cores vs gigahertz

[BenoitWW]

Completely out of curiosity, 

 

presume that there are 2 identical CPUS, each having 4 cores at 4.0ghz

 

Now assume CPU 1 loses half its cores and CPU 2 halves its clock speed. 

 

Which CPU would perform best overall?

[saladcrack]

Depends, gaming usually prefers clock speed over cores so CPU 1 would be better with that but multi-core stuff like rendering I think CPU 2 should be better.

[Princess Luna]

Basically more cores allows you to do more things at once while more clock speed will allow you do the things faster, that's basically it, good enough nowadays high end CPUs can focus on increasing single core/dual core clock speed for the applications even when having 4 or more cores thanks to Turbo Booster.

[SCHISCHKA]

why dont you try it? you can go into BIOS and turn off cores and reduce clock speed. Im pretty sure the higher speed with less cores will be better for most tasks except for certain games that require a quad core. What you will find with older games that will work on a lower clock speed is they also do not make use of 4 cores.

[flibberdipper]

Most applications tend to favor higher clocks over more cores. Hell, that's how I'm able to use a G3258 for everything that I do.

[SCHISCHKA]

There were a lot of discussions when dual & quad core CPUs came out; we were all asking why would anyone need an expensive multi-core core CPU there is no software that can multi-thread. haha

[APasz]

They will perform the exact same but that assumes the program is able to scale correctly between cores and speed.

Of course in the real world it will depend on the application and how well it's able to spread processing or how much over head it has.

[crowy]

Not forgetting the instances where you are not doing a single task and change the Affinity of the programs that are running so that they don't all try to use the same Core

[Glenwing]

This is an interesting question. It really depends on the software and optimization. The purely theoretical answer is that there would be no difference at all, the performance change would be exactly the same. I know the common word is "A quad-core at 3.0 GHz is not the same as a 12.0 GHz single-core, it doesn't work like that", but actually it sort of does work like that in terms of raw computational power. If a CPU core executes X computations per cycle, then creating a chip with four times as many cores or four times as many cycles per second both give you the same result: 4X theoretical performance. It's only in practice that this doesn't follow through, due to the way software utilizes the processor.

 

Here are some benchmarks I did; the first set is done at 3.0 GHz with increasing core counts. The second set is done at 4.5 GHz (50% higher frequency) at each core count. (hyperthreading is disabled in the two sets shown here, but I have more results with hyperthreading enabled for both cases if anyone is interested :P)

 

 

Some interesting comparisons can be made. The second set of tests (+50% frequency) gives roughly +50% performance across the board, dropping slightly as core count increases, but it's not by much. The scaling with core count is subject to much stronger diminishing returns, as some might expect from games.

 

The change from 2 cores to 3 cores (+50% more cores) gives 78% more performance. However the change from 4 cores to 6 cores (again +50% more cores) gives only 36% more performance.