Published Kal-El (TEGRA3) performance: is NVIDIA SoC truly faster than a Core2 ?

Written by Gionatan Danti on . Posted in Hardware analysis

User Rating:  / 55
PoorBest 

GCC 3.4.6 vs 4.4.4 Core2 T7200 performance

While OOO processors as the Core2 T7200 relieve some instructions scheduling issues, compiler performances remain of utmost importance. So, different compilers can give very different speed results. Not only compiler version, but also compiler settings (eg: optimization level) have a crucial role in determining final results.

Here we can see that GCC 4.4.4 compiler outperform its (very) older 3.4.6 brother:

Core2 T7200 with different GCC versions

At the normal O2 optimization level, the newer GCC give us about 12% better performance: a quite good improvements. However, things become much more interesting when we consider the more aggressive O3 level (the one used by NVIDIA to test its own ARM-based hardware): while GCC 3.4.6 shows no real benefit, the newer 4.4.4 version improve its score by 28%, for a total of +41% over GCC 3.4.6 with the same O3 flag. By using even more aggressive optimization flags, the GCC 4.4.4 score slightly higher, but the improvements are almost intangibles here.

So, we have the proof the, when tested with the same, newer compiler version and the same settings that NVIDIA used to test its ARM-based Tegra2 and Kal-El, the Intel Core2 performances can indeed be higher that GCC 3.4.6 scores.

But, in the end, what is the faster processor here: Kal-El or Core2? Let's use our benchmarked score to compare the Core2 to NVIDIA latest and greatest SoC...

Comments   

 
#1 Sina Saeedi 2012-04-21 18:44
Thank you. It's great
 

You have no rights to post comments