NVIDIA L40 CNX

The NVIDIA L40 CNX is a workstation graphics card that launched in Q4 2022. It is built on the Ada Lovelace GPU microarchitecture (codename AD102) and is manufactured on a 5 nm process.

Memory

The L40 CNX has 24 GB of GDDR6 memory, with a 2,250 MHz memory clock and a 384 bit interface. This gives it a memory bandwidth of 864 Gb/s, which affects how fast it can transfer data to and from memory. GPU memory stores temporary data that helps the GPU with complex math and graphics operations. More memory is generally better, as not having enough can cause performance bottlenecks.

Cores and Clock Speeds

The L40 CNX includes 18,176 CUDA cores, the processing units for handling parallel computing tasks. The GPU operates at a core clock speed of 1,005 MHz and can dynamically boost its clock speed up to 2,475 MHz. Complementing the processing units are 568 texture mapping units (TMUs) for efficient texture filtering and 192 render output units (ROPs) for pixel processing. Additionally, the GPU features 568 tensor cores optimized for AI-accelerated workloads and 142 RT cores dedicated to real-time ray tracing calculations.

Compatibility & Power Consumption

The L40 CNX occupies 2 PCIe expansion slots. It supports 1x HDMI 2.1, 3x DisplayPort 1.4a display connections. NVIDIA recommends a power supply of at least 700 W to handle the GPU's thermal design power (TDP) of 300 W.