I am very curious - as a game programmer, you want all the performance the HW can give you, but a lot of complexity (at the hw level) is in the name of performance (OoO, superscalar, etc.) Would you rather have a simple scalar in-order core even in the domain of video games?
-
-
Most of the performance increases you see from year to year in modern CPUs come directly from the complexity. They can't make it go any faster without adding more complexity, that's why they do it. If they could have kept it simple and increased the clock rate, they would have.
-
So really the only way you get "simpler and faster" would be if you made the software much more parallel, using way way way more threads than it does now. Otherwise, simplifying the CPU will only make it slower for current code. I'm pretty sure of that?
- Show replies
New conversation -
-
-
You can make faster CPUs, but you have to rewrite ALL the software. Which isn't practical. Jon is technically correct, except for the problem of legacy. But legacy is exactly what makes these things useful in the first place. Discard legacy and all you have is a games console.
-
I don't actually know what you're referring to here, beyond the part I said about somehow making software use more threads. Are you suggesting _single threaded_ hot loops today would go much faster if we, say, didn't encode them as x64?
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.