Forum: Poser - OFFICIAL


Subject: 4 cores to 6 cores...no significant improvement in Pro2010

Photopium opened this issue on Sep 09, 2011 · 44 posts


kawecki posted Sat, 10 September 2011 at 10:59 PM

I am disappointed with the new processors, for me the best processor design was Pentium I MMX, but it doesn't mean that I shall return to use it just because my Athlon II X2 3GHz runs 15 times faster than my old Pentium I MMX 200 MHz.

I make software, I optimize the software in speed. The problem is how to compare the speed of software if everytime you are using a different processor ?

The only way to compare is to have some measurement unit that is independent of the processor you are using. You cannot measure the time because a 3.3 GHz processor always will give less time than a 2.4 GHz processor.

The unit that is independent of the processor you are using and its speed is the processor clock. I measure the speed of a piece of code in number of CPU clocks that takes to execute this piece of code and the result that I get it will not matter if I running Windows 7, XP or Linux.

More than ten years ago I did some software and optimized it to make the critical parts run as faster as I could. I used that time my old Pentium I MMX-200 MHZ. Time has passed, had new computers and the processors improved with new features, 3dNow, SSE, SSE2 and so on.

Two years ago I decided to re-write my old code using new resources of the new processors. SSE/SSE2 are very useful for 3d applications, in this way I could have a piece of code that will be able to run (in clock units) even much more faster than my old code.

My first big surprise was when I measured the speed of my old code running a modern computer. The result should have been less clocks to execute the piece of code, because modern CPU have more execution units, or at least to have the same speed. But it was not, the speed of the modern CPU was much slower executing the same piece of code.

I re-wrote the code using all the power of the new features of modern CPUs, optimised, but I was only able to achieve the same speed (in clock units), but not faster.

In resume, the 2010 piece of code running a 2010 CPU has the same speed as the 2000 code running a 2000 CPU. And I am using now powerful resources that were not available in 2000.

 

The reason of all this I can explain, but it would be a long technical discussion.

Stupidity also evolves!