d2jsp
Log InRegister
d2jsp Forums > Off-Topic > Computers & IT > Motion Interpolation > On Computer
Prev1234
Add Reply New Topic New Poll
Member
Posts: 10,281
Joined: Jan 7 2015
Gold: Locked
Warn: 60%
Jul 14 2017 11:43am
Quote (ZwiX @ Jul 14 2017 01:36pm)
Still should be more processing power than needed to double the frames?


If the architectures were the same, sure...
clock speed is irrelevant here
if u compare say i7-4770 vs i7-4790k 4.9ghz beats 4ghz by ~20%
sure that makes sense
apples to apples
same instructions same pipeline same cache config p much same everything
and well comparing bulldozer to sandy bridge say fx-8150 4ghz vs i5-2300 2.8ghz... well the fx is not 30% faster it's actually slower
apples and oranges, BUT they still execute the same instructions in those workloads!

comparing ur cpu to that arm cortex cpu is more like apples to foam packing peanuts.
no you shouldnt get double the frames just bcuz u got triple the clock rate..
you don't have the instructions that workload is apparently optimized for if the tv handles it so well @ 1ghz
so you have to compensate
and compensating takes many more cycles
Effectively dividing ur cpu clock speed by the factor that is how many extra cycles it takes to execute
if u take 6x as many cycles to do the same thing the 1ghz does in one.. well.. ur 3ghz cpu may as well be 500mhz in comparison

This is why x86 emulators on ARM cpu's are so much slower than you would expect just by seeing clock speeds. They have to fake it to make it.
Same principle here

This post was edited by DCSS on Jul 14 2017 11:52am
Go Back To Computers & IT Topic List
Prev1234
Add Reply New Topic New Poll