NVIDIA RTX Pro 6000 vs L40S: Key Differences

Q: What is the difference between the NVIDIA RTX Pro 6000 and L40S?

The RTX Pro 6000 is a newer Blackwell-generation GPU with 96GB of GDDR7 and native FP4, while the L40S is an Ada-generation card with 48GB of GDDR6 and an FP8 ceiling; OpenMetal carries the RP6000, not the L40S.

Explore GPU servers

The headline differences are memory and generation. The RP6000 roughly doubles GPU memory (96GB vs 48GB) and raises bandwidth (1.79 TB/s GDDR7 vs 864 GB/s GDDR6), and Blackwell adds native FP4 (NVFP4) on top of FP8. That means larger models and batches per card and higher low-precision inference throughput.

The L40S draws less power (around 350W vs 600W) and has historically been cheaper and more widely available, which can matter for dense, power-constrained inference fleets where 48GB per card is sufficient.

On OpenMetal the RP6000 ships as a single-tenant bare metal server with dual Xeon 6530P, 1TB DDR5-6400, and a 6.4TB NVMe data drive, on fixed monthly pricing. For the highest memory bandwidth (largest-scale training), the HBM-class H200 is the step beyond both cards.