Conclusion
Investigate Cypress performance? Check. Talk about new drivers? Check. Do mean things to the Heaven demo? Check.
On a more serious note, there aren't many conclusions to draw outside of those that are immediately apparent throughout the article:
On a more serious note, there aren't many conclusions to draw outside of those that are immediately apparent throughout the article:
- Cypress is pretty fast versus it's predecessor (+50-60% is not too shabby, in our humble opinion), although it only manages to achieve the 2x performance mark only when DX11 is leveraged in such a way as to improve performance (hey, we warned you back in the architecture piece that there are some points which may limit the upscaling, did we not?)
- It's primarily bound by its engine clock, so we'd expect any future high-end SKU to try to bump that as much as possible
- Heavy tessellation isn't cheap, and explaining it solely by "you're setup limited, silly" is a heavy over-simplification
- We're not really fond of how the Crossfire profiling mechanism ended up being implemented - dear ATI, there's already a solution out there that's well liked by the community, and there have been no reports of users performing ritual seppuku when having to mess with a few extra settings, so please, go and implement something similar rather than reinventing the wheel (yes, we're aware that it was your competitor that came up with it, and yes, we're painfully aware of just how unpleasant it is to program UIs but still, you'd get far far more benefits from doing that)
