Compare two platforms.
Any two processors or accelerators in the series, workload by workload and size by size. The link updates as you choose, so a comparison can be shared as it is.
As a table
Size by size
How many times faster one platform is than the other at every problem size. Hover a cell for both throughputs.
Full measurement record
The two machines
Results are throughput, paths or options priced per second (Black-Scholes counts all 100 passes of a call), at the same problem size on both platforms, so the ratio is the two platforms' relative performance. Energy is watt-hours per million, billion or trillion paths or options at the size the results rank at, from the processor package on CPUs (memory excluded) and the board sensor on GPUs (host excluded), so energy is compared GPU with GPU and CPU with CPU, never CPU against GPU; within 10% is level; lower is better.