If you are willing to sacrifice some realism to get faster rendering, I would actually advise:
- increasing GI vs AA - this will give you better reflection/refraction quality per pass (but each pass will be slower to render) - you can do a few tests in low resolution rendering to the same time limit with various GI vs AA numbers ranging from 4 to 64 (4, 8, 16, 32, 64)
- lowering light samples multiplier to 1 - you don't have any area shadows here (most likely) and this may give you a little speedup
- lowering Max Sample Intensity as much as possible (5-10, but maybe even 1 would work?)
- then going to the System tab and setting Highlight clamping as low as possible - you can run IR and start with something like 1. Then move it up or down and observe the effect. At some point the differences between the default value of 0 and the new value X should be barely visible and this is your sweet spot where you get faster rendering, less noise, and just a little bit darker highlights. Of course this clamps the sample intensity so you won't be able to do "proper" post-processing.
More info:
https://coronarenderer.freshdesk.com/support/solutions/articles/12000006462-what-is-highlight-clamping-Other than that, enable denoising (I would bet on the Intel AI denoiser).