Due to ratio of compute / memory, GPU-powered systems tend to have shorter run times
Shorter run times require higher measurement resolution to achieve accuracy
This poses a challenge for measuring such systems without specialized equipment
Solution
I have no clear path to a solution here. The next step should be to gather concrete information about typical run durations for different architectures.
Problem statement
Solution
I have no clear path to a solution here. The next step should be to gather concrete information about typical run durations for different architectures.