How do multiple cores help increase performance? Why are 4 cores

>>58175819
Fuck off stupid weaboo.
>>>/sqt/

>>58175826
It's not a stupid question or I would've posted it there.

>>58175819
clock comparison will depend on the architecture scaling. more cores will only be faster if the load can actually be distributed.

it isn't faster. simply put, it easier to fabricate a chip with 4 smaller cores instead of one powerful one.

They don't or AMD would be in the lead.

File: Amdahl[1].png (34KB, 648x486px) Image search: [Google]

34KB, 648x486px

>>58175819
>How do multiple cores help increase performance?
It's basically several small CPUs in one.

>Why are 4 cores @ 1GHz faster than 1 core @ 4GHz?
they are not.

>>58175819
depends on the software you want to run you stupid anime poster

>>58175843
it is a stupid question

>>58175867
you're simply wrong, seeing how we had 3GHz pentium 4s before dualcore and quad became a thing, you fucking retard

>>58175883
true

>>58175900
false


As to answer your question: all operating systems nowadays are capable of utilizing more than 1 core.
However, programs can't really benefit from the performance gains, unless they're specifically designed to run on multiple CPU cores, in which case you can expect a 70% or so performance gain vs single core performance

Single core just scales 1:1 more or less, while 2+ cores require some extra work, but can generally give you more performance.

More cores aren't faster.
They are more unstable to work with and require special attention from a developer in terms of software.

Less cores = more stable = no special management software required to make it work and complicate things.

This is why so many people are begging for Graphene, albeit that won't happen for the next 2 decades at least the way Intel are jewing and AMD are failing.

>>58175926
>false
>proceeds to agree with me

Taiga a cute

>>58175819
different but same question: Why bother with making the internals of a cpu smaller? why not make it bigger and fit more cpu stuff inside one?

>>58175998
Do you mean the transitor size or acctual die size?

Smaller transistors (like 14nm) means more transitors in smaller space and more efficient transistors = higher clocks/less heat

Smaller die is because they can fit more into a wafer and the risk of one die being broken is smaller.

File: 2016-12-26_121825.jpg (367KB, 701x544px) Image search: [Google]

367KB, 701x544px

>>58175986
Taiga a slut.

Don't sexualise the Taiga

>>58175954
>This is why so many people are begging for Graphene,
They are to busy begging, so they don't optimize their programs?

4 cores @ 1 GHz are not necessarily faster than 1 core @ 4 GHz, assuming same CPU architecture. But strictly speaking, the 4 GHz CPU is going to run stupidly hot. Intel started pushing multicore CPUs after they were hitting a wall with heat on their Pentium 4s.

Modern CPUs generally don't go higher than 4-5 GHz after overclocking, but adding more cores allows tasks to be run in parallel, thus getting more work done on that same 4 GHz. To put things into perspective, normally my kernel is going to divide up all of my processes and threads into time slices, and schedule them on the CPU. If I have one core, it cannot be running the embedded JavaScript in my Discord tab while decoding audio in my YouTube tab. Instead, it will have to switch between the two. My Quad Core CPU can do both simultaneously, and then let both cores be idle for a bit longer (which subsequently also means less energy usage on top of the performance boost).

>>58175819
If an algorithm is easily parallelizable then it's possible that something like 4x 1.5ghz processors might be faster than 1x 4ghz processor, but there's some overheard with parallelizing tasks so it's never going to be the case that 4x1ghz == 1x4ghz.

If the program's algorithm is not easily parallized, then it's not going to help performance much.

>>58175848
Why do you think apple can make the a10 fusion, which smashes all the 3010101 cored mediakek trash or Qualcomm housefire chipsets

>>58175819
A single core at X Ghz will always be faster than multiple cores with a combined clock rate of X Ghz.

However power consumption increases with higher clocks and voltage quadratically which means Y cores at X/Y Ghz will always consume less total power than a single core at X Ghz.

File: 1333652693240.png (301KB, 1444x1705px) Image search: [Google]

301KB, 1444x1705px

>20 replies
>no one's posted any of the cow2beef edits yet

Because 4 GHz cores can't exist because electronics only work well in low frequencies. That's why we make more cores instead.

Now back to >>>/a/ you dumb weeb.

>>58179471
The 6700k is 4Ghz at stock and easily overclockable to 5Ghz with proper cooling.

>>58175819
1 core, 4 tasks
>do task
>do task
>do task
>do task
>finished

4 cores, 4 tasks
>do task >do task >do task >do task
>finished


asfor 4x1.0ghz vs 1x4.0ghz
who the fuck even does that?

>>58179491
Nice try, CPUcuck. The ALU is 4 GHz but the whole core isn't.

You may be able to distribute work onto them, but there is a lot of hassle with synchronizing work between cores. It is totally possible and not at all far fetched that a poor implementation of multi threading will perform significantly worse on multiple cores than on a single core.

>>58179853
more like

4 cores 4 tasks
>dsk
>ddo ta
>
>dodo tskaskta

>>58180119
kek, or:
>done-dndodoe doed e doe de e

>>58179853
How we imagine it is
>>58180119
How it really is

>>58175926
>... we had 3GHz pentium 4s before dualcore and quad...

AND... now have 9000 cores still around the the 3GHz, because its more practical than to have 6GHz cores

>>58175819
but that is not true desu
if you have a program written without any multirthreading capabilities it will always be faster with a single highly clocked core