak throughput. With 24 GB of GDDR6 and 240 Tensor cores, the L4 is aimed at inference and media rather than large-scale training: LLM and computer-vision inference serving, video transcoding and streaming pipelines, AI-assisted video analytics, virtual workstation and VDI hosting, and cloud gaming or rendering back ends. Buyers are typically cloud and hosting providers, media and streaming platforms, and enterprise AI infrastructure teams scaling inference capacity inside existing power and cooling envelopes.