why is my code faster on a Centrino vs 3Ghz P4?

[gjcsot]

I have a bit of benchmarking code which envolves multiple loops. At the core of each loop are some floating point divides, multiplcations and MACs. I have found that the laptops with centrino chip sets are faster than any other processors including fast desktops (3Ghz P4). It doesn't seem to be RAM or cache size that is the problem. What is different about the centrino?

Cheers

Grant

 

[cdogg]

Are you sure it's just with Centrino's? Its architecture is pretty much identical to the P4M, with the exception of integrated wi-fi and some other power-saving features. Have you tried it on a P4M laptop to see if it has just as much success?

From the surface, it wouldn't make sense that your code runs faster. A P4 desktop (especially at 3GHz) outperforms a P4M and Centrino at all levels that I've seen in benchmarks. It might be that your code is not 100% processor dependant and other system factors are coming into play.


~cdogg
[tab]"All paid jobs absorb and degrade the mind";
[tab][tab]- Aristotle
[tab][navy]For general rules and guidelines to get better answers, click here:[/navy] faq219-2884

 

[gjcsot]

I had assummed that the P4M was the same thing as the centrino.

The data in the code is store in arrays one of them being a square array of 512x512 floats which I would assume wqould be store in the RAM so the bus speed is important I guess. But I still don't understand why the P4 3GHz is not faster?

 

[Nelviticus]

Centrino is just a branding thing. It means that the machine has a Pentium M, the 855 chipset and Intel wireless networking. You can see Intel's description here:

http://support.intel.com/support/notebook/centrino/sb/cs-005967-prd955.htm

The Pentium M is a different processor design to the desktop Pentium. It's better at some things and worse at others, which is why most people use a range of real-world tests to benchmark systems.

If you run your benchmarks all they will tell you is which processor is the best at running your benchmarks - that may or may not be useful to you, depending on what you're trying to find out.

Regards

Nelviticus

 

[gjcsot]

It is useful as the benchmark code is exactly what I want to machine to do, I just wanted to understand why!

 

[cdogg]

It's just like it when the P4 first came out at 1.4GHz. Benchmarks at the time showed it ahead in multimedia apps, but actually "slower" than the 1GHz Athlon and PIII in many office apps. This had to do with a difference in architecture (longer pipeline).

Fast forward to your question...

First, I suggest you take a deeper look into the differences between a P4 desktop and an Pentium-M cpu:

http://www20.tomshardware.com/mobile/20030205/centrino-05.html

(I recommend you go from this page back to the beginning of the article and also make sure you read the conclusion)

You'll see that because of the difference in architecture, it's like comparing apples and oranges. Just because a particular piece of code runs faster on one or the other, doesn't really tell you a whole lot as to why. Where the Pentium-M lacks in speed it attempts to make up for efficiency. It's just using a completely different chipset which is not modeled after a desktop or earlier mobile cpus.


~cdogg
[tab]"All paid jobs absorb and degrade the mind";
[tab][tab]- Aristotle
[tab][navy]For general rules and guidelines to get better answers, click here:[/navy] faq219-2884