SatansReverence
Hipster Princess
The variation is dependent on power, not heat/temperature. It is a known variable, unlike temperature which can vary depending on variation with the APU itself but also room temperature.
Power = Heat created. Again, as I said, the console will be designed to perform at it's stated frequencies in a room at 40+ degrees. It does this by limiting it's overall power draw to prevent it creating more heat than it can dissipate in that ambient temperature. This is a simple bit of overhead that all console manufacturers have to take into account unless they want their console to perform great in Norway but like trash in Australia.
Given it is a known variable, the time + power required to render the frame can be calculated and compared to this limit, just as it can be with the limit from fixed clocks. I agree it appears like there is more work, but that's what the tools are there for.
Player actions aren't a known variable however. Unless it is an entirely scripted sequence, there will be unknowable variables.
What I don't understand is why you feel this requires *far* more work
Go on, quote me where I said it would take >far< more work.
You are looking at it from the negative perspective though, variable down. From the perspective of the same APU with the same cooling at fixed clocks it is variable upward.
It's a negative because it is in direct comparison to a system that is not only more powerful by doesn't have that variance in performance.
The performance is predictable and consistent for a given game though.
Games aren't a CGI movie. The dynamics of players, A.I and physics aren't predictable.
Then where both are over loaded you have the option to optimise the CPU or GPU processes.
>or<
Again, as I've said, if a developer wants a smooth game experience, they will need to leave more performance on the table, or spend more time optimising than they otherwise would need to.
It's a different approach to what's come before, but it's about efficiency and an elegant solution to getting higher performance for your power budget.
It's a sub optimal approach compared to your competition that is already more powerful and doesn't need to account for the variance.
No, the people that replied to me are people that have been attacking variable frequencies with ignorant arguments. Don't worry.
This is hilarious coming from the guy who tried to add the 14%* smartshift performance boost twice in an ass backwards way.
Last edited: