Understanding generational differences between CPUs

Andyk5

[H]ard|Gawd
Joined
Jul 27, 2011
Messages
1,154
I am trying to understand 2 things before I get in to upgrading my computer, thread is not about should I upgrade or not. Talking about a single core of 2 generationally different CPUs

1) Is there a difference in processing speed of a CPU from 4-5 generations ago, running the same clock speed as a current day CPU, like 4770k vs 9700k. If they are both at 4.5Ghz (i know 9700k is 4.9 in reality) do they do the same amount of work and 9700k is basically more efficient as far as power/heat? Is it only the clock speed that matters? Are two 4 core CPUs many generations apart both running 5ghz provide the same performance (heat/power aside)


2) Is the "%in use Core#x" of an older CPU same as a newer generation one? As an example if a monitoring software is reporting Core 5 of 4770k is being used at %50, and we replace the CPU with 9700k will the %used go down or it will stay the same and the application will run faster? For more detail, I am wondering if a CPU core is not being maxed out, does that mean there is more headroom in the CPU and the application is not using it, or the maximum the application can make use of that core is %50 and a faster CPU would have fared better/faster at %50 and provide more fps/processing power.


Sorry for the convoluted post, It is a pretty poor attempt at trying to communicate what is in my head.
 
IPC has increased over the generations. Not as much as it had in the past but usually single digits each generation.

Is the "%in use Core#x" of an older CPU same as a newer generation one? As an example if a monitoring software is reporting Core 5 of 4770k is being used at %50, and we replace the CPU with 9700k will the %used go down or it will stay the same and the application will run faster?

That may depend on the application.
 
Your second question - depends a lot of bottlenecks.

Your %utilitization shouldn't move if something other than the CPU is bottlenecking you.

If the CPU is the bottleneck, then as the IPC goes up, the common sense thinking would be that %utilization for the same amount of work should go down. But then there's "race to idle" to consider - the chip goes up to a very high % utilization, but stays there for a shorter period of time, as time is another factor in utilization, not just % load.
 

Not sure how valid that graph is, but it comes pretty close to answering your first question.

This looks about right.

2 depends on the application. Given each thread is now technically faster at the same clock speed vs your haswell, if only by a small amount, the utilisation for the same workload should be 'less', or the workload will pull the same % but finish/run marginally faster.
 
Back
Top