Dolphin, the GameCube and Wii emulator - Forums

Thank you for the reply!

Well, I always used Geekbench scores as source to compare CPU performance for emulators like PPSSPP and Dolphin that are available on Android and Windows, and they seem to fit. As an engineering student I like a lot to play with numbers, so I always make some calculations comparing the CPU performance (how much of CPU is being used to run the game) to Geekbench single-thread score, than I try to predic how a mobile phone will perform beased on Geekbench score too, my "predictions" are very accurate, I can confirm it by myself using the results getten from my phone and PC running PPSSPP.

(09-25-2017, 01:48 PM)Guilherme Wrote: [ -> ]How mobile score can be compared to desktop scores in Geekbench? Do they match? Because it's weird to a 2.0GHz mobile CPU to do 4000+ points while 7th gen Intel CPU with 3.5GHz (like i5 7600) does 4500 to 5000. Is mobile tech so advanced? I don't have an advanced knowladge about this, but I try to learn as much as I can.

The biggest difference is the instruction set difference.
X86-64 chips are CISC (complex instruction set computers) while ARM chips are RISC (reduced instruction set computers). Though CISC generally gets translated to RISC microcode prior to execution on the chip.

CISC is more complicated so unless particularly well optomized there can be pipeline bottlenecks.
RISC is much simplier but takes more instructions to do the same thing a single CISC instruction can do.

This is a gross oversimplification but basically in very specific tasks pure RISC based chips can achieve remarkable speeds, but they lack the overall flexability and general speeds of more complicated CISC chips. So while geekbench might be using simular tasks to test both mobile and desktop they are not running the same code for both so it is not an exact apples to apples comparison. This is also a good example of why benchmarking is good for a general refrence but the best data comes from testing the exact workload you expect to do under real(ish) world conditions.

(09-25-2017, 04:10 PM)TKSilver Wrote: [ -> ]The biggest difference is the instruction set difference.
X86-64 chips are CISC (complex instruction set computers) while ARM chips are RISC (reduced instruction set computers).  Though CISC generally gets translated to RISC microcode prior to execution on the chip.

CISC is more complicated so unless particularly well optomized there can be pipeline bottlenecks.
RISC is much simplier but takes more instructions to do the same thing a single CISC instruction can do.

This is a gross oversimplification but basically in very specific tasks pure RISC based chips can achieve remarkable speeds, but they lack the overall flexability and general speeds of more complicated CISC chips.  So while geekbench might be using simular tasks to test both mobile and desktop they are not running the same code for both so it is not an exact apples to apples comparison.  This is also a good example of why benchmarking is good for a general refrence but the best data comes from testing the exact workload you expect to do under real(ish) world conditions.

It's clear now. Thanks!