Is Higher Clock Speed or More Cores Better for a CPU?

[MechSoup]

I'm not sure I understand the advantages or disadvantages of each.

I imagine Multi-core (in the dual case) as two processors working on the same information to complete the task (reading, writing, whatever) in half the time.
I imagine Clock Speed (I'm not sure if this is what you call it: ie: 3.4Ghz processor) as the rate at which it processes the information.

If the above definitions are correct, then couldn't a high multi-core unit (with low clock speed) complete the same information in the same amount of time as a single core (with high clock speed)? Yet the single core is much cheaper?

If this is the case, when why is everything going multi core and not increased clock speed?

I have read Wikipedia and different articles, but I just need a down to Earth definition.
Thanks!

 

[bk7794]

I'm not sure I understand the advantages or disadvantages of each.

I imagine Multi-core (in the dual case) as two processors working on the same information to complete the task (reading, writing, whatever) in half the time.
I imagine Clock Speed (I'm not sure if this is what you call it: ie: 3.4Ghz processor) as the rate at which it processes the information.

If the above definitions are correct, then couldn't a high multi-core unit (with low clock speed) complete the same information in the same amount of time as a single core (with high clock speed)? Yet the single core is much cheaper?

If this is the case, when why is everything going multi core and not increased clock speed?

I have read Wikipedia and different articles, but I just need a down to Earth definition.
Thanks!

I believe that they have reached a limit on the speed of the cpus. I am into overclocking myself, which means that I manually bring the speed of the cpu higher then its rated clock speed. I have noticed that some chips are purposely underclocked. This means that they are just rebranded higher end chips generally speaking from what I have heard. However as the clock speeds get higher the heat gets higher. With the more heat a more expensive cooler will be needed and that means the price of the computer will go up. However, there is still a cap on how fast it can go without needing extreme forms of cooling.

In multi core like the intel core 2 series, each core could process a 1 and a 0. So it would take two separate cores to process a 1 and a 0. However with the new series of Core I7s and other cpus before the core 2 supports hyperthreading. This means that each core, or cpu if you will, will be able to process a 0 and a 1 using the same core rather then the old architecture that needed two separate cores.

We are moving to multi core cpus because they can handle more information at a given time and especially with the same speed. So that means more information done faster. The single core could technically keep up, but the speed to do so would have to be very high.

I hope this answers your question, and anyone correct me if I am wrong.

 

[russ_watters]

The multi-core advantage is not n*x as some calculations are linear, with the next calculation based on the results of the last. Such operations can't easily be split to run in parallel.

 

[bk7794]

The multi-core advantage is not n*x as some calculations are linear, with the next calculation based on the results of the last. Such operations can't easily be split to run in parallel.

Right, this is why programs are specially coded for multi core cpus. Right?

 

[Pengwuino]

As you increase a CPUs core speed, you generate a lot of heat. At some point it becomes impractical to continue pushing the envelope. With smaller architectures, it became possible to embed two or four or eight cores onto a single chip. Creating a 4GHz processor for personal computers is a much less convenient task for chip designers than putting two 2GHz processors on one chip. Now you have 8-core CPUs that, even at 2GHz, would be impossible for CPU makers to match and turn into a commercial product (16GHz?).

The catch with multi-core processors is the fact that not all software can use them. Back in the day when they were first coming out they were practically useless. Very few programs were able to use parallel processing and one core would go completely untouched. I can't say for sure, but I would bet it's not a trivial task to redesign software to use parallel processing.

 

[DavidSnider]

It's not easy to rewrite software to be parallel by itself, so not many individual applications get much of a speed increase with mutlicore -- BUT it is pretty easy to split the work of multiple independent programs across cores.

Basically, your spreadsheet doesn't go 8x faster, but it doesn't slow down because you are also chatting on skype, playing an MP3, filtering e-mail, browsing the internet, etc at the same time.

 

[MechSoup]

It would make sense then, that multi core processors are also more efficient then high clock speed single core processors as a single core will generate more heat to do the same amount of work?

Overclocking a multi core would destroy the efficiency gains. Alas! Performance will always prevail! Thank you gentlemen for your input.

 

[Pengwuino]

It would make sense then, that multi core processors are also more efficient then high clock speed single core processors as a single core will generate more heat to do the same amount of work?

Overclocking a multi core would destroy the efficiency gains. Alas! Performance will always prevail! Thank you gentlemen for your input.

Wait, what? Are you talking about efficiency in the thermodynamic aspect? The problem is not with them being too hot to be efficient (in the thermodynamic or data manipulation sense), it's that you can only have a processor become so hot before the cooling requirements become too much for desktop computers.

 

[MechSoup]

Wait, what? Are you talking about efficiency in the thermodynamic aspect? The problem is not with them being too hot to be efficient (in the thermodynamic or data manipulation sense), it's that you can only have a processor become so hot before the cooling requirements become too much for desktop computers.

Sure. It would make sense that a processor that could perform that same work in the same amount of time with less heat is more efficient in its power consumption?

I would assume that you are referring to the limiting factor? As in why we do not create higher frequency CPUs? I suppose that you could look at it on a thermodynamic perspective.

Also, if a Multi Core processor says that it has a clock speed of 2.8Ghz, does that mean that each core operates at this speed? Or is it that the combined output is equal to that speed? Maybe I am looking at it wrong. . . if it is an operating frequency, then they must be that same? Right? I guess I was thinking that each processor (in the dual case) would be 2.8Ghz/2. . .

 

[Pengwuino]

Sure. It would make sense that a processor that could perform that same work in the same amount of time with less heat is more efficient in its power consumption?

Chip designers are really not at all worried about a CPUs power consumption unless you're dealing with laptop CPU. With laptop CPUs, you sometimes don't want to have a higher clock speed. To be honest, none of that is really important to laptops either. CPUs are rarely going at 100% and if you're on a laptop, you're very very rarely hitting 100% on the CPU.

I would assume that you are referring to the limiting factor? As in why we do not create higher frequency CPUs? I suppose that you could look at it on a thermodynamic perspective.

No, you can't. A CPU processing data is not a thermodynamic process. There is heat generated and there are parts of the whole system you can view thermodynamically, but that does not mean the CPU process is subject to the same kind of analysis you'd do on a car engine.

Also, if a Multi Core processor says that it has a clock speed of 2.8Ghz, does that mean that each core operates at this speed? Or is it that the combined output is equal to that speed? Maybe I am looking at it wrong. . . if it is an operating frequency, then they must be that same? Right? I guess I was thinking that each processor (in the dual case) would be 2.8Ghz/2. . .

No, when manufacturers say they have a 3.0ghz quad core CPU, that means there are 4 cores running at 3ghz each.

 

[MechSoup]

Chip designers are really not at all worried about a CPUs power consumption unless you're dealing with laptop CPU. With laptop CPUs, you sometimes don't want to have a higher clock speed. To be honest, none of that is really important to laptops either. CPUs are rarely going at 100% and if you're on a laptop, you're very very rarely hitting 100% on the CPU.
No, you can't. A CPU processing data is not a thermodynamic process. There is heat generated and there are parts of the whole system you can view thermodynamically, but that does not mean the CPU process is subject to the same kind of analysis you'd do on a car engine.

I can see what you mean, and your logic is sound. I even found an article proving your point. I guess what I was thinking was that the power required to run a multi core was the same as a single core, which is not true.

No, when manufacturers say they have a 3.0ghz quad core CPU, that means there are 4 cores running at 3ghz each.

Thank you for the clarification. ;)