Why do Graphics cards clock a lot lower than cpus?

[person223]

Why do Graphics cards clock a lot lower than cpus? 

[minibois]

They are way different and can't be compared with a single number like that.

GPU's are made for a specific purpose and run a different set of instructions.

[PopsicleHustler]

Simply because they're not the same. Its as if I asked you "Why do jet engines and piston engines run at different RPM?"

 

GPUs and CPUs work in different ways and do different tasks.

[person223]

1 minute ago, PopsicleHustler said:

Simply because they're not the same. Its as if I asked you "Why do jet engines and piston engines run at different RPM?"

 

GPUs and CPUs work in different ways and do different tasks.

I fell like i asked a dumb question.

[mariushm]

4 minutes ago, person223 said:

Why do Graphics cards clock a lot lower than cpus? 

 

It's different designs... just 8 / 16 / 32 big cores on a CPU compared to hundreds or thousands of tiny specialized cores that work in parallel to create the image you see on the screen.

GPU chips can also be much bigger in size (so more difficult to dissipate the heat produced) and it becomes more difficult for all the cores to exchange data between them at higher speeds.

[PopsicleHustler]

2 minutes ago, person223 said:

I fell like i asked a dumb question.

No such thing as "dumb question" in PC world. We cannot know everything.

[WereCatf]

3 minutes ago, PopsicleHustler said:

No such thing as "dumb question" in PC world. We cannot know everything.

Go browse e.g. Reddit and you'll quickly see how wrong that claim is; there are plenty of idiots out there that very definitely do ask exceedingly dumb questions. Not saying OP's question falls into that category, but, like Einstein has said: "Only two things are infinite, the universe and human stupidity, and I'm not sure about the former."

[WereCatf]

18 minutes ago, person223 said:

Why do Graphics cards clock a lot lower than cpus? 

You've gotten a good number of replies so far, but I'd like to use an analogy to really put the difference between CPUs and GPUs into easy-to-understand terms:

 

Think of a typical family sedan that people drive around with -- it's good for going shopping with in a city, people drive to work with them, chauffeur their kids to/from hobbies and school and such, maybe even haul small furniture with. Those kinds of cars are kinda like CPUs -- good for a lot of things, though not necessarily the best thing for any of them.

 

Now, think about a rocket-car: it goes really, really fast, but you can't carry even a small bag in one, there's only room for one person and even then just barely, they can't even turn! They do one thing really, really well, though: they go forward extremely fast. That's kind of like a GPU in some ways, ie. it has a specific purpose that it's good at and it does that purpose better than almost anything else.

 

In a manner similar to the two above cars, most comparisons between them just don't make much sense. Technically there are some kinda-sorta similar things, but they're often implemented in such different manner that you just can't apply concepts from one to the other. You probably haven't realized it, but your Ethernet-card has a clock, your SSDs and HDDs contain multiple clocks, your mobo contains a good handful of clocks, USB needs a clock at a certain speed and so on -- comparing those doesn't work, because they're for different purposes.

[Sauron]

27 minutes ago, person223 said:

Why do Graphics cards clock a lot lower than cpus? 

There's no specific reason, they were just engineered in such a way that the maximum frequency is relatively low in exchange for doing a lot more each cycle. For that matter you can have a 500MHz CPU that is faster than a different CPU at 4GHz, it all depends on how much work it can do in a single cycle (which is what we mean when we talk about IPC).

 

Now, obviously GPUs are made for a specific purpose (though more and more they're being used for general purpose computing) which inevitably influences their design towards extreme parallelism at the cost of sequential performance - which means that in order to have a larger, more parallelized chip to work you usually can't clock it as high as a chip with only a handful of cores, and the instructions per clock of each individual "core" are significantly fewer compared to a CPU.

[person223]

ok