Lambda has a very interesting benchmark page https://lambdalabs.com/gpu-benchmarks https://lambdalabs.com/gpu-benchmarks If you look through the throughput/$ m

Lambda has a very interesting benchmark page https://lambdalabs.com/gpu-benchmarks https://lambdalabs.com/gpu-benchmarks If you look through the throughput/$ metric, the V100 16GB looks like a great deal, followed by H100 80GB PCIe 5. For most benchmarks, the A100 looks worse in comparison