NVIDIA L40 48GB vs NVIDIA GeForce RTX 4090 24GB

The L40 is the newer part: Ada Lovelace, launched in 2023, against the GeForce RTX 4090's Ada Lovelace from 2022. Newer architectures typically add lower-precision formats and better throughput per watt, so check the precision rows your workload actually uses.

Dense throughput for the L40 against the GeForce RTX 4090: FP32 91 vs 83 TFLOPS, FP16 181 vs 165 TFLOPS, FP8 362 vs 330 TFLOPS. Sparse figures, where the vendor publishes them, appear under each dense number in the performance table.

Memory is 48 GB against 24 GB, fed at 864 GB/s versus 1.0 TB/s. More memory per GPU means larger models fit before you have to shard across cards, which often matters more than raw TFLOPS for inference.

Power budgets are 300 W for the L40 and 450 W for the GeForce RTX 4090. At FP32 that works out to 0.30 against 0.18 TFLOPS per watt.

For LLM training and inference, weight the FP8 and FP16 rows and memory capacity most heavily. For scientific and HPC workloads, the FP64 row is the one to read.