NVIDIA GeForce RTX 3080 Ti 12GB vs NVIDIA A40 48GB

Both GPUs launched in 2021: the GeForce RTX 3080 Ti on Ampere and the A40 on Ampere.

Dense throughput for the GeForce RTX 3080 Ti against the A40: FP32 34 vs 37 TFLOPS, FP16 34 vs 150 TFLOPS. Sparse figures, where the vendor publishes them, appear under each dense number in the performance table.

Memory is 12 GB against 48 GB, fed at 912 GB/s versus 696 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 350 W for the GeForce RTX 3080 Ti and 300 W for the A40. At FP32 that works out to 0.10 against 0.12 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.