NVIDIA RTX 6000 Ada 48GB vs NVIDIA RTX 5000 Ada 32GB

Both GPUs launched in 2023: the RTX 6000 Ada on Ada Lovelace and the RTX 5000 Ada on Ada Lovelace.

Dense throughput for the RTX 6000 Ada against the RTX 5000 Ada: FP32 91 vs 65 TFLOPS, FP8 729 vs 522 TFLOPS. Sparse figures, where the vendor publishes them, appear under each dense number in the performance table.

Memory is 48 GB against 32 GB, fed at 960 GB/s versus 576 GB/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 RTX 6000 Ada and 250 W for the RTX 5000 Ada. At FP32 that works out to 0.30 against 0.26 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.